CODEWITHSUNDEEP
machine-learningnlpsentiment-analysis
Fahad
Fahad

Reputation: 378

Features for sentiment analysis using Maxent model

I want to implement my own sentiment analysis using maximum entropy model. without using any Api. what could be the best features f(c,d) for my maximum entropy model. I have three classes positive, negative and neutral

Upvotes: 0

Views: 246

Answers (1)

Kurt Bourbaki
Kurt Bourbaki

Reputation: 12616

Some of the most used and effective features in Sentiment Analysis are unigrams. Bigrams can also be employed, but it is quite controversial whether they are really useful or not.

Note that using frequency values of unigrams/bigrams does not significantly improve results in Sentiment Analysis; it is therefore generally sufficient to extract word types and use a boolean value to express their presence/absence in a text.

The important thing is how you preprocess text before you extract these features. For example, apart from lower-casing your tokens, handling negation scopes can improve your results when extracting unigram features.

In any case, Sentiment Analysis is a wide field. You will find that different feature extraction strategies could yield different results depending on the specific type of analysis you need to perform (e.g. feature-based analysis, subjectivity analysis, polarity analysis, etc.).

You can find almost everything you need to get started here:

  • http://sentiment.christopherpotts.net
  • Liu, Bing. "Sentiment analysis and opinion mining." Synthesis Lectures on Human Language Technologies 5.1 (2012): 1-167.
  • Pang, Bo, and Lillian Lee. "Opinion mining and sentiment analysis." Foundations and trends in information retrieval 2.1-2 (2008): 1-135.

Upvotes: 0

Related Questions