Bug fixes

[avinashsai]

#1 Use Infer_vectors:
Doc2Vec model should only be trained using training data and infer_vectors should be used for test
data.Hence modified the model to use only training data.
#2 random.shuffle()
This function shuffles the data given to doc2vec model in every epoch and hence the sentiment labels
will not map to their corresponding vectors. Hence modified get_vectors() function to map labels with
their corresponding vectors.
#3 gensim version updates:
changed epochs=d2v.iter to epochs=epochs=d2v.epochs according to latest gensim version

[ibrahimsharaf]

Hi @avinashsai, thanks for the PR!

Let's split this PR into 3 separate new PRs, that would be better for readability, reviewing and testing.

• 1st one for updating requirements.txt file to the latest versions of needed packages, with needed syntax changes?
• 2nd and 3rd PRs for the first 2 mentioned bugs.

Thanks again for collaborating :)

[ibrahimsharaf]

For the first PR (updating requirements), what do you think about using https://pyup.io/ for dependency updates? I think it's quite cool.

[avinashsai]

Yeah, it's really good. Probably we can consider using it.

[avinashsai]

@ibrahimsharaf What is the next step to proceed further?

[ibrahimsharaf]

@avinashsai the first step is to create a new PR for updating the requirements, using pyup bot, and coding the needed syntax changes.