How steveksh/product-ratings-classifier works
This repository is an academic project exploring Amazon product-review analysis and recommendation. Its visible runtime utilities parse compressed Amazon JSON records and clean review text; the notebooks indicate exploratory analysis, embeddings, NLP, and baseline/proposed modeling. The README links a project-flow image but does not describe its stages in text, and notebook contents were not sampled, so relationships involving those experiments are shown as unverified rather than asserted. No deployed service or persistent application store is evident.
Main entry points: eda_baseline.ipynb Exploratory analysis · sampled_data.csv Sample reviews · json_reader.py JSON data reader · model_baseline.ipynb Baseline model
Components 13
Data preparation
- JSON data reader
modules/json_reader.py - Review cleaner
modules/utils.py - Tiger data reader
tiger/json_reader.py
Analysis and modeling
- Exploratory analysis
baseline_code/eda_baseline.ipynb - Baseline model
baseline_code/model_baseline.ipynb - Proposed model
baseline_code/Proposed model acc_ 0.66 loss_0.546.ipynb - NLP experiments
baseline_code/nlp.ipynb - Embedding experiments
baseline_code/embedding.ipynb - ANR experiment
baseline_code/ANR.ipynb - Tiger analysis
tiger/eda.ipynb
Other components
- Project researcher
- Amazon dataset
- Sample reviews
sample_data/sampled_data.csv
Connections 10
- Project researcher → Amazon dataset: selects data
- Amazon dataset → JSON data reader: provides records
- Project researcher → Exploratory analysis: explores data
- Project researcher → Baseline model: runs baseline
- Project researcher → Proposed model: evaluates model
- Project researcher → NLP experiments: explores NLP
- Project researcher → Embedding experiments: tests embeddings
- Project researcher → ANR experiment: tests ANR
- Project researcher → Tiger analysis: explores data
- Sample reviews → Exploratory analysis: may supply data