2
$\begingroup$

I am trying to find a model or way to classify text which falls into a category and its a positive or negative feedback.

For ex. we have three columns

Review : Camera's not good battery backup is not very good. Ok ok product camera's not very good and battery backup is not very good.
Rating : 2
Topic :['Camera (Neutral)', 'Battery (Neutral)']

My Whole Dataset is like above and Topic is not standard one , Topic value is based on review column.

I want to train model to classify a review that falls into categories and its negative or positive.

I researched a lot but no luck.

$\endgroup$
3
  • $\begingroup$ "Researched a lot but no luck?" Try google "sentiment analysis python" you can find tons of examples. GIYF! $\endgroup$
    – lpounng
    May 23, 2022 at 16:45
  • $\begingroup$ I researched a lot, but none of those worked for me, probably because I dont have experience in this area. $\endgroup$ May 23, 2022 at 17:11
  • $\begingroup$ In this case, I suggest you search for "Twitter sentiment analysis". This is usually the 1st tutorial for beginners (determine positive vs negative sentiment). Get the basics and afterwards, you can explore topic modeling, e.g. the classic Latent Dirichlet Allocation and more recent embedding methods e.g. Word2Vec. $\endgroup$
    – lpounng
    May 24, 2022 at 1:58

1 Answer 1

1
$\begingroup$

This is called sentiment analysis, which is a type of problem in natural language processing.

Sentiment analysis is just a classification problem with a particular type of input. Any of the usual methods for doing feature extraction from text data and classification based on those features (the usual ones will be available in Python) is a possible approach, complete with the usual concerns about model construction, validation, overfitting, etc.

You wrote that you have already tried to use Python sentiment analysis tutorials and have not gotten the performance you want. This is pretty typical of the beginning of a machine learning problem, and many resources on Data Science and on the statistics Stack address improving model performance, such as this post by one of the mods over there.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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