search

sergey-tihon / Stanford.NLP.NET

Stanford NLP for .NET
http://sergey-tihon.github.io/Stanford.NLP.NET/
MIT License
595 stars 123 forks source link

Sentiment analysis train! help me,please! #86

Closed wzjjia closed 5 years ago

wzjjia commented 6 years ago

How does the Sentiment analysis function train me to provide a sentence like, "I'm happy today" and I'm going to train him to be positive

sergey-tihon commented 6 years ago

Sorry, but I am not aware of any way to train your own model.

As I understand, the shipped model is trained using Sentiment Treebank https://nlp.stanford.edu/sentiment/treebank.html So you can your cases there and next model version should be better.

You also may want to search issues on original repo https://github.com/stanfordnlp/CoreNLP or questions on SO https://stackoverflow.com/questions/tagged/stanford-nlp

Seems that there is a way to do this...

wzjjia commented 6 years ago

"If you sometimes like to go to the movies to have fun Wasabi is a good place to start." How can I convert the above statement to the following format？ (3 (3 (2 If) (3 (2 you) (3 (2 sometimes) (2 (2 like) (3 (2 to) (3 (3 (2 go) (2 (2 to) (2 (2 the) (2 movies)))) (3 (2 to) (3 (2 have) (4 fun))))))))) (2 (2 ,) (2 (2 Wasabi) (3 (3 (2 is) (2 (2 a) (2 (3 good) (2 (2 place) (2 (2 to) (2 start)))))) (2 .)))))

cheungtitus commented 6 years ago

@wzjjia I'm new to this as well but it appears BuildBinarizedDataset is the one to use. https://github.com/stanfordnlp/CoreNLP/issues/296 https://github.com/stanfordnlp/CoreNLP/issues/32#issuecomment-61037436 https://www.kaggle.com/c/sentiment-analysis-on-movie-reviews/discussion/12304

What is not clear is how do you break down each sentence into its phrases and then assign sentiment scores for them.

For example why for this sentence. Today is not a good day.

How do we know it has the following phrases.

good
good day
a good day

Perhaps @rstjobs @J38 @manning can comment.

wzjjia commented 5 years ago

Here's the problem How do we know it has the following phrases. good good day a good day

bond- commented 5 years ago

Wow seriously?? no one responded to this. That just shows the amount of support CoreNLP has.

sergey-tihon commented 5 years ago

Sorry, this is the wrong repo for such questions Please ask on StackOverflow or open new issue in official repo https://github.com/stanfordnlp/CoreNLP

bond- commented 5 years ago

Opps sorry didn't look at the repo name