Question about affixes.json

[pca006132]

1. Some are missing metadata. For example, the -able suffix (B) will not add the adjective property to the result.
2. It seems that a lot of common suffixes are missing. E.g. -ation. Considering there are many different suffixes in the dictionary, it seems that we will eventually run out of single symbol mnemonic. What is the plan for this?
3. Composition of suffixes: Certain suffixes can combine, but only in a certain order. For example, -ize and -ation can combine, but we cannot combine them in the reverse order. I guess the cross_product option is for this, but how can we specify order?
4. It seems that we may be able to automatically remove some entries that are just a derivative of some other entry by searching for the suffix. Any plan on this?

[hippietrail]

1. Some are missing metadata. For example, the -able suffix (B) will not add the adjective property to the result.

It actually goes deeper than this. I've been looking into it a bit and done a tiny bit of work on affixes. First little PR is waiting. But I think it needs the main devs really.

2. It seems that a lot of common suffixes are missing. E.g. -ation. Considering there are many different suffixes in the dictionary, it seems that we will eventually run out of single symbol mnemonic. What is the plan for this?

This is mostly due to the affix engine originally being a way to compress a dictionary so the ones that saved the most space were used. Later on it was borrowed for grammar checking and grammar info started to be added.

Part of the "deeper" I mentioned is that the same affix can have multiple meanings, especially depending on the base word: verb+s = 3rd person singular indicative bount noun+s = plural, noun+'s = possessive but noun+'s also = contraction of noun is and also noun has.

But affixes.json actually does two jobs. Not all of the annotation flags are even about affixes. Some only declare whether a word is a noun, verb, adjecitve, or some combination. Some do both. Some declare the category of the noun before and after affixation. Some declare both some property, such as plural, and also the word category. But 1 means both 'noun' and 'singular noun' whereas '9' means both 'noun' and 'plural noun', etc.

3. Composition of suffixes: Certain suffixes can combine, but only in a certain order. For example, -ize and -ation can combine, but we cannot combine them in the reverse order. I guess the cross_product option is for this, but how can we specify order?

4. It seems that we may be able to automatically remove some entries that are just a derivative of some other entry by searching for the suffix. Any plan on this?

I keep thinking about making a just command for auditing the dictionary. I've got a PR waiting for approval that details all the changes to the dictionary and affixes the last time the curated dictionary got a commit. I've also been working on a parser for the English Wiktionary that should have some use in checking our dictionary and maybe importing to it, etc.

Our dictionary also does nothing for multi-word terms, of which there are several kinds, some of which can have features implemented in the same structure that is now named for its affix component.

[pca006132]

Yes this is my main concern, I feel like the current way of writing dictionary entries will need to change significantly, so I want to ask first before trying to throw in more words into the dictionary.

[hippietrail]

I know what you're saying but I keep adding to the dictionary anyway. I think we can make tools to convert it to whatever new format it will have - probably similar but expanded and cleaner, and also do manual cleanup. I think the main devs have at least some of these problems in mind and is surely following these threads.

[pca006132]

I see. I can also submit some patches to the dictionary, but wouldn't bother too much with finding the correct affixes I guess.

[hippietrail]

I see. I can also submit some patches to the dictionary, but wouldn't bother too much with finding the correct affixes I guess.

Are you familiar with just printaffixes to print out a quick basic guide to affixes and just getforms to expand an entry into all the forms after applying the affix rules? The latter doesn't print out stuff like word category or properties though, yet anyway.

[pca006132]

Will check. Thanks for the pointer.