Resolução de topônimos em textos não estruturados baseada em heurísticas

Everyday, people use place names and spatial relationships to give directions and inform the location of events. Mentions of places, also called toponyms, are present in the most varied types of documents with geographic content, such as articles, blogs, reports and criminal reports. The geographic...

Descripción completa

Detalles Bibliográficos
Autor: Sá, Breno Alef Dourado
Tipo de recurso: tesis de maestría
Estado:Versión publicada
Fecha de publicación:2022
País:Brasil
Institución:Universidade Federal do Ceará (UFC)
Repositorio:Repositório Institucional da Universidade Federal do Ceará (UFC)
Idioma:portugués
OAI Identifier:oai:repositorio.ufc.br:riufc/69874
Acceso en línea:http://www.repositorio.ufc.br/handle/riufc/69874
Access Level:acceso abierto
Palabra clave:Geocoding
Resolução de topônimos
Desambiguação baseada em heurística
Topônimos adjetivos
Tipos de topônimo
Descripción
Sumario:Everyday, people use place names and spatial relationships to give directions and inform the location of events. Mentions of places, also called toponyms, are present in the most varied types of documents with geographic content, such as articles, blogs, reports and criminal reports. The geographic information extracted from these documents can be used in emergency response applications, epidemic monitoring, news gathering, tourism planning, among others. However, due to the absence of metadata, extracting this information from unstructured texts is not a trivial task. One of the challenges in this process is the mapping of toponyms to geographic coordinates due to the ambiguity of the names of the places, which commonly have homonyms. The process of solving toponyms to their coordinates, obtaining candidates and disambiguating them, is called geocoding. The present work proposes and evaluates two heuristics for geocoding: normalization of adjectival toponyms and geometric optimization by toponym type. Initially, the baseline is defined through experiments with heuristics. Then, two geocoders are created by modifying the baseline to use each of the heuristics proposed in this work. Finally, a third geocoder is similarly created to use the combination of the two heuristics. The results indicate an improvement in the performance of geocoding using these heuristics compared to the baseline, even surpassing state-of-the-art geocoders in the databases evaluated.