How chiayewken/span-aste works
Span-ASTE extracts aspect–opinion spans and sentiment polarity from sentences, with a scikit-style API for training, prediction, and scoring. The main path converts the API’s triplet format to model input, reads sentences into instances, applies span-based NER and relation extraction, then formats predictions back for users. The README also describes evaluation and ATE/OTE capabilities. Edges reflect README statements or visible references in the supplied excerpts and source index; training configuration and some runtime wiring were not sampled.
Main entry points: wrapper.py SpanModel API · ner.py Span NER · relation_proper.py Relation Extractor · span_model.py Span Model
Components 16
User API
- SpanModel API
aste/wrapper.py
Input Processing
- Triplet Data
aste/data_utils.py - Document Types
span_model/data/dataset_readers/document.py - Dataset Reader
span_model/data/dataset_readers/span_model.py
Span Model
- Text Embedder
span_model/models/embedder.py - Span Utilities
span_model/models/shared.py - Span Model
span_model/models/span_model.py - Span NER
span_model/models/ner.py - Relation Extractor
span_model/models/relation_proper.py - Entity Beam Pruner
span_model/models/entity_beam_pruner.py
Prediction and Evaluation
- Model Predictor
span_model/predictors/span_model.py - Result Analysis
aste/analysis.py - F1 Calculation
span_model/training/f1.py - NER Metrics
span_model/training/ner_metrics.py - Relation Metrics
span_model/training/relation_metrics.py
Other components
- Researcher
Connections 17
- Researcher → SpanModel API: uses API
README.md - SpanModel API → Triplet Data: converts data
aste/wrapper.py - SpanModel API → Dataset Reader: trains and predicts
README.md - Dataset Reader → Document Types: reads documents
span_model/data/dataset_readers/span_model.py - Document Types → Span Utilities: uses batching
span_model/data/dataset_readers/document.py - Span Model → Span NER: runs NER
span_model/models/span_model.py - Span Model → Relation Extractor: runs relations
span_model/models/span_model.py - Span NER → Span Utilities: uses span utilities
span_model/models/ner.py - Relation Extractor → Entity Beam Pruner: prunes spans
span_model/models/relation_proper.py - Span NER → NER Metrics: reports metrics
span_model/models/ner.py - Relation Extractor → Relation Metrics: reports metrics
span_model/models/relation_proper.py - NER Metrics → F1 Calculation: computes F1
span_model/training/ner_metrics.py - Relation Metrics → F1 Calculation: computes F1
span_model/training/relation_metrics.py - SpanModel API → Model Predictor: invokes prediction
aste/wrapper.py - Model Predictor → SpanModel API: returns predictions
aste/wrapper.py - SpanModel API → Result Analysis: scores results
README.md - SpanModel API → Text Embedder: configures embeddings