Structured vs. Unstructured Inputs in LLMs: Evaluating the Semantic and Pragmatic Predictive Power in Abnormal Event Forecasting

Jou-An Chi, Shu-Kai Hsieh


Abstract
Large Language Models (LLMs) are increasingly applied to temporally grounded reasoning tasks, yet the role of input representation remains unclear. This paper compares structured temporal inputs, represented as Temporal Knowledge Graphs (TKGs), with unstructured captions in two settings: forecasting future events and detecting anomalies in surveillance video descriptions. To enable direct comparison, we build a unified dataset by aligning anomaly labels from UCF-Crime with caption annotations from UCA. Experiments show that unstructured captions consistently yield slightly higher scores across both tasks, but the differences do not reach statistical significance. Their trade-offs, however, differ: captions provide richer semantic cues for generation, while TKGs reduce input length, suppress noise, and enhance interpretability. These findings suggest that action-centric corpora, such as surveillance or forensic narratives, naturally lend themselves to structured representations, which can provide temporal scaffolds for timeline reconstruction and more traceable reasoning. All code, data processing scripts, and experimental results are available at our GitHub repository.
Anthology ID:
2025.rocling-main.25
Volume:
Proceedings of the 37th Conference on Computational Linguistics and Speech Processing (ROCLING 2025)
Month:
November
Year:
2025
Address:
National Taiwan University, Taipei City, Taiwan
Editors:
Kai-Wei Chang, Ke-Han Lu, Chih-Kai Yang, Zhi-Rui Tam, Wen-Yu Chang, Chung-Che Wang
Venue:
ROCLING
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
237–248
Language:
URL:
https://aclanthology.org/2025.rocling-main.25/
DOI:
Bibkey:
Cite (ACL):
Jou-An Chi and Shu-Kai Hsieh. 2025. Structured vs. Unstructured Inputs in LLMs: Evaluating the Semantic and Pragmatic Predictive Power in Abnormal Event Forecasting. In Proceedings of the 37th Conference on Computational Linguistics and Speech Processing (ROCLING 2025), pages 237–248, National Taiwan University, Taipei City, Taiwan. Association for Computational Linguistics.
Cite (Informal):
Structured vs. Unstructured Inputs in LLMs: Evaluating the Semantic and Pragmatic Predictive Power in Abnormal Event Forecasting (Chi & Hsieh, ROCLING 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.rocling-main.25.pdf