sentiment analysis from the tweet data

mratanusarkar / twitter-sentiment-analysis

a demo poc for sentiment analysis of tweets

0 stars 0 forks source link

sentiment analysis from the tweet data #7

Open mratanusarkar opened 1 year ago

mratanusarkar commented 1 year ago

We need to have a new feature built on the existing codebase.

From all the scraped tweet data, we have to come up with some "analysis feature" to analyze:

sentiment analysis of the tweet topic
emotion detection of the topic
violence or negative vibe or propaganda detection in the topic
fraud detection in the topic
trolling, bullies, and similar thread detection from the search topic
use GPT models to give a summary on the topic (PS: topic here is referred to the sentiment of the search query)

We can use various NLP techniques, binary or multi-class classification techniques, and various other ML techniques to achieve the same.

Note: The topic list above is just a draft list, and the analysis points above are to be added, edited, or deleted during this feature development.

mratanusarkar commented 1 year ago

@dsaha21 please take up this issue. Add a notebook with the implementation and analysis, and then we can decide on how to merge the same in the module when done.

Let's use this thread to discuss the same.

dsaha21 commented 1 year ago

Hi @mratanusarkar , Was going through this repo please check once. Should I perform such multi class or perform seperate classifications. please let me know once.

THE EDA part is done. only the final model building is left.

https://github.com/bentrevett/pytorch-sentiment-analysis/blob/master/5%20-%20Multi-class%20Sentiment%20Analysis.ipynb

mratanusarkar commented 1 year ago

@dsaha21 could you please share some links to your EDA or snippets to the analysis that you did? It would be easier to comment on the choice of model, or direction we decide to take, only after I get some insights on the analysis you did.

dsaha21 commented 1 year ago

Ladakh_Tweet.zip Please verify this once

dsaha21 commented 1 year ago

@dsaha21 could you please share some links to your EDA or snippets to the analysis that you did? It would be easier to comment on the choice of model, or direction we decide to take, only after I get some insights on the analysis you did.

https://colab.research.google.com/drive/1rt7nCPqnFscWzmymYoJvZcjVEFrXNjEn?usp=sharing