Human readable, machine processable, ethically collected and shared linguistic data.
Version 0.5.0 Alpha · Windows Downloadgithub.com/smojos/Lemmatrix/releases/latest
What Lemmatrix does
Lemmatrix is a fieldwork database for linguists: instances of real language data called "tokens" are tagged for grammatical, phonological, or other properties, and anchored to "lemmas" (real or hypothetical roots or stems). Tokens can be laid out in a matrix, either lemma vs. tag, or tag vs. tag, so endless paradigms can be produced from a single underlying dataset, facilitating well-organized and flexible data for phonological and grammatical analysis. Utterances and text features allow for syntactic and discourse analyses, and these are seamlessly integrated into the lemma/token data (i.e. lexicon).
LemmaA root or stem, real or hypothetical, with its gloss, word class and your own fields. The lemma list is your lexicon.
TokenOne recorded form of a lemma: its transcriptions (phonetic, phonemic), morpheme glosses, pitch, translation and soundfile.
TagWhat a token is: Plural, 1SG.POSS, Imperative, a frame, a tone context… Tags come from a structure you shape yourself.
Lemma / Tag
Citation
Plural
1SG.POSS
Plural + 1SG.POSS
goat
▶ form · gloss
▶ form
▶ form
▶ form
house
▶ form · gloss
▶ form
▶ form
▶ form
dog
▶ form · gloss
▶ form
▶ form
(not yet recorded)
A Lemma/Tag matrix: each cell is a token you can play, edit or record. Change the rows or columns and the same tokens give another paradigm; save the setups you use, and the gaps show what is still to record.
There are countless additional features to be discovered:
Praat side-loading: open any recording or text in Praat; texts are TextGrids, lined up with their recordings.
Notebooks and session recording: who you worked with, where and when, with a timer and the day's notes.
Comments: a deep, linkable and searchable comment system, with #tags, questions and To Dos.
Hypotheses: link individual data to your theories about the language and grammar.
ɛλicit: request recorded data remotely using a simple link; speakers record on their phone.
Sharing ethically: archives that follow each person's consent, and wordlists (EACW, CAWL, Swadesh) to start from.
Your data stays yours, in plain, readable XML files on your own computer, beside its recordings. Your notebook and hypotheses are your own work: they never go into an archive unless you include them.
Unblock it before unzipping: right-click the zip → Properties → select Unblock → OK. (Or, after unzipping, in PowerShell: Get-ChildItem "C:\Program Files\Lemmatrix" -Recurse -Force | Unblock-File, with your own folder's path.)
Unzip it anywhere: Documents is simplest. Keep the folder as it is.
Double-click Lemmatrix.exe. If Windows says it protected your PC, choose More info → Run anyway. Pin it to Start to keep it handy.
2. Ten minutes with the sample
The welcome window offers the sample database: an invented language with recordings, texts and a week of fieldwork notes (also under Help → Sample databases).
In All data, every token is listed: select one and press Ctrl+Space to hear it.
Open the Lemma/Tag tab and pick one of the saved matrices: a paradigm, ready to listen through.
Open ☰ to reach the other pages: Texts, Utterances, the Notebook, Hypotheses, ɛλicit.
3. Your own database
File → New database: your language and variety, and a wordlist to start from (the EACW, recommended; the CAWL; the Swadesh list) or a blank database.
Set up your tags, then record: each token's soundfile is named for you and kept beside the database.
Keep the database's folder together (it holds the recordings); a OneDrive or Dropbox folder works, so you can carry on at another computer. Every save keeps the previous version in Backups.
Install Praat separately if you use it: Lemmatrix finds it.
4. Help and feedback
Help → I found a tick! reports a bug (a tick is a bug); Help → Request a feature sends an idea.
Help → Check for updates… says when a new version is out: unzip it in place of the old one; your databases and settings stay as they are.
This is an alpha: work on copies of your data, and tell us what goes wrong.
