Sentiment analysis in linguistic research

Hello. I am having hard time in getting the results from my sentiment analysis workflow based on IMDB reviews dataset. How can I see those texts, that were classified by the neural network? It didnt classified everything correctly, so I want to conduct a linguistic analysis of which texts can be classified correctly and which ones cannot, and why.
The link on the workflow which I used below

@ActingCushion Welcome to the Forum.

Try this. It identifies the misclassified and correctly classified rows and reassembles them with the original text. Same approach should work for the XGBoost branch. I made no attempt to do any downstream processing. That’s up to you. Let me know if this helps.


WOW. Thanks a lot. Do you know how to find out on the basis of which words the neural network decided to assign a negative or positive sentiment for each text?

There’s no simple answer to your question. I’d suggest you start by reading this:

Second, this workflow doesn’t include a neural network. The algorithms are standard machine learning algorithms. If you want to try a neural network approach, try this as a starting point although the installation is a bit tricky.

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.