0

I am working with Word2vec (Gensim, in python) to understand meaning of sentences (by each word in them). My goal is to be able to realize if the sentence indicates about the feeling of the speaker. Where can I find this kind of a dictionary of words? For example one dictionary for words that indicate happiness and other for sadness. Thanks

rdo123
  • 149
  • 1
  • 1
  • 9
  • Your end goal – determining if a sentence indicates more 'happiness' or 'sadness' – doesn't necessarily require any external lexicon of words associated with 'happy'/'sad'. It's a classification problem, & if you simply had enough example sentences labeled with the "right" answers, you could try many alt methods. (Many of these methods might have as a side-effect creation of such word-to-emotion info. But they might not need them as an input - & they might work better without importing any other word set's assumptions.) So, is the goal more important, or the specific method you've asked about? – gojomo Apr 09 '20 at 19:03
  • It would be great to use sentences, but I don't have them either. I thought that the most simple way is to compare the similarity of the words in the sentence to these dictionaries and then if it is close, I can understand what is the meaning. – rdo123 Apr 09 '20 at 20:55
  • Do you mean, you don't have sentences labeled with the 'right' answer? (Because you clearly have the sentences you want to evaluate.) Can *you* rank a sentence on a sadness-to-happiness scale? If so, then take a few dozen (or a few hundred) of the sentences you want to evaluate with code, and hand-annotate them with *your* judgements, as training data. (If you don't do this, even using other word-assessments will just be flying blind: how will you know if the evaluations make sense, or get better or worse as you try different steps/weightings/etc?) – gojomo Apr 09 '20 at 23:23
  • I don't have the sentences yet, this is the problem. I need to build something that will be able to work with the live data. Thus I thought that using dictionaries and comparing similarly with Gensim I will be able to get a prediction – rdo123 Apr 10 '20 at 00:20
  • It's nearly impossible to build anything sensible without some example data to test your assumptions, so I'd suggest finding your own set of sentences that you think may be a good stand-in for the sentences you'll eventually have to classify – or just waiting until actual data is available. – gojomo Apr 10 '20 at 08:32

1 Answers1

1

Try SentiWordNet

"SentiWordNet is a lexical resource for opinion mining that assigns to each synset of WordNet three sentiment scores: positivity, negativity, and objectivity"

ashutosh singh
  • 511
  • 3
  • 15