Data : Kaggle
https://www.kaggle.com/datasets/niraliivaghani/flipkart-product-customer-reviews-dataset
- This analysis consists of Exploratory Data Analysis for free text data from the Flipkart Product Reviews dataset.
- Sentiment analysis was performed using DistilBERT model from huggingface by fine-tuning it for the product reviews data.
- Topic modelling was performed using BERTopic with a spacy model for embeddings.
Notebook Details:
| Notebook | Description |
|---|---|
| flipkart_pr_eda.ipynb | Exploratory Data Analysis |
| flipkart_pr_topic_modelling | Topic Modelling for Product Reviews and Summary |
| flipkart_pr_v2.ipynb | Fine-tune DistilBERT model - local directory |
| flipkart_pr_huggingface_push.ipynb | Fine-tune DistilBERT model - Push to Huggingface |
| flipkart_pr_use_finetuned_huggingface_model.ipynb | Use fine-tuned Huggingface model |
- The counts for sentiments (Positive, Negative, Neutral) are analysed as per the price range of products.
- Wordclouds analysed for product names and summary of review features.
- DistilBERT model fine-tuned on the data
Recall: 0.94 Precision: 0.94
Note Negation are also identified in the input string
E.g "I am not unhappy with this product." -> Sentiment: Positive
- Topic Modelling performed for product names and summary of review features. Topics visualized in a 2D space for inter-topic distance
