Shanraq.org Shanraq.org
IT

AI without an LLM: a Kazakh text model Lesson 5 of 20

Lesson 5. Build a model you can run by hand

Before trusting a program to analyze words, we should be able to perform the steps ourselves. A manual model is a written procedure that lets any reader reproduce the result. It exposes missing rules.

This text was translated with AI.

Why this matters

Before trusting a program to analyze words, we should be able to perform the steps ourselves. A manual model is a written procedure that lets any reader reproduce the result. It exposes missing rules.

See the whole process

Our dictionary currently has two stems: үй (“house”) and мектеп (“school”). Our teaching table allows only these sequences: stem; stem + plural part; stem + plural part + location part. Here + means “the next part of the analysis”, not arithmetic addition. Use -лер after үй and -тер after мектеп; then add -де. This is a small teaching table, not a complete rule for Kazakh.

Analyze үйлерде: 1) recognize үй; 2) recognize -лер; 3) recognize -де; 4) read “in the houses”. мектептерде gives “in the schools”. For кітаптарда, say “unknown for now”: кітап is absent from our dictionary, even if a person knows the word.

Recall map

Dictionary → find stem → check next part → check order → output an analysis or “unknown”.

Recall without looking

Describe every step for үйлерде. Name a boundary of our table: which stems and endings does it not yet know?

Task

Use the table to analyze үй, үйлер, мектептерде, кітаптарда, and мектеплерде. Give the result or the first step at which checking stops.

Check: үй — house; үйлер — houses; мектептерде — in the schools; кітаптарда — unknown stem; мектеплерде — after the known stem, our table expects -тер, not -лер. We will learn the fuller sound rules later.

If you have found a mistake or a typo in this article, tell us about it

Comments (0)

No comments yet. Be the first.