0

I am trying to work on a sentiment analysis tool for reviews data. I came across a few edge-cases where the overall sentiment of a sentence may be positive but contain a negative word.
For example :

I want this iPad so bad.
F*ck yes, it looks good!

I tried 2-3 sentiment analyzing libraries and they infer such sentences with negative sentiment. I have not found any work which is looking into such solutions. Is there any known solution to handle sentiment in such contextual cases?

I tried mostly lexicon based approaches. I used NLTK, SPACY, IBM tone analyzer, TextBlob, VADER. Currently I am averaging the summation of results from all of them.

User54211
  • 121
  • 2
  • 11
  • What library did you use? What have you tried so far? please edit your question and give more details. – Amir Jan 17 '19 at 07:31
  • I tried mostly lexicon based approaches. I used NLTK, SPACY, TextBlob, VADER. Currently, I am averaging the summation of results from all of them. – User54211 Jan 17 '19 at 08:53
  • How prevalent are these edge cases? Could you live with a few false negatives? I would imagine this is an active area of sentiment analysis research. Have you tried looking through Google Scholar or arxiv? – Alex L Feb 24 '19 at 01:18
  • tried looking, did not find any strong relevant work- I may take a stab at it. Thanks for the input. – User54211 Feb 25 '19 at 12:20

0 Answers0