L3IRIT at SemEval-2026 Task 4: Learning Narrative Similarity from Aligned Film Plot Summaries

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, short

Summary

The L3IRIT team, a joint research group from the IRIT and L3i laboratories, participated in SemEval-2026 Task 4, focusing on learning narrative similarity. Their approach involved constructing a novel bilingual resource by automatically aligning film plots extracted from Wikipedia. This dataset enabled training embedding models using contrastive learning objectives to effectively capture higher-level narrative structures, moving beyond surface-level lexical overlap without requiring manual annotation. The team also introduced a named-entity masking strategy to enhance narrative abstraction and minimize superficial entity-based matching. This method aims to improve representation learning for structural and event-level similarities across stories in different languages. Their system achieved accuracies of 65.75% for Task A and 61.00% for Task B, ranking in 24 of 44 and 20 of 27 scoreboards, respectively.

Key takeaway

For NLP engineers developing cross-lingual narrative understanding systems, consider integrating automatically aligned bilingual plot summaries and named-entity masking. This approach, demonstrated by L3IRIT's SemEval-2026 Task 4 performance, offers a path to build robust embedding models that capture deeper structural and event-level similarities, reducing reliance on costly manual annotations and superficial lexical matching. You can adapt this methodology to enhance your representation learning for diverse story analysis tasks.

Key insights

A novel bilingual resource and named-entity masking improve narrative similarity learning via contrastive embeddings.

Principles

• Automated plot alignment creates valuable bilingual resources.
• Contrastive learning captures higher-level narrative structures.
• Named-entity masking promotes narrative abstraction.

Method

Construct a bilingual film plot resource from Wikipedia, train contrastive embedding models, and apply named-entity masking to abstract narratives and reduce entity-based matching.

In practice

• Develop cross-lingual narrative understanding systems.
• Generate training data without manual annotation.
• Improve story summarization and comparison.

Topics

• Narrative Similarity
• SemEval-2026
• Contrastive Learning
• Embedding Models
• Named-Entity Masking
• Bilingual Resources
• Film Plot Summaries

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• ttda704 at SemEval-2026 Task 4: Modeling Narrative Structures via Pseudonymization and Multi-View Sentence Alignment — Using contrastive learning with fine-tuned sentence transformers and multi-view alignment effectively models narrative structures and similarity.
• SemEval-2026 Task 4: Narrative Story Similarity and Narrative Representation Learning — The NSNRL task operationalizes narrative similarity for evaluation and identifies areas for automated system improvement.
• FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings — FLiP models effectively interpret sentence embeddings by linearly extracting lexical content, revealing underlying biases.
• Which sentence is doing the most work in your favourite novel ? I tried to find out. — Machine learning can identify semantically distinctive sentences in texts, revealing structural or narrative pivots.
• Narrative Knowledge Weaver: Narrative-Centric Retrieval-Augmented Reasoning for Long-Form Text Understanding — Narrative Knowledge Weaver (NKW) enhances RAG by modeling evolving story worlds through structured narrative assets and multi-channel evidence assembly.
• The Cosine Similarity Trap: Why Embeddings Can’t Distinguish “War” from “Union” — Language models often cannot distinguish complementarity from conflict due to how embeddings encode relatedness, not relationship.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.