Making Dictionaries Serve Translation

1 Introduction

“What a fine mess you’ve got us into.” (Parrott 1930)

None of the translators I have worked with in Nigeria have known what a difficult task they were letting themselves in for when they first expressed an interest in translating the Bible for their people. What seemed like a relatively simple task (pick up an English Bible, write what it says in our words) rapidly becomes more and more complex. Some are disheartened and others increasingly gripped as they discover the differences between languages, the challenges of translating anything and the particular difficulties all language pioneers face. Although in multilingual societies people are ‘translating’ every day, there’s admittedly a lot of muddle in that communication as people stumble over the mismatches between languages. Written translation (as opposed to immediate oral interpretation) gives us a great opportunity to take time to anticipate problems and to get it right. Unfortunately so much of the time I’ve found translation teams going round and round in circles while increasingly desperate to see progress. Why is this so common? What solace is there? What could we do better?

New strategies, methods, software and training have been suggested,1 but few translators and consultants really expect a dictionary to be much help. Indeed, while in times past missionary linguists would often labour to produce a dictionary, grammar and translation of the scriptures,2 for a variety of reasons dictionaries are no longer routinely produced except possibly as an afterthought once the whole Bible has been ‘finished’. This is unfortunate, not merely for ideological reasons but for profoundly practical ones. We would do well to remember how dictionaries contributed to past Bible translation efforts and consider how (mother-tongue) translators might build them to help in current methodologies of Bible translation.

John Roberts has written a helpful case study of how linguistic study informed the Amele translation (Roberts 2013). Building on this and my own observations, I contend that many problems and inefficiencies faced by Bible translation projects could be addressed by producing and using some kind of dictionary throughout the lifetime of the project.

• I begin by analysing some of the key difficulties present in translation (and checking), with a focus on issues where some kind of dictionary could help.
• Then I examine what kinds of writing support translators need, examining what features of dictionary entries might be most helpful for translators and what common mistakes people make in the construction and application of dictionaries.
• Putting this theory into practical shape, I propose some low-cost, high return ways that translators can begin with the tools they already use (eg Paratext) to start building a useful dictionary, gradually growing into a more sophisticated effort, but always with a focus on investment and immediate return.
• Finally we return to some of the dangers and discouragements that we need to avoid in this enterprise.

My own lexicography and translation experience

I must confess to being one of those strange people who loves reading dictionaries, loves discovering and using new words, delights in groan-inducing wordplay and is frequently caught in the cross-fire between clarity and circumlocution, mired in mixed metaphors. I spent nearly a year collaborating on the C’Lela dictionary in NW Nigeria in 2001 which cured me of any lingering prescriptivist leanings and inspired me to switch my focus from software engineering to Bible translation. Subsequently I have taught dictionary making and computers in linguistics at the Theological College of Northern Nigeria and have taken a keen interest in rapid word collection and software such as WeSay that enables people to collaborate in lexicography. (In 2001 I wrote my own collaboration system to support the C’Lela dictionary development and so I’ve been delighted to see others implement better systems.) Since 2011 I have worked as a translation consultant (in training) with several languages in central/northern Nigeria, with a particular focus on Gworog3 since 2013.

My focus in this paper is on support for undocumented, underdeveloped languages where it is likely that the only significant writing will be a translation of the Bible. It is possible that translators of fairly well developed languages and languages of wider communication (LWC) may be able to gain some insights and ideas, but in such cases I assume that others may already have worked on dictionaries for those languages. The challenge may be encouraging translators to use (and improve) the dictionaries that exist in a careful manner.

1cf BT2013 and other recent papers on accelerating Bible translation, such as Mark Gaddis, Expediting Bible Translation, 2013.

2cf the 18th, 19th and early 20th century translations in Africa, such as mentioned in (‘Robert Moffat (missionary)’ 2015) and (Moffat 1869, 42).

3ISO639:3 KCG-Gworog.

Extracted text from 15dic_Making_Dictionaries_Serve_Translation_150dpi.pdf:

Making dictionaries serve translation David Rowbory, Translation Consultant in Training, SIL Nigeria A paper presented at the 2015 Bible Translation Conference hosted by GIAL Dallas, Texas, 16-20 October relating to the sub-themes Technology and other Tools, Theory & Practice and Translator Training. Abstract John Roberts has lamented the tendency of Bible translators to ignore lexicography until after a New Testament has been completed and printed. The consequence is that while the translation process necessarily reveals much of the lexical richness of a language, few dictionaries are ever /nished and little of the e0ort of creating such a dictionary ends up bene/tting the translation itself. It does tend to be a peculiar minority of people who attack the task of lexicography with relish, but I want to outline the many ways that a working dictionary can and should support better writing. Recent developments have eroded many of the di2culties which have hindered the development and use of dictionaries. There is no need to typeset a full dictionary before it is used; software-based dictionaries can be useful even when incomplete. Rather than throwing knowledge away, every translator or pioneer writer should see the dictionary as a place to store the riches of their language and conserve the fruit of their wrestling with the language. Mother- tongue translators need dictionaries too. Where a diverse range of community members contribute their knowledge of the language to make a good, growing, living dictionary it can provide consultants, reviewers and translators alike with a wider evidence base for their decisions than mere individual opinion. I survey recent developments that make dictionary development more achievable than ever before, and propose procedures for Bible translators to use and maintain a dictionary with examples from projects that have done this. Acknowledgements I am very grateful to all the colleagues and friends who have contributed to my thinking about dictionaries and translation, particularly those in Nigeria. Thanks too to all who attended my presentation at the Bible Translation conference in Dallas in October 2015, asking questions and making suggestions. Of course I must thank my wife Julie who has continually encouraged me as I wrote this paper. 1Making Dictionaries Serve Translation David Rowbory, 2015 1 Introduction What a /ne mess youve got us into. (Parrott 1930) None of the translators I have worked with in Nigeria have known what a di2cult task they were letting themselves in for when they /rst expressed an interest in translating the Bible for their people. What seemed like a relatively simple task (pick up an English Bible, write what it says in our words) rapidly becomes more and more complex. Some are disheartened and others increasingly gripped as they discover the di0erences between languages, the challenges of translating anything and the particular di2culties all language pioneers face. Although in multilingual societies people are translating every day, theres admittedly a lot of muddle in that communication as people stumble over the mismatches between languages. Written translation (as opposed to immediate oral interpretation) gives us a great opportunity to take time to anticipate problems and to get it right. Unfortunately so much of the time Ive found translation teams going round and round in circles while increasingly desperate to see progress. Why is this so common? What solace is there? What could we do better? New strategies, methods, software and training have been suggested,1 but few translators and consultants really expect a dictionary to be much help. Indeed, while in times past missionary linguists would often labour to produce a dictionary, grammar and translation of the scriptures,2 for a variety of reasons dictionaries are no longer routinely produced except possibly as an afterthought once the whole Bible has been /nished. This is unfortunate, not merely for ideological reasons but for profoundly practical ones. We would do well to remember how dictionaries contributed to past Bible translation e0orts and consider how (mother-tongue) translators might build them to help in current methodologies of Bible translation. John Roberts has written a helpful case study of how linguistic study informed the Amele translation (Roberts 2013). Building on this and my own observations, I contend that many problems and ine2ciencies faced by Bible translation projects could be addressed by producing and using some kind of dictionary throughout the lifetime of the project. I begin by analysing some of the key di2culties present in translation (and checking), with a focus on issues where some kind of dictionary could help. Then I examine what kinds of writing support translators need, examining what features of dictionary entries might be most helpful for translators and what common mistakes people make in the construction and application of dictionaries. Putting this theory into practical shape, I propose some low-cost, high return ways that translators can begin with the tools they already use (eg Paratext) to start building a useful dictionary, gradually growing into a more sophisticated e0ort, but always with a focus on investment and immediate return. 1 cf BT2013 and other recent papers on accelerating Bible translation, such as Mark Gaddis, Expediting Bible Translation, 2013. 2 cf the 18th, 19th and early 20th century translations in Africa, such as mentioned in (Robert Mo0at (missionary) 2015) and (Mo0at 1869, 42). 2Making Dictionaries Serve Translation David Rowbory, 2015 Finally we return to some of the dangers and discouragements that we need to avoid in this enterprise. My own lexicography and translation experience I must confess to being one of those strange people who loves reading dictionaries, loves discovering and using new words, delights in groan-inducing wordplay and is frequently caught in the cross-/re between clarity and circumlocution, mired in mixed metaphors. I spent nearly a year collaborating on the CLela dictionary in NW Nigeria in 2001 which cured me of any lingering prescriptivist leanings and inspired me to switch my focus from software engineering to Bible translation. Subsequently I have taught dictionary making and computers in linguistics at the Theological College of Northern Nigeria and have taken a keen interest in rapid word collection and software such as WeSay that enables people to collaborate in lexicography. (In 2001 I wrote my own collaboration system to support the CLela dictionary development and so Ive been delighted to see others implement better systems.) Since 2011 I have worked as a translation consultant (in training) with several languages in central/northern Nigeria, with a particular focus on Gworog3 since 2013. My focus in this paper is on support for undocumented, underdeveloped languages where it is likely that the only signi/cant writing will be a translation of the Bible. It is possible that translators of fairly well developed languages and languages of wider communication (LWC) may be able to gain some insights and ideas, but in such cases I assume that others may already have worked on dictionaries for those languages. The challenge may be encouraging translators to use (and improve) the dictionaries that exist in a careful manner. 2 Analysis: Translation is difcult While theoreticians and new students of translation regularly conclude that translation is impossible (Zweerde 2010, 266) the rest of the world relies on it happening all the time. How do we reconcile such a clash between theory and practice? Its helpful for translators and their associates to have some awareness of the common di2culties translators face so that we can examine what writing tools might be available to help. Building a road while driving along it: translation and language development Pioneering Bible translators4 are frequently doing two jobs at the same time: pioneering a standard written form of their hitherto unwritten language, and actually translating the Bible. Both tasks are formidable. A major pragmatic advantage of keeping everything oral5 is that we avoid the di2culties and confusions of doing both tasks at once. Likewise it is easy to wish that all the di2culties of language development and analysis could be neatly settled before translation begins. However, in many contexts the desire for Bible translation is the major driving force behind people developing their language and local 3 ISO639:3 KCG-Gworog. 4 In this paper I refer to as pioneering Bible translators people who are following the steps of John Wycli0e with the English Bible, being among the /rst to make a concerted e0ort to translate the Bible into their mother tongue and also amongst the /rst to put the common spoken language in written form to that end. Others may build on their work and improve upon it but every language and every translation needs a pioneer to blaze the /rst trail. 5 cf BT2013, BT2015 and other recent papers on oral Bible translation. 3Making Dictionaries Serve Translation David Rowbory, 2015 churches want to see practical, tangible results quickly. So it seems inevitable that in this generation of pioneering Bible translation, translators will be forced to wrestle with the di2culties of writing a language for the /rst time. How do we cope with developing the language whilst also using it in the translation? While translators naturally focus on the di2culties of understanding the original text and /nding a suitable way of expressing the same meaning in their language, it is vital they recognise that many of their frustrations are actually due to the di2culty of an oral language giving birth to a written one. For example, team checking or review of a translation in a language that has not yet been well standardised will often descend into endless debates as to the best way to spell certain words. One person likes to divide words one way, and another person perceives it di0erently. Dialect di0erences cause contention. If translation teams can identify what are primarily language development issues, then they can deal with them appropriately and avoid running out of time to handle the real problems in translating ancient texts across cultures adequately accurately, clearly and naturally. Maintaining a dictionary is time-consuming, but if everyone uses it as framework for developing and recording consensus on language issues, it can be an immediate and enduring saver of time and relationships. No Harry! Dont look at the light!6 Looking beyond the challenges of language development, most translators /nd it far to easy to be /xated by the form of the original, whether following the originals grammatical patterns or discourse structure or falling for faux amis (False Friend 2015). A translator needs to refer to decent dictionaries for both languages to help remind themselves of the meaning both of the original words/phrases and the proposed translation. In a similar way editors can /nd it hard to think of other ways of expressing things once they have been shown one way. A thesaurus (Glock 1976) or a dictionary with growing notes on normal usage ideally based on a corpus of texts (Roberts 2013, 3,67) can help editors to see other ways of expressing things which might be better. Grappling with polysemy, semantic range and ambiguity In every language some words are more Oexible than others. While sometimes it is arguably unnecessary and impossible to distinguish polysemy and homonymy in lexicography (Bergenholtz and Agerbo 2015) it is certainly clear that a single word can have well-distinguished senses or meanings. The context, whether grammatical or lexical, must determine the meaning that writers intend and readers infer otherwise communication will be intolerably ambiguous. Word play (such as punning) may involve some abuse of this principle, misdirecting the hearer or reader before exposing the deception. Sometimes polysemy can be leveraged for rhetorical e0ect. Translators often face the challenge of mismatched polysemy: one source language word may have 3 senses but the target language requires di0erent terms to express each sense. The reverse is also true; a translation equivalent might have more senses than the original word had. A small example can illustrate the di2culties of /nding appropriate matches cross- linguistically. 6 (Lasseter and Stanton 1998) 4Making Dictionaries Serve Translation David Rowbory, 2015 Example: / in James 1 The Greek noun has two entries in BDAG (Bauer 2001, p. 793), which could be glossed as test/trial or temptation (abbreviated by me):

1. an attempt to learn the nature or character of something, test, trial
2. an attempt to make one do something wrong, temptation, enticement to sin. The related verb has 4 senses described in BDAG (again abbreviated by me):
3. make an e0ort to do something, try, attempt (at times in a context indicating futility)
4. endeavour to discover the nature or character of something by testing, try, make trial of, put to the test
5. attempt to entrap through a process of inquiry, test
6. entice to improper behaviour, James 1 appears to play on this polysemy, causing real di2culties for the translator: 2 , , ,... 13
   , . 14

(Count it all joy, my brothers when you face of various kinds... No-one who is tempted should say I am by God; God is by evil, and he no-one. But each is by his own desires, lured and enticed...) Exacerbating the problem for English translations, test and tempt have other senses beyond what is meant by as these (abbreviated) dictionary entries of the verbs illustrate (Stevenson 2014, tempt, test): tempt: (v)

1. entice or try to entice (someone) to do something that they )nd attractive but know to be wrong or unwise
2. be tempted: have an urge or inclination to do something
3. persuade someone to do something
4. archaic: risk provoking a deity usually with undesirable consequences. test: (v) Generally: take measures to check the quality, performance or reliability of something
5. give someone a short examination of their pro/ciency or knowledge
6. judge or measure pro/ciency/knowledge by means of a test
7. reveal the strengths or capabilities of someone/something...
8. carry out a medical test
9. [with no object], produce a speci/ed result in a medical test...
10. Chemistry: examine a substance by means of a reagent
11. touch or taste something to check it is acceptable before proceeding further. 5Making Dictionaries Serve Translation David Rowbory, 2015 The senses in heavy type are those I consider to correspond most appropriately to the apparent senses of / in the original. Simply using either test or tempt concordantly to represent / in James 1:2-14 will end up being internally inconsistent and incoherent.7 The list of senses and de/nitions helps to reveal the similarities and di0erences between the English and Greek terms in a way that otherwise might not be very clear. A Venn diagram (see right) can roughly illustrate the semantic overlap and disjoints, where the semantic range is represented with partially overlapping circles. Some senses of test, tempt (and also try) seem to lie within the range of senses of /, but several do not, and there are also aspects of the Greek terms which are not conveyed by any of those English words. A Greek-English dictionary helps us to identify English words that can express the meaning of / and thus to internalise something of the way those Greek words describe reality. It is helpful to see the distinct senses for the words but this is not enough support for the translator. There are two bene/ts an English (target language) dictionary brings.
12. The English dictionary clearly shows no overlap between test and tempt and translators are forced to distinguish between the senses when translating /. The term evil in James 1:13 clearly indicates in that context the enticement to bad action sense is in view, rather than the sense of learning the nature of something.
13. It also reveals that these translation equivalents themselves carry various senses which are not part of /. Having found an apparent match, the translator is often tempted to disregard the mismatches.8 Most current Bible translators have access to a dictionary explaining the Biblical Languages via English, yet no clear way of analysing and comparing the senses of similar words in their own languages. This presents us with a very real danger that translators using the Biblical languages will be skewed by the English (or other LWC) analysis undermining full and careful use of their own language resources. 7 I once unwittingly but memorably provided a similar example when speaking via interpreter in Tanzania, comparing and contrasting grace and mercy. When the interpreter represented both as rehema the rhetoric was ruined and communication confused. 8 In the case above, given the context is clearly neither medicine or chemistry, I consider it unlikely that anyone would misunderstand test but given the widespread use in English of a glib and non-negative form of tempt to mean being persuaded or attracted to something (tempt.3) we may want to watch out for possible misunderstandings in that way. 6Making Dictionaries Serve Translation David Rowbory, 2015 Collocations Since words do not normally exist in isolation, it is important to understand the normal patterns that govern their usage. Some phrases, such as father and son and bread and butter have a normal order or common words which collocate, or are often found together. Translating the component words individually may lead to a stilted translation. On the other hand, writers may deliberately break the usual order to surprise the reader, such as a man and his father... in Amos 2:7. If we represent an unnatural or marked collocation with an unmarked, or ordinary collocation, then we lose an aspect of the communication. So, as well as de/nitions, notes about normal usage and common collocations are helpful, both for source language dictionaries and target- language dictionaries. The Collins COBUILD dictionaries (The History of COBUILD 2015) pioneered a corpus- driven approach to English language dictionaries for learners of English, showing the natural and common patterns of language use. A strong motivating factor was the realisation that a dictionary which treats words in isolation does the translator or the learner of a new language a disservice. Idioms, usage Idioms, in the sense of a group of words established by usage as having a meaning not deducible from those of the individual words (Stevenson 2014, idiom) cause translators problems (Barnwell 1986, 16) because idiom conventions vary between language. Since idioms are typically brittle if one word is changed, the special meaning of the idiom will be lost on the listener. On the other hand, if a translation accidentally invokes an idiom where none is intended (such as Psalm 1 stands in the way of sinners or Deuteronomy 11:19 ...when you are walking by the way, and when you lie down... [ESV]) it is distracting for the reader. Moving between languages the translator may forget the natural patterns of usage, and so need the support of a dictionary which notes idioms with their meanings. Connotations & Associations Just as idioms leverage conventional patterns to give special, particular meanings not deducible from their components, words in each language will have associations with other words and wider connotations which cannot be easily predicted. It is very helpful for the translator to be reminded about the negative, positive or other connotations and associations a word has, since this helps pre-empt unfortunate choices of words. For example, in English woman used as a direct address as in John 2, seems unspeakably rude, particularly given the situation of Jesus addressing his own mother. To describe the word woman without noting the cultural and emotive connotations of certain usage is unhelpful. When reviewers or hearers object to certain translations, sometimes it may be because of unfortunate associations. Decision-making is tiring Making decisions is a very tiring exercise (Tierney 2011) and the more considerations that need to be held in the balance mentally, the more exhausting. Translation involves many tricky decisions and while the very mentally agile may be able to carry out a good deal of translation without any external aid, the likelihood is that tired translators and consultants take short-cuts. The result may be unnaturally concordant translation, verbosity, a lack of clarity or a blindness to possible misunderstandings. 7Making Dictionaries Serve Translation David Rowbory, 2015 3 What kinds of writing support do translators need? I recently was catching up with a translator9 who has struggled for some years to make progress translating the Bible. I asked him what he thought made the work hard and progress slow. To my surprise one of his /rst answers was the lack of any kind of dictionary of his language. He is not alone. John Roberts (Roberts 2013, 54,55,65) gives several examples where the Amele translators needed the support of a dictionary to properly understand the English source language and consider the best translation options available to them. Without the record of careful discussions, their mother tongue intuition alone was insu2cient. In some situations the translator is always surrounded by a community ready and willing to discuss language issues with them, but often this is not terribly easy. David Francis, the Nikyob translator, noted several lacks:
14. without a dictionary he found it hard to know the best ways of using the language (including coping with dialect variation),
15. without a standardised written form its hard to write anything down and be con/dent of it being understood well,
16. when there is uncertainty about Biblical terms he is unsure whether he is capturing the meaning as it should be from the source,
17. similarly reviewers are at a loss to comment beyond giving their personal preferences about what the translation should be. In short, without any writing support for the translators it seems the Nikyob translators were Ooundering and expending e0ort while making little real progress. Whenever they did make some progress, the underlying thinking and discussion was liable to be forgotten. And when experienced translators die without recording some of their accumulated knowledge in some useful form, they take with them to the grave the fruit of many di2cult and necessary discussions. Since languages of wider communication are frequently languages of wider confusion, it is very dangerous to assume that a translator is actually understanding what is really intended by the English versions they read, even when the translator relies extensively on English and appears to function well in it. In my experience in Nigeria, most people assume a one-to-one equivalence between words in di0erent languages, with the semantic categories being laid down by the underlying mother tongue. In younger generations there has been more widespread but often more super/cial education, with non-standard or idiosyncratic understandings of English being passed down from teacher to pupil. While a new Nigerian English may be emerging from this morass, in the interim it leaves translators often thinking they understand their reference materials better than they really do. So clearly some writing support needs to challenge writers assumptions about the language of wider communication, whether highlighting the polysemy, semantic range, collocations or idioms, or even expressing the connotations of particular words. 9 September 2015. Conversation with David Francis, from the Nikyob community in Kaduna State, Nigeria. 8Making Dictionaries Serve Translation David Rowbory, 2015 As already described above, while the driving force for much language development may be the translation of the Bible, this means that translators are often the /rst people to seriously wrestle with spelling, word division, grammatical norms and dialect di0erences. The wise translator needs a good way of keeping track of every pertinent discovery so as not to be either overwhelmed or (more likely) to forget it and miss the opportunity to develop the language well. However, for a variety of reasons, the kind of dictionaries most helpful for translators are seldom developed before or even during a Bible translation programme. We discuss below some of the reasons for this, then outline a number of ways in which a growing dictionary can bene/t a Bible translation. Why dictionaries are often ignored (making them and using them) Dictionaries are certainly ignored in terms of production and use. There are some good reasons for this. Well formatted, printed, general-purpose dictionaries take an extraordinary long time to prepare and cost a lot to print; they are a risky endeavour. Few are the people, and somewhat peculiar, who genuinely love the tedium of exhaustive documentation especially when the most work may be exerted on obscure entries that quite possibly no-one will read. It is good to admit this. Dictionaries undoubtedly can have many uses, but many preconceptions can undermine the practical utility of dictionaries. Some people envisage a dictionary as a status symbol for their language and in that case a terse word list would do without requiring anywhere near the time or money required for a full dictionary; it need not even be carefully checked if its purpose is just to sit on a shelf looking impressive.10 Beyond the status symbol, it is understandable but unfortunate that people often focus on the least imaginative uses of a dictionary and so given the e0ort required to produce such a work, consider it of little importance. Most mature readers would only turn to a dictionary when coming across esoteric, rare or technical words. Younger readers or readers of a foreign language will have much more cause to turn to a dictionary. Writers and editors may imagine the main use of a dictionary is to ensure correct spelling, especially when orthography has diverged from phonology to the extent of English. But if words are written as they sound then who needs a dictionary? And then some entertaining souls swallow a dictionary to impress their listeners, such as the honourable and notable Patrick Obahiagbon, a Nigerian Member of the House of Representatives:11 Every word I use is in the dictionary but the only di$erence is that I use a multiplicity of dictionaries. My elliptic and polyvalent modus of communication has been something that I have acclimatised myself to for donkey-years and it would be unfair to me to say that it is just razzmatazz... This is my logomachy. 10 A well-intentioned translator once showed me his surprisingly large dictionary of a Nigerian language. It consisted of a standard English dictionary with his Nigerian language equivalents written beside the English headword. This too would take a modest time to complete and would yield an impressive looking tome with numerous senses and well-edited de/nitions but is undermined by being one of the most misleading and dangerous approaches to lexicography. In essence we had a word list masquerading as a dictionary. The de/nitions were only appropriate for the English headword, 11 Quote from youtube.com/watch?v=kBZn2pg2k78, see also youtube.com/watch?v=OF0Re7gJBb8 youtube.com/watch?v=96A4ZctoBdw 9Making Dictionaries Serve Translation David Rowbory, 2015 Such behaviour is so markedly at odds with the ideals of clear and natural Bible translation that such (ab)use of dictionaries may give them a bad name. Most Bible translators (and consultants) are less interested in building a linguistic museum of curiosities than helping themselves and others wield the language well. The helpful lexicographer will identify obscure or technical words to help an honest user avoid unwitting abuse of the dictionary in this way. This is particularly important when we have jargon such as novel Biblical terms included in a dictionary which may be only understood well by a subset of the community. A general purpose dictionary may be deadly dull, irrelevant to most users or may try illegitimately to /t the language into some ideal pattern perhaps following the language of wider communication. Searching through a dictionary may be intimidating and frustrating for highly inOectional languages or root-based languages such as Arabic and Hebrew. There are many reasons to ignore dictionaries and rely on mother tongue intuition and the mental dictionaries that everyone has. Finally, the busy writer perhaps feels under too much time pressure to consult a work of reference. Or perhaps the dictionary is theoretically available but practically hard to access. We should not downplay the challenges of building and using dictionaries, but recognise that there are many ways to dodge the di2culties and bene/t. Spell checking Checking spelling is one of the most frequently cited, yet least imaginative reasons for using a dictionary. There are two sides of spell-checking: discovering the canonical spelling and verifying spelling that the writer believes to be correct. Even experienced writers may be uncertain of the canonical spelling of a word especially a rare or commonly-confused word and so may want to look it up when drafting and proo/ng. Even experienced writers make typing errors to the annoyance or confusion of the reader; it is considerate to remove every such error that may hinder comprehension. With paper dictionaries spell checking is a time-consuming process. Computers are usually much better than people at checking spelling, providing you supply them with all the correct rules. If there is no printed dictionary then the pioneering Bible translator has no authority to refer to and so spelling may never actually be checked. Translators are often in a rush to just get something translated, but simply writing the way something sounds often turns out to be much harder than people expect. Sounds are used di0erently in di0erent languages and often there are several reasonable-looking ways of dividing words and combining letters to convey the sounds as well as meaningful aspects such as vowel length, tone and intonation which wont necessarily have a self- evident representation. A concern for consistent spelling and use of punctuation may chieOy have developed with printing technology, but it is genuinely helpful to give readers and advisors/consultants more certainty of what the translation is saying and make reading less arduous. So automated spell checking and quick veri/cation is a genuine purpose for target- language dictionaries, but there are some major discouragements, challenges and dangers we must avoid. These are noted in section There is also much more that a dictionary can o0er to the translator. 10Making Dictionaries Serve Translation David Rowbory, 2015 Glossary & Biblical terms Technical writers are concerned about using technical terms correctly. The Bible is a technical work, with specialist vocabulary from many areas of life and incorporating concepts and terminology from di0erent cultures. Translators need to keep track of their standard ways of translating technical terms, to help ensure consistency and to save wasting time repeatedly wrestling to /nd a good translation. Many translation projects have kept a note of Key Biblical Terms, often with a back translation into a language of wider communication, for the sake of consultants. Some translators produce a glossary to better help readers and new translators understand the Biblical concept being communicated. Both the list of key terms, functioning as a translation memory or reference list and the glossary are kinds of dictionaries and helpful in their own way, but they focus on the meaning of the biblical terms rather than the meaning of the translation equivalents. In searching for good translations of biblical and technical terms, the Bible translator may need to become an expert in many domains out of their immediate experience. He may need to research the terms used by blacksmiths, potters, builders, farmers and soldiers, amongst others.12 A loyal translator-lexicographer will serve his community well by documenting and sharing all he learns; a dictionary should be a storehouse of a communitys linguistic treasure. However, alongside harvesting specialist jargon from their community, a translator who is keenly aware of his audience needs to document the less precise or less technical ways ordinary people may talk about specialised industries. If he documents and labels both the technical and the informal terms and senses in a dictionary then he provides a reference for reviewers to consult when they are ensuring that the translation communicates accurately and clearly to the target audience. A highly technical translation that ordinary people do not understand (without looking up a dictionary) may be impressive without being helpful. Beyond mere glosses: de.nitions, grammatical notes, usage, idioms Given the challenges of translation, a dictionary which is really useful for writers and translators must at a minimum move beyond glosses to de/nitions and indicate di0erent senses of a word. As we have already seen we can in some places add extra value by giving grammatical and usage notes, identifying who understands the word(s), and listing idioms which use the headword. A wordlist with simple glosses reinforces the mistaken notion of one-to-one correspondence between languages. However glosses are attractive because they e2ciently get you very close to the meaning. The astute dictionary user knows that a simple one word gloss is misleading. I think a good compromise is to use both a gloss and a clarifying de/nition. Some illustrative examples from Hebrew, English and Hausa We already saw above that a uniform gloss of / as tempt or test is misleading because both the Greek and English words have a variety of di0erent senses. 12 cf. much biblical imagery, such as military: Ephesians 6:130; construction: Psalm 118:22; pottery: Isaiah 29:16, 41:25, 64:8, Jeremiah 18:4,6, Romans 9:21. See also Mark Sweetings paper/poster The Cornerstone: A Biblical Term of Foundational Signi)cance to be presented at BT2015. 11Making Dictionaries Serve Translation David Rowbory, 2015 However, even giving di0erent glosses for di0erent senses may still be misleading, because semantic boundaries are drawn di0erently, with some words being much more broad in one language than the gloss is in another language. For example, we could gloss the Hebrew word /yad/ as hand, but while English hand means usually the end part of a person's arm beyond the wrist, including the palm, /ngers, and thumb (Stevenson 2014, hand) the Hebrew term has a much broader sense. So a more helpful entry in a Hebrew-English dictionary might read: noun:
18. hand, arm; the upper limb from shoulder to /ngers, not just the palm and back of the hand as in English. Then some extended or /gurative senses would be helpfully listed, such as:
19. side, part...
20. strength... Similarly a Hausa dictionary is misleading if it simply de/nes hannu as hand (Bargery and Westermann 1951), since a better de/nition would likely resemble that of Hebrew above, though with di0erent extended//gurative senses:
21. hand, arm; the upper limb of a person from shoulder to /ngers.
22. trunk (of an elephant)
23. sleeve (of a shirt) In each case, the gloss does help the reader get a quick grasp of the sense of the word, but the de/nition shows clearly where the gloss falls short. Again, instead of merely being satis/ed with glossing /ets/ as tree, we would serve the translator better if we made it clear how the word is actually used in Hebrew, and how the gloss may be misleading. English maintains a clear distinction between a living tree and (felled) logs//rewood/wood. Hausa has a term itace which corresponds to better than English, but in the secondary sense has some di0erences which are worth pointing out. So a Hebrew/English/Hausa references might resemble this: Hebrew
24. tree Hausa: itace-1; a large plant with a strong central stem (trunk) that bears fruit or seed, larger than a vegetable plant
25. wood, timber Hausa itace-2; but in the Bible cut wood in this sense is more often for building materials or sacri/ce than just ordinary /rewood. 12Making Dictionaries Serve Translation David Rowbory, 2015 The secondary sense for Hausa itace is generally wood for burning (/rewood) whereas it seems that the secondary sense for is more likely to be wood for building (timber). Very common words like heart and bitter are very easily misunderstood because the English words used as glosses di0er markedly in their /gurative and idiomatic senses, though the original or basic sense may correspond well. A dictionary entry for the Hebrew term should point out that it is not used /guratively about emotions (as English heart is) but about decision-making. Similarly bitter has a primary sense connected with taste, but the English /gurative sense of bitter implies holding a grudge against someone while the Hebrew /gurative sense is more about intensity of su0ering. What might a translation-aware dictionary entry look like? It should be clear that not every kind of dictionary will serve translation. Some example entries may give some ideas about what is helpful and what falls short. From the 2001 CLela dictionary (Rikoto 2002) here are three main entries, chosen fairly arbitrarily: The CLela dictionary intended for print publication from the outset began with a trilingual wordlist (CLela with English and Hausa glosses), gaining IPA pronunciations and example sentences later. In the editing phase, as duplicates were eliminated, the editors tried to understand the varying senses of words which seemed to have several possible glosses and uses. While the source material often began with a gloss, we realised that this was sometimes misleading, and so we sought to expand glosses such as pack to distinguish related words. Given that CLela grammar distinguishes nouns and verbs, we separated homonyms for doko from di0erent senses of doko2, based on the part of speech. The entry for the noun doko3 gave both the more technical term biceps but also a non-technical description. With the bene/t of hindsight and with electronic publishing for translators in view, there are several de/ciencies:
26. There were no example sentences because the example sentences in the source material were not su2ciently illustrative and spelling and grammar checking all of the example sentences in the database took too long.
27. The entries lack any semantic domain marking, or explicit links to related words which might be helpful for translators. The English reverse /nder list did at least 13Making Dictionaries Serve Translation David Rowbory, 2015 group together other words for pack etc, but direct links to related words would be much better, to enable translators to consider better alternatives.
28. The lengthy secondary sense for dedga appears to be somewhat contradictory, perhaps suggesting more senses that need disambiguation and illustration. In the Gworog WeSay dictionary, I added the following entry during translation checking: jyam verb De)nition: roam; wander around aimlessly Glosses: roam, wander Note: DR: September 2015 team discussed whether this was appropriate for 'prowl' in 1 Peter 5:8 but decided that there "wurug kh" was more appropriate. In section 4 I will go on to explain how translators and consultants can work towards making similar entries as they discuss words in the translation. Eventually this minimal entry could be expanded, but for now it records something useful we have discovered about the word, even though it was eventually deemed inappropriate for the translation of 1 Peter. It would be good to identify whether it can be used equally for animals and humans, but that distinction may not be relevant for the Bible translation. An example dictionary speci.cally designed for Bible translators The UBS Flora and Fauna of the Bible (UBS 1980) is so richly descriptive as to border on being an encyclopedia. It supplies a good model for translator-friendly dictionaries in anticipating misunderstandings or misconceptions people may bring to each term. So the Translation section for bush includes this warning, particularly pertinent to Nigeria: Unfortunately, many languages do not have a generic word for bush or shrub; they only have tree and grass. In West African English the word bush refers to an uninhabited wilderness area. Every year people burn the bush, and in a number of places translators have mistakenly used this concept in their translations of this passage and in Deuteronomy 33:16 where him that dwelt in the bush is particularly confusing to many West Africans. In the absence of a word for shrub, it may be necessary to use tree. Such a pattern could be helpfully followed for target language dictionaries. 4 Suggested ways forward The busy Bible translator, advisor or consultant can very easily be overwhelmed with all the best practice advice which is available. We may be tempted to ignore it all or be totally distracted spending endless time trying to come to grips with every new idea. But instead I would like to present a gradual way forward that aims to add as little cost (in time) as possible for maximum immediate gain. The more elaborate the system we try to adopt, the greater the risk that it will all be a monumental waste of time and e0ort. There are already some relatively simple tools within the grasp of most Bible translators and assistants. Since most Bible translators I work with have access to the joint UBS/SIL Paratext software, I will concentrate on that and on free software tools. (Obviously 14Making Dictionaries Serve Translation David Rowbory, 2015 without Paratext some steps would be di0erent, but very similar results can be achieved using standard free software.) Overview Paratexts Biblical terms tool gives a simple structure for building a useful dictionary, though with some signi/cant limitations we should not ignore. It is a good way to start accounting for polysemy and with the glossary tool (from Paratext 7.5 and later) it provides a good area to practice writing de/nitions in the style of a monolingual dictionary. Since all information is shared via Send/Receive this is an easy way for translators and advisors to build their knowledge of the language as it applies to biblical terms. The Wordlist tool allows us to approve correct spellings, then check the spelling of existing translation and new drafting. This is helpful, but has limitations and some dangers if the orthography is still unstable. The Project interlinearizer (File menu) keeps a basic lexicon of glosses to re/ne its guesswork about what the translated words mean. Often di0erent glosses will reOect di0erent senses of a word. Finally, Paratext is being increasingly integrated with FieldWorks. Any user can look up a word from Paratext in the FieldWorks lexicon, building and re/ning the dictionary entry in FieldWorks, with all of the distinct senses and notes that anyone has added. Paratext Biblical Terms As already mentioned, many translators of technical writing /nd it helpful to keep a reference of technical terms and past translations.13 Paratexts Biblical Terms tool is a very good way to start developing a useful dictionary, gaining some immediate bene/t and practising techniques which can then be used for the language in general. There are three possible tasks associated with each Biblical Term:
29. Listing valid translation options (taking account of polysemy of source and target language).
30. Verifying that the actual translation uses the agreed renderings appropriately.
31. Building a glossary to help readers of the translation. Begin with names, for simplicitys sake Select the Major Biblical Terms list14, limit the range of terms to just one book and then focus on the names. The names are the easiest place to start and the major decision you need to make is about the spelling/transliteration of each name.15 This is a translation policy decision which the translation committee or team need to agree on and document at some point in a translation brief. The options are generally to begin either with the Greek/Hebrew, or with a language of wider communication, then adjust in some ways to 13 See translation memory products such as www.wordfast.com/clients_translators etc 14 Major Biblical Terms is a hand-picked list of biblical terms drawn from a number of sources. There are arguably many omissions, but its a good place to start. 15 If quite a lot of translation has already been drafted then you can Guess Renderings to get Paratext to make an educated guess about how the names are being spelled in the translation. In this case the renderings will be added in red. Its probably best to do this for names only rather than other kinds of biblical term as the guesswork becomes increasingly fraught with di2culty for more Oexible terms. 15Making Dictionaries Serve Translation David Rowbory, 2015 suit the phonology of the target language. So what kind of information do we need to record about these names?
32. The spelling of the name.
33. A note about the source and the basis for the spelling.
34. Optionally:
35. some glossary information to help readers understand anything which is signi/cant about the meaning of the name or the identities of the people with this name.
36. a list of common mistakes in spelling the name, or old standards.
37. discussion about other options for spelling the name which are under consideration, but not currently standard. Within the Biblical Terms tool we double-click on a row to open the editing window. Enter the standard transliteration in the Rendering /eld. Then I recommend putting additional information in the Notes16 /eld. Many translators concentrate only on the /rst item, recording the spelling of the name, but if we start to build more information about the entry we can prevent confusion and wasted time later on. The most basic information to add in the Notes for a name is the rationale for the transliteration, such as following Hausa or following English, or possibly explaining a phonological constraint. To give an example, churches have been established in Gworog-speaking lands in central Nigeria for nearly a century but the formal Bible translation project is quite new. They apparently have an established history of rendering the name Joseph as Isuwu. This is unlike both Hausa and English, the languages of wider communication. Given that the normal translation policy was to follow the familiar Hausa forms, I was surprised at the rendering and asked about it. Then, in the notes /eld I added this explanation together with anything that I was uncertain of: 16 Notes may be titled Description from Paratext version 7.6. 16Making Dictionaries Serve Translation David Rowbory, 2015 Of course such information should also be in a consultant report, but by adding it to the Biblical Terms tool it is more obviously available to any other consultant, advisor, theological student or new translator who does not know the reason for the rendering. Going further Once you have established a pattern giving the rendering and an explanation in the notes you can start working on more complex terms that have multiple senses, such as
    lord. The Greek term has several senses which will not necessarily correspond directly to the target language. The Good News and NIV use the following renderings: lord, Lord, sir, master since the English terms have a narrower semantic range than the Greek. That is English uses di0erent words to convey the appropriate meaning in context. In Paratext 7.5 or earlier, di0erent renderings are separated by commas, but from 7.6 you enter each on a new line. Paratext conveniently allows you to type a note or gloss in (brackets) afterwards. So in Gworog we might have the display as below: Making the most of notes/description The Notes area of the biblical term window is a freeform area to describe the term. The lack of structure makes it simpler in some ways, but to make the most of the tool it may be best for a translation team to develop a standard pattern for the description, possibly along these lines:
38. De/nitions.
39. Mismatches or concerns about the rendering(s)
40. Misspellings
41. Alternatives either rejected or under consideration If we place a gloss for each biblical term in (brackets) following it, then in the notes box we should add a fuller de/nition, making clear the signi/cance of the di0erent senses. 17Making Dictionaries Serve Translation David Rowbory, 2015 This is a good place to identify semantic mismatches or possible misunderstandings. Considering the examples outlined earlier, it is very unlikely that any rendering will be an exact match for the original term. If we understand where our rendering falls short and how it may be misunderstood, then we can more easily check that it does still function satisfactorily in context. Misspellings are bound to occur in the future, and if we clearly identify them now then we can save time wondering what the correct spelling is. Finally it is often necessary to consider several possible renderings. As a translator builds an entry in Biblical Terms it would be wise to consider other terms that may seem similar, and to account for the inevitable mismatches between the semantics of the biblical term and the standard renderings. If we only ever mention the current standard then no-one reading the entry knows whether an alternative has been considered. When a rendering has been found to be unhelpful and has been replaced by another rendering, we should explain the di0erences between all of the alternatives. This may help prevent regression17 to a less satisfactory term. Comparing and contrasting similar terms in the target language is a necessary part of the process of /nding the most appropriate standard translation. Sometimes a translator may think of a new rendering which seems superior to the current agreed standard. In a team project, it would be inappropriate for a single team member to change the rendering without discussion, but to avoid losing the insight, the translator could add the proposal to the Notes box. However, the Notes box is perhaps a clumsy place to try to store too much information. Eventually it would be good to document the words used in the renderings so that you approach the translation from the other angle, just as we saw with the test/tempt/ example from James 1 earlier. Translators or other advisors may create dictionary entries (say in a WeSay of FieldWorks dictionary) for all of the candidate renderings, their component words and related words. Making use of History Paratext 7.5 and later prompt18 the user to explain any change in the Biblical Terms rendering through the History button. This can warn a team away from chasing their tail, unwittingly changing back and forth between alternative renderings. Its a suitable place to note brieOy the motivation for a particular change. Later anyone can look at the record to understand what changes were made in the past, whether wise or not.19 This can complement a comment in the Note /eld that explains why the agreed rendering should be used instead of anything else. Glossary entries Paratext 7.4 and later includes a very helpful Glossary tab in the Biblical terms tool. Glossaries are really specialised mini-dictionaries which focus on explaining terms used in the Bible which may be unfamiliar to or misunderstood by some readers. While the 17 Regression in software engineering describes the surprisingly common situation of old errors recurring after having been apparently /xed. 18 with a Oashing exclamation mark sign 19 This kind of chronological, change-focussed record-keeping is very helpful in the software development world where it is equally possible to go round and round in circles, especially in a team environment. 18Making Dictionaries Serve Translation David Rowbory, 2015 information in the Notes box should usually be written in a language of wider communication, so that all translators, advisors and consultants have access to the information, the glossary entries should be written in the same target language as the translation. Somewhat confusingly the glossary is really a kind of a mono-lingual dictionary or encyclopedia rather than simply a list of glosses. So if translators think of a dictionary as a word list, they may be uncertain how to write a glossary entry, and may be tempted simply to translate something from a resource. This would be a major mistake. Glossary entries can compensate for the mismatches between the source and target languages, helping the reader to avoid drawing any unhelpful inferences due to di0erences in culture and semantics. For this reason the entries should probably contain some negative statements. Caution: some drawbacks We need to remember throughout that the Biblical Terms tool is organised according to the original source language terms. Because semantic range, polysemy and usage vary between languages we must keep reminding ourselves that you cannot reverse a bilingual dictionary and end up with anything useful. The Biblical Terms tool is really a dictionary of Hebrew/Greek with glosses and de/nitions in the translation language. Paratext Spell checking Paratext keeps a Wordlist,20 all the words actually used in a translation, so that you can mark each spelling as correct or wrong. Once you have enough words marked good or bad (rather than the default of unknown) then Paratext adds a bit of statistical guesswork and o0ers two di0erent kinds of automated spell check: step-by-step or wiggly underlines. This is clearly the best way to satisfy the need for consistent spelling and picking up typing errors, but using these tools well is much harder for pioneering Bible translators than many people expect. We will save the pitfalls to avoid for the next section, but concentrate here on what I consider a prudent way of working through the word list and then transitioning to spell checking. First we need to assess how stable the orthography seems to be. Apart from just reading through the translated text, Paratexts Wordlist has several ways of identifying instability in the writing system. Start with the mistakes that translators and checkers can identify, then ask What general problem might be underlying these speci/c mistakes? For example, if there are more vowel symbols in the target language (eg ) than in the languages of wider communication (LWC), then people may forget to use them. Is there a transition primer to help people transition into literacy from a LWC? That may give a clue about likely mistakes. If there is no such guide then the team need to make one. Use the Word Listss Similar Words tool to get a sense for the scale of di2culties. 20 In the Tools menu 19Making Dictionaries Serve Translation David Rowbory, 2015 For example, in Gworog several vowels are commonly confused and other special characters are not used consistently. If we use the settings as below, then a long list of potentially confused words is generated: Then we need to to work out which are erroneous. Out of context it can be very di2cult to know what these words actually mean, so clicking on each word in turn shows in the panel below where the words occurs in the translation work done so far. This can help us to understand the variation. In the example above, we can see the signs of an orthography change; the // symbol was adopted instead of /ng/ but because it would be dangerous to replace all /ng/ with // there are still many words with the old spelling. Are words divided chaotically? 20Making Dictionaries Serve Translation David Rowbory, 2015 Use the Wordlist tool Find Incorrectly Joined or Split Words. This will group together words which all contain the same letters but are sometimes divided di0erently. In the example above, we can see that an initial /n/ sometimes is written on the preposition /ba/ in because of phrase level phonological rules. ba nshyo and ban shyo di0er only in word division. Pedantic non-native speakers (such as translation advisors or consultants) may be much better than native speakers at identifying inconsistency. If the orthography is as Ouid as the Gworog currently is, hasty automated spell checking will be of little help, probably will cause damage and certainly will waste considerable time. If the orthography is still quite young I would recommend that the team checks spelling as a team and only approves the easy, obvious words /rst, and only where there is clear consensus. 21Making Dictionaries Serve Translation David Rowbory, 2015 When the orthography has stabilised then: Work through all of the words in the wordlist either correct or wrong (and supply the correction). Note: It may be more encouraging and manageable to do this for only one book of the Bible at a time. Use the Wordlist View Set Scripture Range command to do this. Then, having clicked on the project window, turn on the Checking Display Spelling option. In the drafting window, Paratext will add red (and grey) wiggly underlines to words which have not been approved. Right-click on such a questionable word in order to change the spelling to a similar word, or mark the spelling as OK. Clearly this only makes sense if there are very clear guidelines for deciding what the canonical spelling is. If it is just a matter of opinion and preference then this tool will waste time. While this can be done individually, it is important to ensure that every individual is implementing consensus decisions about spelling. Spell checking in other software If you have developed an extensive list of approved words in Paratext, it is possible to transform that list and use it with other software such as LibreO2ce21 and Microsoft O2ce. It is also possible to produce a spell checking list from WeSay or another dictionary to use in Word Processing and layout software apart from Paratext, whether 21 Tools Spelling and Grammar... Options... New... to create a new dictionary then look in Tools Options (or on a Mac look in LibreO>ce Preferences...) and look at the LibreO2ce Paths section to /nd where the dictionary /les are stored. They are just plain UTF-8 text /les with a list of correctly-spelled words sorted in alphabetical order. You can move the approved words from Paratext by opening the Wordlist and exporting as XML, then massaging the words into a plain text /le. 22Making Dictionaries Serve Translation David Rowbory, 2015 through a series of manual conversions of the dictionary /le or by writing software to do it automatically. But unfortunately there is not space or time to describe the process here. Interlinearizer Paratexts Project Interlinearizer (opened from the File menu) begins with statistical guesswork to generate guesses of the meaning of words in a translation based on a reference text. (Sometimes the results are astonishingly good and other times, as you would expect, they are nonsense.) As you teach the interlinearizer correct glosses it uses them in the future in place of the guesses. Helpfully, you can also supply both a phrase level and morpheme level translation. Paratext displays all the glosses for a particular word when you click on the current gloss. Complementing the work done in the Biblical Terms tool, the interlinearizer can help us work towards a bilingual dictionary of the target language as we build a bilingual word list. We should expect a mismatch in semantic range and senses between the translation language and the glossing language which will likely give us an opportunity to account for di0erent senses of the target language words. For example the Gworog lyyad has the basic sense of word and an extended sense of issues. Quite likely the word lyyad will have a variety of other uses which are not yet accounted for in the interlinearizer. Note also all the red guesses, which generally are irrelevant. The Gworog pronoun nyin also appears to have a variety of glosses: your.pl, you.pl and yours.pl. This is not because of polysemy so much as grammatical di0erences with English. It is not the form of the pronoun but the word order that determines whether it is in a possessive relationship with other words. (Since Gworog distinguishes between second person singular and plural, to avoid confusion I have made the plural explicit in the gloss, hence you.pl. The gloss you should really be replaced by you.pl.) Help in nding ambiguous homographs, wrong words or misspellings Gworog uses tone extensively to mark lexical and grammatical distinctions but the Latin- based writing system is still developing and given that the Latin alphabetic is not very suitable for marking such distinctions we have quite a number of tone-di0erentiated ambiguous homographs. Several other sound distinctions are not being di0erentiated in writing reliably yet so this leads to other situations where the wrong spelling is used. The interlinearizer can help reveal these occurrences because it deals purely with the orthographic form of the text. For example in two occurrences shown below, the word nnay (long initial consonant) is properly the number 4. However, it may be wrongly written as nay which instead means request. When glossing, the error should be obvious. (My convention is to pre/x the gloss of each misspelled word with *.) 23Making Dictionaries Serve Translation David Rowbory, 2015

Idioms and gurative language A lively translation will endeavour to use idioms in the target language where appropriate and following the usual patterns of the language. However idioms can pose problems during checking because the literal meaning of the individual words is at odds with the combined meaning of the idiom. It will help translators and consultants if we capture both the meanings of the constituent words and the actual message conveyed by the whole idiom. Record the real meaning of idioms by linking together the constituent words to gloss it as a phrase. It may help to identify such entries by a pre/x such as idiom:. (Later, if we migrate the interlinear lexicon to WeSay or FieldWorks, then we can search for such pre/xes on phrases and /gure out how to expand the entry most helpfully.) Similarly, if we have a word glossed as seed in the text it may be helpful to identify the actual plain meaning at the phrase level as descendent. Alternatively, add a gloss marked with a pre/x such as )g:, ie. /g:descendent. Moving interlinearizer data into other software to develop a richer dictionary The list of translations (most likely glosses) is stored in the project folder, inside the Lexicon.xml /le. As a translation consultant I have found it helpful to browse the gloss list interactively.22 Others have wondered whether they could move these data into WeSay or FieldWorks. This is possible but it is not without its problems. A simplistic method is:

1. Use my glossy tool to display the whole lexicon in a table in a web browser
2. Edit Select All the whole text, Edit Copy
3. Paste it into a new spreadsheet starting from row 2.
4. In row one title the /rst column \lx (for lexeme) and the second \gl (for gloss).
5. Use SheetSwiper (Coretta 2015) to export SFM and import into FieldWorks. Note that you can only do this once before you have to worry about creating duplicate entries. However, for users with a well developed interlinearizer lexicon, this is a good way to leverage that investment and develop it further as a dictionary. Having migrated your glosses into FieldWorks you will need to do considerable editing. It can be helpful to search for all the words which share a gloss and then compare and contrast these words. 22 Software to do this is available here http://blog.rowbory.co.uk/sw/glossy 24Making Dictionaries Serve Translation David Rowbory, 2015 FieldWorks integration From Paratext 7.5 (and improved and simpli/ed in later versions) you can link a translation project to a FieldWorks (dictionary) project. Right-click on any word in a project and Find in Dictionary.... If the word is not yet in the FieldWorks lexicon then you will have an opportunity to create an entry, with a basic gloss. If it is in the lexicon then you can read the dictionary entry, and click Open Lexicon if you need to change anything: It would be worthwhile Opening the Lexicon to develop the initial terse gloss into a full de/nition, separating senses as appropriate and adding any notes on usage etc that may be helpful. So in the example above, you can see what this might look like for de/nitions of two senses of the Gworog word kh. If semantic domain information is present in the lexicon (for example from a Rapid Word Collection exercise) then we can right-click and Find Related Words..., using the lexicon as a thesaurus. This might help remind a translator of a more distinctive or appropriate word for that part of the translation. How dictionaries .t into the translation work6ow: At each point in the translation workOow the translation team should be
6. looking at what is currently written about the words they are using, and
7. considering what extra information could be kept in a dictionary for later reference. Setting up a translation project to use a dictionary It may be helpful to set up a WeSay dictionary or a FieldWorks project on an administrator computer, then use Send/Receive to transfer that initial dictionary to other translators computers. (FieldWorks is best for integration with Paratext, but WeSay is 25Making Dictionaries Serve Translation David Rowbory, 2015 much less overwhelming. For looking up words in WeSay we need to use the Dictionary Browse & Edit button.) While drafting The individual (or team) drafting may want to check the spelling of a word and check whether they are using it correctly and if there is any uncertainty about dialect variation. If at all uncertain about the teams standard translation of a biblical term, then they should check the biblical terms tool. Whenever the drafter encounters an unfamiliar English (or LWC) word in the model or reference text, he should look it up in the dictionary in case it has an entry. If it is not there then he will need to check an English dictionary, ensure he understands the word well, then add an appropriate entry to the target language dictionary to help other translators. While proof-reading Solitary proof-reading may rarely be done in practice, but unless following a team- drafting approach it is often ideal for a drafter to proof-read their work before troubling others with it. The drafter needs to make sure they have followed the spelling and word division conventions, but also they need to ask themselves whether the translation uses the language appropriately. Here a growing dictionary can help with recalling appropriate spelling and word division, but also the drafter can look up some notes made in a dictionary about usage. While team-checking When a translation team has a discussion about the best words to use, they should make sure all words/phrases are in the dictionary not only the words which are chosen to be used in the translation but also those that were rejected. Then the team also need to check that the dictionary entries all reOect the teams understanding of the language and ensure that the evidence from the dictionary would support their decision for the translation. If there are debates about spelling or dialect, the team should ensure that all evidence is stored in the dictionary. This may mean adding notes (as in English dictionaries which mention American and British usages) explaining which words are used by which dialect. The team should also check the biblical terms tool to ensure both that there is an appropriate agreed rendering recorded, with appropriate explanatory notes and also that the translation is using the renderings as agreed. Common mis-spellings should be recorded in the dictionary notes /eld to help later writers avoid that mistake and to make them more con/dent of the correct spelling. During back-translation and consultant check A linguistically-aware back translator or supervisor, or a consultant can add appropriate glosses in the project interlinearizer. Where there is uncertainty it might be a good opportunity to Find in Dictionary... and ensure that relevant material from the consultant check and the consultant report are included in the dictionary. After testing and in review sessions During testing and review, it is especially important for translators and testers to note misunderstandings that occur, either when people are reading or listening. These may be evidence of ambiguous homographs (possibly due to under-representation of tone) or 26Making Dictionaries Serve Translation David Rowbory, 2015 unnatural usage or a confusion over what sense is intended, where words have more than one sense. All of this should be recorded in the dictionary to help translators in the future become aware of potential misunderstandings. It may be that usage notes are required, or a note saying not to be confused with... just as modern English dictionaries sometimes need to distinguish biennial and biannual, disinterested and uninterested. When enough work is done in the biblical terms tool, we could import the terms into WeSay, however this is not necessarily an easy task. It may be simpler and more reliable to add dictionary entries for the biblical terms into WeSay manually. Other dictionary development methods It is quite possible and even desirable for dictionaries to develop with several methodologies at once, as long as everyone understands what is happening. This paper has focussed on some ways in which translators and consultants can contribute to a dictionary in the course of their work on the translated text, but Rapid Word Collection and natural text analysis are helpful complementary methods. Rapid Word Collection (http://rapidwords.net) can be an e0ective way to engage whole communities in the building of a dictionary. Some semantic domains (such as social, familial, religious) may be very helpful for translators. The most domains that translators /nd most helpful often, in my experience, seem to be the ones that people /nd easier to generate words for. Beginning a RWC exercise with some semantic domains closely related to the Bible translation work may be a good way of motivating translators to be involved in building a dictionary. The WeSay software encourages de/nitions to be collected for words, rather than just glosses. This will make for a more useful dictionary entry. Since words are collected together with other related words, it should be natural to compare, contrast and document what is distinctive about each word rather than, say, simply glossing 4 di0erent words as shout or relative. Natural text collection and analysis (from /ctional stories, historical narratives and other speeches) gives a valuable insight into the natural patterns of language usage and example sentences drawn from natural texts may provide a helpful reference point for translators. There is a long tradition of dictionaries (such as the Oxford English Dictionary) relying on quotations from texts to evince de/nitions. More recently computerised corpus-based dictionaries have helped ensure that dictionaries reOect language as it actually is used rather than the idealised form people sometimes imagine. (Nabirye and Schryver 2012) Publishing In recent years the options for publishing dictionaries have multiplied considerably. Even as recently as 2001 when I participated in the CLela dictionary project, Unicode was still fairly new and barely supported by the dictionary development tools available, mobile telephones were not a viable target, laptops were expensive, tablets were virtually unknown, the internet was usually inaccessible from locations where Bible translation was going on, and Paratext was limited to use by consultants and typesetters. While a print dictionary may be desirable for status reasons and for simplicity of distribution, it is very time-consulting and expensive to produce a print dictionary, slow 27Making Dictionaries Serve Translation David Rowbory, 2015 to update and it is slow to look up random entries. Just as most Bible translators now carry a library of reference works within Paratext and other software on their computer, it is much better for translators to carry the dictionary electronically. Paratexts integration with FieldWorks and the ability of FieldWorks to Send/Receive with WeSay (through USB or internet) may be su2cient for translators and consultants to make use of and improve a dictionary. It is unfortunate that there are no user roles and permissions for this yet, as there are for the translated texts in Paratext. So every user added to the project must be trusted to edit carefully. It is probably prudent to back up the FieldWorks database from time to time in case one user makes a catastrophic mistake. If we want to open up the dictionary to wider access, outside of the Bible translation team, then other software is more helpful. (Links are given in the Key Software section before the main Bibliography.) Publishing your dictionary on the web is good for wide, public distribution, especially where internet access is cheap and reliable. Webonary takes FieldWorks databases and builds a dictionary website around the popular content management system WordPress. Webonary encourages publishing early and often (New Dictionaries on Webonary | Webonary.org 2015). It can cope with updates (without producing duplicates) as long as any dictionary entries that have changed are appropriately date-stamped. In many parts of Nigeria however, web access is still very limited and unreliable and so Lexique Pro is a great way to install a searchable and browsable electronic dictionary on any Windows computer, with easy cross-references, links, pictures and audio. The Dictionary App Builder (in testing now, for Android devices) is an e0ort to put dictionaries on mobile phones. In these cases the convenience of using the dictionary ofine is o0set by it being slower to update dictionaries than with FieldWorks/WeSay or Webonary. Pathway provides more electronic publishing options, including e-Readers such as Kindle. Again these avoid the need to be constantly online, but then it will be harder for everyone to keep up to date as the dictionary develops. For those who want to collaborate on developing their dictionary online, whether with Rapid Word Collection or editing, Language Forge is developing into a powerful web- based environment to do this. There are now so many di0erent ways to distribute a dictionary electronically that it would be foolish not to take advantage of them. It may be that the only dictionaries worth printing are limited semantic domain literacy books, such as a mini dictionary of animals, or social terms etc. 5 Steering away from some dangers and discouragements My approach here is purposely pragmatic. Many lovers of words and lovers of dictionaries expect everyone else to love the lexicon as much as they do. Lexicographers understandable enthusiasm for dictionaries may blind them from the humbling reality that most people /nd dictionaries exceedingly dull or overwhelming. But we need not 28Making Dictionaries Serve Translation David Rowbory, 2015 persuade everyone that a dictionary is a worthy end in itself.23 I simply want to suggest some easy ways in which writers, editors and advisors can store and refer to information about a language. Objection: Wont maintaining and using a dictionary slow down the translation? It certainly could slow down and utterly distract the translators unless they remember that the translation is what is primary and the dictionary is merely a place for recording and organising and cross-checking discoveries about the language. Only if translators can use and build the dictionary easily and e0ectively will it be worth the time invested in the exercise. That is, if translators dont refer to the dictionary or abuse it in some way, then we would be better without it. Objection: Shouldnt a dictionary be made and language fully analysed before translation begins? Given that language documentation (dictionary and grammar) were for many years standard pre-requisites for SIL and other organisations translating the Bible, it is understandable that /eld linguists may be worried by quick-and-dirty approaches to linguistics (Unger 2013). The haphazard approach of building a dictionary (and doing language analysis) in parallel with the translation may indeed look foolish. However, local Christian believers are very frequently the instigators of much Bible translation and they often begin with little initial appetite for academic study whose point is not immediately clear, as Unger and others on the Linguistics Discussion List bear witness. Rather than making a dictionary an annoying distraction that outside funders impose, this paper argues that we can achieve similar or better input by collecting dictionary-type data as translation work proceeds. The likely alternative to a dictionary-as-we-go approach is no dictionary at all, or in a few cases a dictionary created by outsiders which the translators ignore. Worse, translators may grudgingly build a dictionary which is really just a list of words and glosses,24 then slavishly adhere to it and ignore the rich description that a dictionary should furnish. Given the advantages that we have shown a dictionary can bring even to a Ouent translator it is desirable not just for someone to build a dictionary, but for translators to buy in, reaping the bene/ts as immediately as possible. Warning: Status and ownership In many Nigerian cultures people are much more likely to share possessions than hard- won or exclusive knowledge (cf Maranz 1997). This kind of cultural background may quietly undermine e0orts to collect and to share a store of language wealth freely within the community. For this reason I think it is more helpful for a dictionary to be from the start clearly framed as a collaborative e0ort which is used and relied on by everyone who is contributing to it, rather than a status symbol to be exploited by any particular individual. 23 Even the translation of the Bible is not an end in itself but a means to translating a knowledge of God into the understanding of people who dont read Hebrew, Aramaic and Greek. 24 For example the Comparative African Word List, which is more useful for phonological study than semantics. 29Making Dictionaries Serve Translation David Rowbory, 2015 Warning: Dictionaries are dangerous and thesauri treacherous Dictionaries often get a bad name because users abuse them, leading to obscure or stilted communication, slavishly concordant translation or any number of sins. As long as a language is alive and developing, a dictionary is a work in progress and all the more so for the kind of dictionary development I am advocating here. So everyone who contributes to and uses the dictionary must remember that the dictionary is not an unchallengeable authority, but a work in progress: There may be obvious errors or omissions; they should be corrected. There may be confusion or disagreement; more evidence needs to be sought to improve the entry. Di8culties with the interlinearizer Good use of the interlinearizer requires linguistic awareness. Nave bilinguals and well- trained translators will sometimes back-translate the translation into natural English (or another LWC) phrase by phrase and then use the LWC word order when glossing. This wrecks the lexicon! The interlinearizer is not the place for such a free translation. We need to account for the contribution each word makes to the whole sentence. Some words are not easily glossed because they perform more of a grammatical function. For such words you have a choice between giving a functional gloss (eg PAST, CONDITION, RELATIVISER, END-OF-RELATIVE-PHRASE or an abbreviated version) or simply giving the most appropriate English/LWC gloss in each situation. My preference leans towards the former, since this helps to avoid illegitimately supposing the target language to follow the grammar of the LWC. A massive list of wildly divergent alternative glosses is a tell-tale sign of something being wrong in the way you use the interlinearizer. Finally, when glossing in the interlinearizer it may be apparent that the spelling or word division of the word in the translation is wrong. This poses a di2cult problem, because we may want to gloss the word, but without perpetuating spelling errors. I think it best to:
8. Make a note of the spelling error in the text so that translators can correct it.
9. Enter the appropriate gloss, but with a * pre/x. This is a common way in linguistic writing of identifying something as being incorrect or defective. Later we can search the glosses for this * marker and attempt to correct the entry. In the example right the article ku is incorrectly joined to the end-of-relative-phrase marker ni. This is a very frequent error. Unfortunately when the spelling error in the translation is corrected the gloss will disappear, unless the corrected spelling is already in the lexicon. But on the other hand we do hope that as errors are corrected the lexicon will gradually improve. Di8culties: Unstable orthography If the orthography is still not certain when you are adding entries, then a dictionary will quickly gain both variant spellings of the one word and apparent homographs which actually conceal a di0erence in sound. (For example in Gworog as written in 2015 30Making Dictionaries Serve Translation David Rowbory, 2015 mbiyang/mbyyang/mbyang are all well attested variants of the one word meaning and.) This is bound to cause confusion. However, rather than waiting until the orthography completely stabilises before we work on the dictionary, we need some way to tame the chaos. Here are some suggestions: When you come across several