Automatic Itinerary Reconstruction from Texts

Abstract

This paper proposes an approach for the reconstruction of itineraries extracted from narrative texts. This approach is divided into two main tasks. The first extracts geographical information with natural language processing. Its outputs are annotations of so called expanded entities and expressions of displacement or perception from hiking descriptions. In order to reconstruct a plausible footprint of an itinerary described in the text, the second task uses the outputs of the first task to compute a minimum spanning tree.

Location Estimation Using Crowdsourced Spatial Relations

“…Specifically, the Moncla et al [2014aMoncla et al [ , 2014b propose an unsupervised geocoding algorithm that employs clustering techniques to estimate a spatial footprint of toponyms not found in gazetteers. The authors evaluate their approach with a corpus of real hiking descriptions in three different languages.…”

A multi-layer markup language for geospatial semantic annotations

“…The existing markup languages dedicated to the annotation are more focused on the specification of relations (spatial or spatio-temporal) than in named entities (ne). Standard markup languages consider ne as being only composed of a pure proper name, whereas we consider more complex expressions known as expanded named entities (ene) [7,15]. ene concept is particularly essential in any task that aims to retrieve all or part of evocative context of the NE, such as disambiguation task.…”

“…Although still at an early stage of development, the proposed markup language was applied for a problem of automatic information extraction and toponym resolution described in [8] and for a problem of automatic itinerary reconstruction described in [7].…”

Geocoding for texts with fine-grain toponyms

“…The first step is the geoparsing (Fig. 1a) : we annotate geographic entities thanks to a well-established process [23] based on the use of POS taggers and the application of syntactico-semantic combined patterns (cascade of transducers). Then, we start with the geocoding part of the proposal.…”

“…The objective of this work is to propose a method establishing a processing chain to support the geoparsing and geocoding of text documents describing events strongly linked with space and with a frequent use of fine-grain toponyms. The geoparsing part of the workflow is a NLP approach based on a previous work of the authors [23], which combines the use of a Part Of Speech tagger and a cascade of transducers. The real novelty of this work lies in the geocoding part of the method.…”