Now you can easily gauge your products sentiment on e-commerce sites like Amazon, FlipKart etc. This is a sentiment analysis model built using state of the art ML libraries and models like NLTK, TextBlob, Shap and Lime.
This project is used by the following companies:
- Neurotech Computer Systems Pvt. Ltd.
How to deal with text based data was unknown to me for a long time and so with the help of this project I finally made my acquiantance with NLP!
The biggest challenge was to identify which keywords caused the review to be negative or positive. Sure I could create a wordcloud of the negative and positive reviews but it would only give me the most frequent appearig words in the corpus. Not much useful so I had to find another way. Eventually I came across model interpreatbility libraries like Shap and Lime. I used Shap for global interpretation and Lime for local interpretation which finally gave me the keywords which were responsible for classifying negative or positive reviews.
To improve my baseline model, I used NestedCV to get a robust estimate of my model. Also I used RandomizedSearchCV inside the NestedCV to tune the hyperparameters of all the models I tried (9 to be precise).
Also before feeding the data to the model, I removed all the smileys and emoticons from the corpus.
Lisztomaniac Data Scientist...
👩💻 I'm currently working on Fake Currency Detector
🧠 I'm currently learning SQL/CNN
👯♀️ I'm looking to collaborate on a CNN project (apart from image classification!)
🤔 I'm looking for help with Deploying ML/NLP/DL models
💬 Ask me about my current/completed projects or anything related to Data Science!
📫 How to reach me [email protected] or LinkedIN!
Python, Machine Learning, NLP, Deep Learning, PowerBI, Web Scraping, Automation of data related tasks
Here are some of my other projects
Contributions are always welcome!
If you have any feedback, please reach out at [email protected]