Skip to content

NLP approach

Always start a new project with simple method instead of diving in a deep network straightaway

Heuristics-based Methods

  • i.e. Rule-based approach
  • e.g. Count how many positive & negative words to judge a sentiment of the text.
  • using Wordnet - db of words, and semantic relationships between them
  • using regex (i.e. StandfordCoreNLP , TokensRegex, Pregex)
  • Context-Free-Grammar (CFG) - a formal grammar to model natural lang
    • Earley Parser
    • JAPE (Java Annotation Patterns Engine)
  • Rule based system - General Architecture for Text Engineering (GATE)

Machine Learning Methods

  • Classification
    • Naive Bayes - classic algorithm for classification tasks
    • Support Vector Machine(SVM) - learn a decision boundary to classify classes
  • Regression
  • Hidden Markov Model - a statistical mdoel
    • used for POS tagging
  • Conditional Random Field(CRF)
    • POS tagging

Deep Learning Methods

to be continue....

Comments