Point WordEmbeddingSimilarityIndex documentation to gensim.similarities

[Witiko]

Currently, all documentation, Jupyter notebooks and code examples assume that WordEmbeddingSimilarityIndex is located in gensim.models, although it lives in gensim.similarities since 4.0.0. This pull request fixes the documentation.

[piskvorky]

Thanks @Witiko ! The main place for documentation is the gallery. I fear the notebooks directory is hard to discover, and not really maintained / tested.

Would it make sense to convert this into a gallery tutorial / howto script? With the notebook and html formats auto-generated from the converted source, instructions.

[Witiko]

@piskvorky I like the idea behind the gallery. I don't have the time to convert now, but I will take a look during the next week.

[piskvorky]

@Witiko we're planning the 4.0 release now. Can you please finish this? Thanks.

[Witiko]

@piskvorky Thanks for the reminder. I will give it a look this weekend.

[Witiko]

@piskvorky Would it be OK to merge the SCM notebook with the WMD autoexample? This would rid us of a lot of duplication and there would exist a single lean and mean example for word-embedding-based document similarity measures.

[piskvorky]

Anything that cleans up & simplifies documentation is highly desirable. @mpenkov WDYT?

[mpenkov]

No objections from me.

[Witiko]

In the end, I kept the tutorials separate. Merging the tutorials seemed as a good idea at first, but the preprocessing pipelines for the SCM and the WMD are quite different and the merged tutorial would become significantly more complex as a result.

Perhaps I can give it another stab if we are ever adding sections about WmdSimilarity and SoftCosineSimilarity M:N document similarities, since this will make the autoexamples complex anyways. For now, I added some fixes for the WMD autoexample, such as removing a section about deprecated init_sims method, and redirecting the documentation to the Gallery.

[piskvorky]

Thanks so much @Witiko . Is the failing test related?

[Witiko]

@piskvorky I don't think it is. The failing test is for Doc2Vec, which this PR does not touch.

[mpenkov]

I re-ran the tests, they all passed.

[mpenkov]

Merged, thank you @Witiko !