similarity feature

[kezihlmann]

Semantic Textual Similarity (STS) can be used to analyze text of any two projects and find if they are similar or not. This is then used to update the similarity table and ultimately suggest similar/relevant projects to the user (recommender system)

I want to use the BERT model to do this task (https://towardsdatascience.com/semantic-textual-similarity-83b3ca4a840e)

• implement SBERT cross_encoder
• implement SBERT bi_encoder
• implement TF-IDF
• implement pre-process text
• connect with the azure database
• save and load similarity matrix as json
• tell the database the three most similar projects