[WIP] Add doc2vec similarity algorithm

[rock420]

Fixes: #674

Implemented the PV-DM based doc2vec similarity algorithm with cosine similarity as similarity metrics between two documents.

Todo:

• [x] Try different other doc2vec algorithms
• [ ] Try different other similarity metrics
• [x] Tuning different hyperparameters
• [ ] Discussion about the class structure

[rock420]

@marco-c can you guide me on why the docker build is failing?

[rock420]

Tested both the PV-DM and PV-DBOW doc2vec algorithm on the bug data (tuned other hyperparameters also). PV-DBOW is working better compare to PV-DM in terms of both precision and recall. Comparison of the evaluation result (for the same set of hyperparameters and same number of data) -

	PV-DM	PV-DBOW
Recall @ 1:	9.739%	14.285%
Recall @ 5:	10.922%	16.041%
Recall @ 10:	11.005%	16.178%
Precision @ 1:	7.373%	9.896%
Precision @ 5:	2.376%	3.397%
Precision @ 10:	1.339%	1.969%
Recall:	11.038%	16.232%
Precision:	1.983%	2.302%
MAP@k :	8.404%	11.673%

[suhaibmujahid]

Closing for lack of updates, feel free to reopen if you still intend to work on this.