A Natural Language-Based Automatic Identification System Trajectory Query Approach Using Large Language Models DOI Creative Commons
Xuan Guo,

Shutong Yu,

Jinxue Zhang

и другие.

ISPRS International Journal of Geo-Information, Год журнала: 2025, Номер 14(5), С. 204 - 204

Опубликована: Май 16, 2025

The trajectory data collected by an Automatic Identification System (AIS) are essential resource for various ships, and effective filtering querying approaches fundamental managing these data. Natural language has become the preferred way to express complex query requirements intents, due its intuitiveness universal applicability. In light of this, we propose a natural language-based AIS approach using large models. Firstly, textualization was designed convert time sequences trajectories into semantic descriptions segmenting trajectories, extracting semantics, constructing documents. Then, completed rewriting queries, retrieving generating answers. Finally, comparative experiments were conducted highlight improvements in accuracy relevance achieved our proposed method over traditional approaches. Furthermore, human study demonstrated user-friendly interaction experience enabled approach. Additionally, ablation illustrate significant contributions each module within framework. results demonstrate that effectively bridges gap between offering intuitive, user-friendly, accessible solution domain experts novices.

Язык: Английский

A Natural Language-Based Automatic Identification System Trajectory Query Approach Using Large Language Models DOI Creative Commons
Xuan Guo,

Shutong Yu,

Jinxue Zhang

и другие.

ISPRS International Journal of Geo-Information, Год журнала: 2025, Номер 14(5), С. 204 - 204

Опубликована: Май 16, 2025

The trajectory data collected by an Automatic Identification System (AIS) are essential resource for various ships, and effective filtering querying approaches fundamental managing these data. Natural language has become the preferred way to express complex query requirements intents, due its intuitiveness universal applicability. In light of this, we propose a natural language-based AIS approach using large models. Firstly, textualization was designed convert time sequences trajectories into semantic descriptions segmenting trajectories, extracting semantics, constructing documents. Then, completed rewriting queries, retrieving generating answers. Finally, comparative experiments were conducted highlight improvements in accuracy relevance achieved our proposed method over traditional approaches. Furthermore, human study demonstrated user-friendly interaction experience enabled approach. Additionally, ablation illustrate significant contributions each module within framework. results demonstrate that effectively bridges gap between offering intuitive, user-friendly, accessible solution domain experts novices.

Язык: Английский

Процитировано

0