Classify messages of a live in positive, negative and neutral

Hello, I want to classify a series of comments on a video, to analyze the behavior of the viewers, the idea is that they will show if it was positive, negative or neutral, the comments are made in Spanish, thanks

I enclose the base in case someone wants to help me, I have tried to adapt the flows that currently exist but I can’t get any of them to give the requested result
sentimientos7 - copia.xlsx (346.5 KB)

Thanks

Hello @Melilla0705 and welcome to the KNIME forum

I’ve found this post from @victor_palacios in a Sentiment Analysis on Amazon Reviews topic that could be helpful:

He may have some more answers for you.
BR

I really tried to make it work with that flow but it was not successful having many inconveniences I think it is due to my lack of experience that I could not adapt it, after giving up it was that I came here

Most of these are salutations “hi, good evening, etc.” and there are no labels for the data. So to do this you will have to manually label at least a number of samples as negative or positive (you don’t need to classify neutral since you can use thresholding for that).

Words like “label, threshold” come from machine learning so you should be familiar with that if you want to perform sentiment analysis. As well, you will also need to know Natural Language Processing (Text Mining) to deal with your text.

We have articles dedicated to this which you will find helpful to understand what you need.

So a good way to start is to make a sample with a small classification of what for me that message is categorized without a doubt Hello, good evening, I would like it to be something positive

because it is an interaction in the activity, the article you send is the same one that was indicated to me, however, as I said, I could not adapt it to my needs

Yes, start by labeling some 30-50 rows and make sure you have enough samples for positive and negative classes (roughly half of each).

Then, please attach your workflow and explain what parts could not be adapted.