SymbolicMethods_Project

In this project, we aim to build a research interest summarization tool using LDA. The 'R' packages 'RISMed' and 'textmineR' were used. Process Flow

1. Author name is the input and the package 'RISMed' is used to extract the abstracts published by the author on PubMed.
2. A Document-Term matrix is constructed with each 'document' being a vector a MeSH terms extracted from a given abstract.
3. Each vector of MeSH terms (for a given abstract) is fed as a document to the LDA model.
4. Two approaches have been taken to assign topic names:
   1. By Using the bi- and tri-grams outputted by the LDA model
   2. By using the pre-generated topic labels
5. Fuzzy string match conducted between bi-grams and tri-grams and the long list of all MeSH terms extracted from all abstracts associated with an author to reveal the main topics the research interest corresponds to.