Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8564f81d7a
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| isInflected: /^ξανα/, | ||
| // cf. import {removeAlphabeticDiacritics} from '../text-processors.js'; | ||
| deinflect: (term) => term.replace(/^ξανα/, '').normalize('NFD').replace(/[\u0300-\u036f]/g, ''), |
There was a problem hiding this comment.
Narrow ξανα fallback to the forms it is meant to handle
This fallback rule currently matches every ξανα... verb (/^ξανα/) and strips all diacritics after removing the prefix, so inputs beyond the intended cases (commented as ξαναπάς/ξαναλές) also produce extra candidates like ξαναρώτησε -> ρωτησε. Because these candidates are still tagged as verbs (conditionsOut: ['v']), they can introduce incorrect matches when a dictionary contains accentless entries and always increase lookup fanout for ξανα- terms. Restricting the pattern to the targeted forms (or otherwise gating when deaccenting is applied) would avoid these false deinflections.
Useful? React with 👍 / 👎.
|
As for the codex review, it is true about the fanout, if that means that it will do some extra useless work, but the alternative is a bunch of edge cases (ξαναδεί, ξαναφάς etc.), or being able to know what a syllable is in Greek, which is even more code... It will never be *Edit: it was the copyright year :/ serves me well for not updating my local repo. |
Kuuuube
left a comment
There was a problem hiding this comment.
Idk anything about greek, I'll trust you on this one.
This now works in
wtythanks to yomidevs/wiktionary-to-yomitan@fe99f75 (non-lemmas now have rule identifiers)Closes yomidevs/wiktionary-to-yomitan#295
Before, it wouldn't show any results for anything but ξανανθίζω (because ανθίζω was a lemma)
Now it is capable of ξαναρώτησε > ρώτησε > ρωτάω (in English, re-asked > asked > ask)
demo.webm