IgnitionHacks_Sentiments

Inspiration

One application for sentiment analysis is to recognize social media content that is inappropriate or fake and removing it. This is a very relevant problem in an era where fake news and inappropriate content is prevalent. I always wondered how this was possible, and realized that in this hackathon we are solving a very similar but simpler problem were we evaluate the sentiment of tweets. This was my inspiration while building this project.

What it does

My model predicts the sentiment(good or bad) of tweets.

How I built it

I tried 2 popular methods used in sentiment analysis which are logistic regression and a neural network. My focus through out the project was to pre-process and clean data as the model is only as good as the data that its fed!

Challenges I ran into

Some challenges that I ran into are running out of RAM on collab!

Accomplishments that I'm proud of

I am proud that I have tried 2 different models and verified for myself which model works better and find the reasoning behind why this is the case. I am also proud that I did this by myself, but I will probably work in a group next time!

What I learned

I learned the neural networks work better with larger datasets. As well I learned many NLP pre-processing techniques like lemmatization and using the Spacy library.

What's next for Sentiment Analysis!

I'm planning on further building on this model to use it on larger pieces of text and use it to detect fake news.

Built With

• jupyter-notebook
• python
• pytoch
• spacy