search

erickguan / cppjieba_rb

Cppjieba Ruby binding
MIT License
17 stars 3 forks source link

JRuby support using jieba-analysis library for Java #1

Open headius opened 6 years ago

headius commented 6 years ago

cppjieba_rb is a dependency of Discourse, which I am trying to get to run on JRuby. However, JRuby does not support CRuby extensions.

There is a Java library (https://github.com/huaban/jieba-analysis) that could be used from JRuby to provide the same functionality (or at least, I think so...I am not familiar with the jieba library and English documentation online is rather sparse).

Perhaps we can work together to get a jieba library for Ruby that works with both the native extension and the JVM library?

erickguan commented 6 years ago

Do you prefer to replace this gem with another?

Perhaps we can work together to get a jieba library for Ruby that works with both the native extension and the JVM library?

I don't have any experience about that. But sure, we can work on this together and provide you necessary information.

headius commented 6 years ago

The library I pointed at can be used from Ruby directly in JRuby, so it should be possible to mimic the API you've provided in cppjieba without writing any Java code at all.

erickguan commented 6 years ago

It's essentially the same algorithm but with different API.

So they provided with two mode here. We can patch this. For example, :mix can be mapped to SEARCH here. Others can be INDEX. (It doesn't make sense for NLP task we tried to perform)

https://github.com/huaban/jieba-analysis/blob/master/src/main/java/com/huaban/analysis/jieba/JiebaSegmenter.java#L15-L18

jieba-analysis has its own DICT_PATH and a customizable USER_DICT. I think we could change file path here.

Could you go for a PR with these information?