search

facebookresearch / StarSpace

Learning embeddings for classification, retrieval and ranking.
MIT License
3.94k stars 531 forks source link

python - Information Retrieval #223

Closed ctrado18 closed 5 years ago

ctrado18 commented 5 years ago

Hi,

I have some question to the python wrapper. Is the Information Retrieval part already implemented? As the infos are a bit sparse but it seems that can get now document vectors from a labeled data set? but the test data seems unlabeled. What format I need for labeled documents?

I suppose question and answer need to be trained together somehow, so a docvec seems not enough.

Thanks.

freakeinstein commented 5 years ago

@ctrado18 I believe you are referring to the Utility functions available for StarSpace binary. The python library doesn't bind those utility functions. The utility functions are backed by the StarSpace class, which is actually binded in the python wrapper. So, if you want to achieve those high-level functionalities, you need to implement the same idea in python.

Additional info: Bindings for getNgramVector, printDoc, predictOne are available here