CODEWITHSUNDEEP
javanlptext-miningsentiment-analysis
Dmytro Titov
Dmytro Titov

Reputation: 3201

Sentiment analysis: more than 3 sentiments

My app needs sentiment analysis functionality. I've found plenty of services and libraries which can help with this task. But most of them have "three-dimensional" output: the text may be classified as "positive", "negative" or "neutral.

But what if I need larger variety of options? For example: "confident/doubtful", "calm/alerted", "kind/aggressive" or something like that.

Is it even possible to perform such classification? May be there are already some services/frameworks/libraries available?

Upvotes: 3

Views: 595

Answers (2)

clemtoy
clemtoy

Reputation: 1731

You should try WordNet-Affect. This ressource provides a tree of emotions. As it is a quite old ressource, you will have to manually parsed it and to map the IDs with WordNet 1.6 synsets (I did this work in Python here).

Upvotes: 1

Walter
Walter

Reputation: 279

You can use Machine Learning algorithm. I have used "Support Vector Machine" (https://en.wikipedia.org/wiki/Support_vector_machine) for sentiment analysis.

Support Vector Machine is a supervised algorithm, so you need to train the algorithm with data previously classified (confident, doubtful, calm, alerted, kind, aggressive). Finally, you will get a model that you can use to classify new text.

I have used the LibSVM library with phyton (https://www.csie.ntu.edu.tw/~cjlin/libsvm/) with good results. I think you can also use it with java.

Upvotes: 0

Related Questions