Aspect-Based Sentiment Analysis for Restaurant

Tool

Features

• Classify a sentence
• Annotate data (for training, testing...)
• Compare results (between annotators)

Offline

1. pip install -r requirements.txt
2. streamlit run app.py

Online

👉 Live App 👈

Note

• SAEvaluate.java: official evaluation tool from vlsp.org.vn
• app.py: web app to classify, annotate data, and compare annotated results
• change_format.py: convert original data to dataframe, and vice versa
• notebook.ipynb: training notebook
• model/pipe.joblib: trained model (based on Logistic Regression)
• data/original: original data
• data/csv: new format (converted from original data using change_format.py)

Screenshots

Classify user sentence

Annotate raw data

Compare labeled results between annotators

References

1. VLSP 2018 - Aspect Based Sentiment Analysis (VABSA 2018)
2. Annotation Guidelines
3. NLP@UIT at VLSP 2018: A SUPERVISED METHOD FOR ASPECT BASED SENTIMENT ANALYSIS