IMG_20180428_104348 The three pillars of Natural Language Processing
From Becky's presentation, here are the three cornerstone approaches: sentiment analysis, topic modeling, and summarization.
Sentiment analysis quantifies the subjective "emotion" in the text. It is commonly rule-based, but can also use knowledge base, Latent Semantic Analysis (LSA), and Support Vector Machines (SVM).
Topic modeling finds abstract concepts that occur in the body of texts (also known as corpus). It commonly uses Latent Dirichlet Allocation (LDA) and Latent Semantic Analysis (LSA), but can also use Non-Negative Matrix Factorization (NMF).
Summarization reduces a text to several key phrases or a representative sentence. It commonly uses TextRank and LexRank, but can also use knowledge base / knowledge graphs.