1
$\begingroup$

I am going to do Sentiment Analysis over some tweets. The goal is to find out which post is with and which one is against a specific topic(which tweet is saying this product is good and which on is saying that is not good). I have about 6000 tweets for each Positive, Negative and Neutral. I have tested some models like Naive Bayes, NN, Decision Tree and Random Forest but I saw no good results. When I refer to confusion matrix, I see that many Positive and Negative are predicted interchangeably. Also, when I try to add some layers (for example in NN), it is going to over-fit. I use these models, but almost all the results are the same:

[(TF-IDF)] + [(Naive Bayes), (Decision Tree), (Random Forest)]

[(BERT), (Distilbert)] + [(Fully connected NN)]
Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Did you try Convolutional Neural Networks? Even though, they were introduced for computer vision, they proved to be effective for sentiment analysis as well.

You can use this tutorial on building a CNN for sentiment analysis from scratch, if you need.

Improve this answer
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.