PyABSA Documentation

PyABSA is an open-source framework focusing on Aspect-Based Sentiment Analysis (ABSA) and related tasks, providing end-to-end capabilities for training, inference, augmentation, and visualization.

Getting Started

Learn the goals, scope, and typical use cases of PyABSA.

• Installation

• Quick start

• FAQ

Introduction to PyABSA

User Guide

Configuration, data, annotation, and visualization.

• Supported tasks and data formats

• Customizing configuration

• Datasets and annotation

• Metrics visualization

Configuration Guide

Advanced Topics

Improve performance and robustness.

• Text augmentation

• Adversarial robustness

• BPE and pretraining

Data Augmentation for Aspect-Based Sentiment Classification

Tutorials & Demos

Reproduce and get started quickly with notebooks.

• Classification / Extraction / Triplet

• Flask inference demos

Aspect-based Sentiment Classification

Getting started

• Introduction to PyABSA
  • What is PyABSA?
  • Key Features
  • Who is this for?
  • Getting Help
• Installation
  • Prerequisites
  • Setting up a Virtual Environment
  • Installing PyABSA
  • Verifying the Installation
  • Optional Dependencies

Quick start

• Quickstart: Aspect-Based Sentiment Classification (ABSC)
  • Inference with a Pre-trained Model
  • Training a Custom Model
• Aspect Term Extraction (ATE)
  • Inference with Pre-trained Models
• Aspect Term Extraction and Sentiment Classification (ATESC)
  • Inference with Pre-trained Models
  • Training a New Model
• Annotating Custom Datasets
  • Annotation Workflow
  • Using the Annotation Tool
• Adversarial Defense for Text Classification
  • Inference with Adversarial Defense
  • Generating Adversarial Examples
• Text Auto-Augmentation for Classification
  • Augmentation for Aspect-Based Sentiment Classification
  • Augmentation for Text Classification
• Training Custom Tokenizers and Word Embeddings
  • Training a BPE Tokenizer
  • Training Word2Vec Embeddings

User guide

• Supported Tasks
  • Core ABSA Tasks
  • Extended Tasks
  • Specialized Domains
  • Getting Started with Tasks
  • Model Architecture Support
  • Performance and Benchmarks
• Configuration Guide
  • Understanding Configuration Objects
  • Getting Configuration Objects
  • Common Configuration Parameters
  • Advanced Configuration Examples
  • Task-Specific Configurations
  • Saving and Loading Configurations
  • Best Practices
  • Troubleshooting
• Working with Datasets
  • Built-in Datasets
  • Custom Dataset Preparation
  • Data Preprocessing and Validation
  • Data Augmentation
  • Advanced Dataset Features
  • Best Practices
  • Troubleshooting Common Issues
• Visualizing Metrics
  • How It Works
  • Generating Metric Files
  • Visualizing the Metrics

Tutorials and notebooks

• Aspect-based Sentiment Classification
  • APCModelList
  • APCDatasetList
  • Training
  • Inference
  • Simple Prediction
  • Batch Inference
  • Annotate your own datasets via PyABSA
  • Deploy a APC demo
• Aspect Term Extraction and Sentiment Classification
  • ATEPCModelList
  • ATEPCDatasetList
  • Training
  • Inference
  • Simple Prediction
  • Batch Inference
  • Annotate your own datasets based on PyABSA
  • Deploy a ATEPC demo
• Aspect Sentiment Triplet Extraction (ASTE)
  • Training Stage
  • Start training
  • Visualizations
  • Inference Stage
  • Simple Inference
  • Batch Inference
  • Annotate your own datasets based on PyABSA
  • Deploy a ASTE demo
• Aspect Sentiment Classification
  • Flask Server
  • API Request
• Aspect setiment triplet extraction
  • Flask Server
  • API Request
• Aspect Term Extraction and Sentiment Classification
  • Flask Server
  • API Request

Advanced topics

• Data Augmentation for Aspect-Based Sentiment Classification
  • Why Use Data Augmentation for ABSA?
  • Automatic Augmentation Pipeline
  • Manual Augmentation Control
  • Domain-Specific Augmentation
  • Augmentation Quality Control
  • Integration with Training Pipeline
  • Best Practices
  • Measuring Augmentation Impact
• Text Classification

API reference

• API Reference
  • API Reference

Citation

• Citing PyABSA
• API Reference