Sentiment analysis is “the process of computationally identifying and categorizing opinions expressed in a piece of text, especially in order to determine whether the writer’s attitude towards a particular topic, product, etc. is positive, negative, or neutral.”
On the CorText platform, we offer two different algorithms to perform sentiment analysis:
- Google Natural Language API (reference needed) enables a user to measure the sentiment polarity and magnitude of a text . A normalized version of magnitude is also computed. Polarity spreads from -10 to +10. Magnitude from 0 to 10 (only integer values)
- Textblob is a python library (Textblob Documentation) computing sentiment score for a textual snippet as well as its degree of subjectivity. Sentiment varies from -10 (very negative) to 10 (very positive). Subjectivity is 0 for factual statement and 10 for very opinionated claims.
The results are stored in the database for further analysis.
Sentiment analysis works in English or French (only Textblob is allowed in that case).
Italian language is also being tested. Only polarity score is issued (no subjectivity scoring)
Google API is limited to 1000 results only.
Additionally, it is possible to appraise the quality of this scoring reading texts and their sentiments in a dedicated corpus explorer interface.
Field you wish to analyse
Language used in the text
Methods (for french, textblog only)
Example of usage (CorText Manager | Raw graphs | Inkscape)
Comments from students on a cours