abhinav-neil / nlp-cord19-research-papers

A jupyter notebook for topic-modelling, clustering and question-answering on COVID-19 research papers.

Search and IR on COVID-19 Research Papers

Perform NLP on COVID-19 research papers to extract useful information.

1. Exploratory Data Analysis (EDA)

   • Find most common words and bigrams in title
   • Topic modelling using Latent Dirichlet Allocation (LDA) and gensim, visualize with pyLDAvis

2. Find similar papers

   • Get embeddings using Universal Sentence Encoder (USE) and find similar titles using cosine similarity

3. Find papers matching query

   • Using cosine similarity & similarity matrix of embeddings

4. Keyword extraction

   • Extact keywords from abstracts using Rake

5. Knowledge graphs

   • Entity detection, dependency parsing, and knowledge graphs from paper abstracts