harapalb at SemEval-2026 Task 4: Multi-Signal Neuro-Symbolic Ensembles for Narrative Similarity

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

Summary

Andrei Tiberiu Carp's "harapalb" system, presented at SemEval-2026 Task 4, introduces a multi-signal neuro-symbolic ensemble designed to determine narrative similarity. This system aims to transcend basic text matching by focusing on structural and causal alignment within narratives. Its architecture integrates three distinct signals: action-focused neural embeddings for isolating event trajectories, a symbolic Structural Survival Ratio (SSR) that quantifies the preservation of discrete event tuples through dependency parsing, and high-level structural comparisons performed by the gpt-5-mini model. When evaluated on the SemEval-2026 Task 4 test set, this integrated ensemble achieved an accuracy of 68.25%. The work was published in the Proceedings of the 20th International Workshop on Semantic Evaluation (2026) in July, San Diego.

Key takeaway

For NLP Engineers developing advanced narrative understanding systems, this research suggests moving beyond simple text matching. You should explore neuro-symbolic ensembles that integrate diverse signals, such as action-focused neural embeddings for event trajectories and symbolic Structural Survival Ratios from dependency parsing. Incorporating large language models like gpt-5-mini for high-level structural comparisons can significantly improve accuracy in determining narrative similarity, as demonstrated by the 68.25% accuracy on SemEval-2026 Task 4.

Key insights

A neuro-symbolic ensemble combining neural embeddings, symbolic parsing, and LLM comparisons enhances narrative similarity by focusing on structural and causal alignment.

Principles

• Narrative similarity needs structural and causal alignment.
• Fusing neural and symbolic signals enhances understanding.
• Beyond surface-level text matching is crucial.

Method

The method fuses action-focused neural embeddings for event trajectories, a symbolic Structural Survival Ratio (SSR) using dependency parsing for event tuple preservation, and gpt-5-mini for high-level structural comparisons to determine narrative similarity.

In practice

• Integrate gpt-5-mini for structural comparisons.
• Apply dependency parsing for event tuple analysis.
• Combine neural embeddings with symbolic signals.

Topics

• Neuro-symbolic AI
• Narrative Similarity
• SemEval-2026
• Neural Embeddings
• Dependency Parsing
• gpt-5-mini

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• 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.
• Robust Dual-Signal Fusion: Hybrid Neuro-Symbolic Gating with Compressed Chain-of-Thought Refinement for Irony Detection in Social Media Texts — Hybrid neuro-symbolic fusion with compressed Chain-of-Thought effectively overcomes LLM literalism for zero-shot irony detection.
• Modeling semantic association in self-paced reading with language model embeddings — Methodological choices, especially embedding model type, critically impact semantic association quantification in reading comprehension.
• 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.
• Chronicle: A Multimodal Foundation Model for Joint Language and Time Series Understanding — Jointly pretraining a single transformer from scratch on text and time series yields competitive multimodal representations.
• IVIE: A Neuro-symbolic Approach to Incremental and Validated Generation of Interactive Fiction Worlds — IVIE's neuro-symbolic approach balances LLM creativity with symbolic consistency for coherent, engaging interactive fiction world generation.

Open in AIssential →

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