0

I would like to build a clean "dictionary" (or some sort of "base word form" collection) for various languages. I am used to English, or even Chinese, because it is analytic and especially for Chinese, each word is an atomic unit. But in agglutinative languages like Turkish or Inuktitut, I don't know what you should put into this thing.

In Inuktitut, we have:

ailiruk
Go get it (command)

aivviaqtuq
walrus hunting (he has gone...)

alianaigusuttuq
he/she is having fun doing something 

These are sentences.

Turkish seems to have an infinite amount of words, such as:

muvaffakiyetsizleştiricileştiriveremeyebileceklerimizdenmişsinizcesine
As though you are from those whom we may not be able to easily make into a maker of unsuccessful ones.

I can't find any better examples off the top of my head, but basically, words can be complete sentences.

So what do you include in the dictionary in these cases, because these types of things don't seem to fit into the same "atomic" category as words like you'd think of them in Chinese. What is the general best practice here?

5
  • This is a decision that needs to be made by an individual lexicographer. And in case it needs to be said—you generally need to really understand a language to make a dictionary for it.
    – Draconis
    Mar 23 at 16:25
  • Turkish, like other agglutinative languages, cannot in practice create an infinite number of words by agglutination alone – there are a finite number of affixes, and affix recursion is not generally allowed (except for certain subtypes, like negations, which are recursive in English as well, e.g., ‘un-un-un-un-un-un-un-un-undo’). Many non-agglutinative languages (including most Germanic ones), however, can create infinite words from compounding (e.g., Swedish nordöstersjökustartilleriflygspaningssimulatoranläggningsmaterielunderhållsuppföljningssystemdiskussionsinläggsförberedelsearbete). Mar 23 at 18:21
  • Also, Eskimo languages can do much better than those Inuktitut examples you cite. My West Greenlandic grammar has anigugassaajunnaangajalluinnarsimassagaluartut ‘…that they must really almost have become unavoidable, but…’, for example. Mar 23 at 18:25
  • 1
    You might find this article interesting; it's an overview of Inuktitut aimed at non-linguists building computer systems to deal with it, and covers the sorts of information you'd want from a good Inuktitut dictionary.
    – Draconis
    Mar 24 at 0:26
  • Jerry Sadock (who majored in Eskimo while attending college in Denmark, before he became a linguist) says that there are many recursive morphological structures in Inuktitut, and therefore the lexicon is countably infinite, since there's no real distinction between word and sentence. So get counting, Eskimo lexicographers. You've got a lot on your plate
    – jlawler
    Mar 24 at 19:28
2

I assume you mean "electronic dictionary", and I assume infinite programming skills, plus access to recording facilities and speakers.

Such a dictionary needs to be paired with a grammar (and, obviously, a grammatical analysis), and an annotated text corpus of some sort. What a dictionary does is distill out the essential unpredictable properties of individual lexical items. The English word "cat" is pretty simple in terms of form – the singular and plural are completely predictable (read the grammar), there are no case forms (read the grammar). "Child" is a bit more complicated so you have to list the plural. Since this dictionary is not exclusively about word forms, you may also assemble idiomatic usages of "cat" including special meanings in compounds, for example "catwalk", or "child" (as in "child node"). Consult the OED for zillions of examples of that type.

In the complete dictionary of Logoori, you would list the (formal) causative of "buy" because it actually means "sell", and is semantically unpredictable compares to the formal causative of "plow" which is "make a person plow". There would be information on transitivity and stativity because "boil" doesn't distinguish "I boiled the water" vs. "The water boiled" or "The water is boiled", and different roots interact with the tense system differently to yield an interpretation – you need to include that kind of information. You need to indicate how the "perfective" stem is formed since that is not entirely predicable from the usual citation forms. It would take years to enumerate all of the forms that would have to be listed in the perfect dictionary, so I'll cut this short.

Semantic information is particularly difficult, since 1 or 2 word English translations are usually very inadequate. Pairing lexical entries with actual usage (from the corpus) is ideal, and why I stipulated electronic and programming skills. Providing all of the desired information for the root -gáraba is TMI (pages and pages for "bean leaf"), but with an electronic resource you can structure the presentation of information, suppressing the dozen corpus examples until it's relevant.

Usually, dictionaries are organized around the language equivalent of "noun, adjective, verb" (the major lexical classes), with "preposition" having an uneasy place (can't you just put them in the grammar? Yes in Bantu usually, no in English). One does not ordinarily list grammatical morphemes in a dictionary, for example "mʊ-. Cl. 1 [sg], plural usu. Cl. 2". But in an electronic dictionary this can also be listed, and arguably should be available (revealed when you click "more").

As a toy example, you could start to sketch a Swahili dictionary, because there are already such dictionaries, grammars, and speakers, and the level of complexity of the language is relatively low (minimal arbitrariness, not as ludicrously prolific in word-formation as some Bantu languages or the languages of the Eskimo family).

The usual problem in dictionary construction, in terms of knowing what to include, is that you do not know what to include. I recently learned that "e" in Turkish spelling is ambiguous between two vowels, so a dictionary needs to indicate whether that vowel is open mid or close mid. If you leave that information out, you have a less-than-perfect dictionary. If you have links to pronunciations by speakers, you at least in principle have supplied the necessary information (which the Turkish government dictionary does). Most dictionaries of Bantu languages are less than perfect because they omit tone. So there is no general and precise answer to the question "what should I include", but the direction of the answer is "Anything that cannot be provided by rule", and "anything needed to prove the analysis. With my proposed toy dictionary, you can discover in one instance what should be included.

1
  • A good example of a dictionary (print, though, not electronic) is the Hopi language dictionary that Ken Hill produced. It uses a lot of print technology to bring out the character of a highly inflected
    – jlawler
    Mar 24 at 19:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.