As discussed in en-wl/wordlist#429, I am now using LLMs to help decide whether or not to add a word. I am also now using LLMs to help with a number of other word related decisions.

The chat interface for size decisions is now available as GPT at https://chatgpt.com/g/g-6972895467f88191b57a7d071a15b5bf-esdb-word-evaluation. The prompts used for that GPT are in eval-gpt.

The rest of this repo contains code to programmatically query LLMs though an API.

The req dir contains the base code for the other LLM tasks. The other tasks are as follows:

• size-score: scripts to ask other LLMs what size a word should be to get a second opinion. It used the same prompts as the GPT.

• size-score-simple: a simplified version of the size-score task

• assign-pos: scripts to ask LLMs what POS a lemma should be assigned

• find-variants: scripts to ask LLMs to identify variants

• plurals: scripts to ask LLMs to classify possible plurals