Capturing a Lexical Item (Kpelegbe)

In this blog post, I walk through the steps that I have been following in capturing lexical items for the Kpelegbe dictionary.

1. Finding

There are many different sources for lexical items. The only constraint is your imagination. There should be no dogmatic constraints on the process. There are no “rules” of fieldwork, only ideas that work, and ideas that don’t work so well.

One great source of new lexical items is having causal conversations with your consultants and people in the village. If they use a word that you have never heard before, ask them what it means. In general, a great (undervalued) source of information on a language is just trying to learn how to speak the language and taking note of the kinds of interesting things that people say. 

Another great source of information is the consultants themselves. If they know how to write, buy them each a notebook and pens, and ask them to keep an eye out for new lexical items. I have already got a few dozen words this way in less than two weeks of working on the dictionary this summer. 

Yet a third source is going to some location and asking the consultants to name the objects in that location. For example, there is a kitchen right next to my office space (on the porch of a house). We just turned our heads toward the kitchen and started naming objects. I got quite a few new words in this way. The same strategy can be used almost anywhere. 

A great strategy is to search for words by semantic domain. One day, I asked the group of lady consultants to name things that women wear. Right off the bat we got about 10 new lexical items, without batting an eye. It was quick and painless and even fun. 

2. Transcribing

Segmental transcription of Kpelegbe words is not especially challenging. I already know the segmental inventory from earlier work, and there are no very challenging distinctions (unlike for the Khoisan languages, where different click releases can be very difficult to distinguish).

But for tonal transcription, Kpelegbe is often challenging. There are three level tones: H, M, L. While M is not fully distinctive (there are no L versus M minimal pairs), it varies in interesting ways and needs to be carefully distinguished from L and H. Also, there are lots of floating tones in Kpelegbe (and Ewe more generally) and all of these need to be transcribed. Often getting an accurate tonal transcription for a word involves a careful series of questions.

One method that I have found particularly useful in working with Kpelegbe (and Khoisan) is the tone matching test. If you know the tone pattern of X, and suspect that it is the same as the tone patter of Y, ask your consultant to compare X and Y and see if they match. Some consultants can do this well. For the Kpelegbe consultants, I do this via whistling: whistle X and then whistle Y, are they the same? The consultants have a very sharp ear for this, and can distinguish such close cases as H.M versus H.L easily. 

For the tone matching tests, it helps to develop of small list of words where you are absolutely certain of the tonal transcriptions. Then you can keep coming back to this list, and the consultants become comfortable and very efficient in the task.

I have also developed a set of linguistic contexts that help me transcribe tone. For example, if I suspect that X (ending in a mid-tone) has a floating H tone, I will put it in the context of a following mid-tone. The word ta “head” is mid-tone in isolation, but has a floating high tone. If you put it in the context of a following mid-tone, the tone of ta “head” shifts to MH (because the floating H tone docks to the left). Just a few simple contexts like this help me a lot. 

3. Translating

My dictionary is based on translations of Kpelegbe words into English and French. In this dictionary, I am not giving Kpelegbe definitions to Kpelegbe words. I may do this in the future as a long-term project, but for now, I want to get a tri-lingual dictionary into the hands of readers as soon as possible.

To my great surprise, I have found that Gemini is a great source of information for these translations. To begin, Gemini can easily do English-French translations at a very high level of accuracy. If you tell Gemini the context (Togo, west Africa), it can even give translations that take that into account. The greatest surprise for me has been how Gemini can help me with animal and bird and plant names. It is sometimes the case that the consultants know the Kpelegbe name for an animal or bird or plant, but have no idea what the French is (and they do not generally speak English either). In this case, I give Gemini a complete description of the object according to the consultants. Surprisingly, Gemini is able to come up with a good hypothesis as to the name of the object in English and French. If I can then verify this name, using one of my books, I am fairly confident that I have the right translation.

4. Examples

Especially for verbs, it is impossible to understand a lexical item without accompanying examples. The goal should be an example for each lexical item, and sometimes two or three. The easiest way to get great examples is to get the consultants to generate them. That is, they learn the skill of generating natural examples on their own for particular lexical items. This is not a translation task. I do not present any sentence to them in French beforehand. Rather, they learn the skill of generating completely new and interesting examples. I have also used this strategy quite a bit with the Sasi dictionary

Another great source of examples is oral texts. I populated a large portion of the Sasi dictionary with examples using the oral texts that I had transcribed in FLEx. The fact that the dictionary was also in FLEx made the process almost effortless. I do have a number of oral texts for Kpelegbe, but I gathered them in the 90s and have not yet transported them into FLEx. I intend to do this during the academic year 2026-2027.

5. Verifying

An essential part of the process is verification. If I gather a lexical item with one of the teams, then I need to verify that lexical item with another team. I have fixed many errors with this strategy.

The number one source of confusion for the consultants is the influence of other dialects. Every Togolese Ewe speaker needs to master three dialects of Ewe: (a) their village dialect, (b) standard Ewe (for church and school), and (c) Gengbe, the lingua franca of Togo. These three dialects are all very close to one another, and people often mix them together when communicating on a day-to-day basis. Since I am trying to be a purist about using Kpelegbe in the dictionary, it is important to try to filter out non-Kpelegbe forms. Verifying with another team is an excellent strategy.

6. Recording

Every word of the Kpelegbe dictionary is recorded with at least two men and two women. I had the same minimal standard with the Sasi dictionary. The reason for this is to get a good idea of the phonetic variation in the pronunciation of particular lexical items. Another reason is to have a record for the future in case I want to change some aspect of the transcription. A third reason is to set myself up for future electronic versions of the dictionary, where people will be able to listen to the lexical items on their smart phones.

All of my recordings are done using a Zoom H4n as an audio interface to my computer. Each word is repeated three times (in isolation). Often, there is an example sentence for the word that is recorded too. If I have any questions about tonal transcription, I often record the word in a series of linguistic contexts meant to bring out tonal properties.

Because of the number of speakers and the different example sentences, the number of recordings is much greater than the number of lexical items. For example, for the first two weeks I worked with Kpelegbe, I gathered 191 new lexical items, but I have 1,131 sound files for them.

7. FLEx

In the end, all lexical items must be entered into FLEx. The final printed (and published) dictionary will be created from FLEx, using an SIL program call Pathway (with the help of Zach). Using FLEx greatly reduces the amount of effort that it takes to make a dictionary. It has all kinds of built in features that facilitate the task. 

Here is an example of a lexical entry from the Kpelegbe dictionary, as seen in the “Entry” window of FLEx:

edru [L.LH H] n. yam mound (Fr. butte d’igname) Meƒo dru alɔfaɖeka. ‘I dug one hundred yam mounds.”

This example shows that the word edru has the tonal contour L.LH with a floating H tone. Its part of speech is noun. The English translation is “yam mound”. The French translation is “butte d’igname”. There is also an example (translated into English, but not glossed) to show how it is used.