STPar: A Structure-Aware Triaffine Parser for Screenplay Character Coreference Resolution

2025-12-25 · Source: Transactions of the Association for Computational Linguistics · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A new Structure-aware Triaffine Parser (STPar) has been developed to address the complex task of Character Coreference Resolution in Movie Screenplays (MovieCoref). This task is more challenging than traditional coreference resolution due to the intricate narrative structures and character interactions found in screenplays. STPar integrates discourse and syntactic structures during feature encoding to analyze ternary relationships and complex interactions. It employs a triaffine scorer for high-order relations between candidate mention pairs and incorporates multi-task learning, including singleton and span detection. Evaluated on the MovieCoref dataset, STPar significantly surpassed the previous best baseline, achieving F1 score improvements of 7.4% (B^3), 21.5% (CEAF_e), 7.1% (LEA), and 10.2% (CoNLL). The integration of structural discourse and syntactic information, alongside triaffine and multi-task learning, proved beneficial.

Key takeaway

For research scientists working on natural language understanding in complex narrative domains like screenplays, STPar offers a robust framework. You should consider integrating structure-aware parsing, triaffine scoring, and multi-task learning into your coreference resolution models. This approach significantly improves F1 scores across multiple metrics, suggesting a path to more accurate character relationship mapping in challenging textual data.

Key insights

STPar enhances movie screenplay coreference resolution by integrating structural, triaffine, and multi-task learning approaches.

Principles

• Narrative structure is key to coreference.
• High-order relations improve coreference accuracy.
• Multi-task learning boosts resolution performance.

Method

STPar combines discourse and syntactic structures in feature encoding, uses a triaffine scorer for high-order relations, and integrates multi-task learning for singleton and span detection.

In practice

• Apply triaffine scoring for complex relations.
• Integrate multi-task learning for better F1 scores.
• Leverage discourse and syntactic structures.

Topics

• Character Coreference Resolution
• Movie Screenplays
• Triaffine Parsing
• Multi-task Learning
• Discourse Structures

Best for: Research Scientist, AI Researcher, AI Scientist, NLP Engineer

Related on AIssential

• Disco-RAG: Discourse-Aware Retrieval-Augmented Generation — Explicitly modeling discourse structure and planning significantly improves RAG system performance, coherence, and factual consistency.
• Coreference Resolution in spaCy — spaCy's experimental "coref" component enables robust coreference resolution for NLP applications.
• Revisiting Semantic Role Labeling: Efficient Structured Inference with Dependency-Informed Analysis — A new SRL framework offers explicit predicate-argument structure and 10x faster inference with modern encoders.
• MPDocBench-Parse: Benchmarking Practical Multi-page Document Parsing — Practical multi-page document parsing requires benchmarks assessing semantic continuity, visual content, and hierarchical structure beyond basic text.
• TantraTagger: A Benchmark Dataset for Tantrayukti-Based Discourse Structure Labelling in Sanskrit Śāstra Texts — TantraTagger is a benchmark dataset for discourse structure labelling in Sanskrit Śāstra Texts.
• Discourse Relations and Connectives in Higher Text Structure — Local discourse analysis can reveal clues about higher text structure and global coherence.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Transactions of the Association for Computational Linguistics.