Unsupervised Location Mapping for Narrative Corpora

2025-04-08

AI Summary

This paper formally defines and addresses the unsupervised narrative spatial mapping task: automatically constructing cross-textual geographic maps from large-scale unlabeled narrative corpora and precisely localizing individual narrative trajectories on them. Methodologically, we propose an end-to-end unsupervised pipeline integrating zero-shot geographic entity recognition and relational inference via long-context large language models, graph-structured modeling of location co-occurrence patterns, and trajectory alignment with embedding-space projection. We validate the approach’s internal consistency and historical plausibility on two heterogeneous domains—Holocaust testimonies and lakeside literature. We introduce the first benchmark dataset and evaluation protocol for this task, achieving key metric improvements over supervised baselines. Our core innovation lies in enabling generalizable spatiotemporal narrative modeling across centuries and genres—without requiring predefined location labels or manual annotation.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This work presents the task of unsupervised location mapping, which seeks to map the trajectory of an individual narrative on a spatial map of locations in which a large set of narratives take place. Despite the fundamentality and generality of the task, very little work addressed the spatial mapping of narrative texts. The task consists of two parts: (1) inducing a ``map'' with the locations mentioned in a set of texts, and (2) extracting a trajectory from a single narrative and positioning it on the map. Following recent advances in increasing the context length of large language models, we propose a pipeline for this task in a completely unsupervised manner without predefining the set of labels. We test our method on two different domains: (1) Holocaust testimonies and (2) Lake District writing, namely multi-century literature on travels in the English Lake District. We perform both intrinsic and extrinsic evaluations for the task, with encouraging results, thereby setting a benchmark and evaluation practices for the task, as well as highlighting challenges.

Problem

• Mapping narrative trajectories on spatial maps unsupervisedly
• Inducing location maps from text sets without predefined labels
• Evaluating method on Holocaust testimonies and Lake District literature

Innovation

• Unsupervised location mapping from narrative texts
• Pipeline using large language models' extended context
• Intrinsic and extrinsic evaluations on diverse domains