This repository provides a Chinese text classifier for short product reviews, primarily for sentiment analysis. It offers a range of models achieving over 90% accuracy, targeting developers and researchers working with Chinese e-commerce data.

How It Works

The classifier implements several deep learning architectures, including Transformer, word2vec combined with TextCNN, FastText, and recurrent networks (LSTM/GRU) with Attention mechanisms. This multi-model approach allows for flexibility and comparison, with word embeddings pre-trained on a large dataset to capture semantic meaning.

Quick Start & Requirements

• Install via pip install tensorflow==2.0.
• Requires Python 3.
• Dataset: 100,000京东 (JD.com) product reviews (data/goods_zh.txt), labeled as 0 (negative) or 1 (positive).

Highlighted Details

• Achieves >90% accuracy across implemented models.
• Models include Transformer, word2vec+TextCNN, FastText, word2vec+LSTM/GRU, word2vec+LSTM/GRU+Attention, and word2vec+Bi_LSTM+Attention.
• Future improvements planned for GloVe, GPT, BERT, and ERNIE.

Maintenance & Community

• Project maintained by ami66.
• WeChat public account ID: datanlp for more ML/DL project knowledge.

Licensing & Compatibility

• License not specified in the README.
• Compatibility for commercial use or closed-source linking is undetermined.

Limitations & Caveats

The project is built on TensorFlow 2.0, which is an older version. The README does not specify the license, which may impact commercial use.