Prompt Engineering for Domain-Specific Geo-spatial Named Entity Disambiguation
Abstract: Despite the scarcity of employing transformer approaches for toponym resolution, this study leverages oral and transcribed text data to address the disambiguation of diverse named entities, including place names such as camps, ghettos, and streets. We utilise generative AI techniques, incorporating prompt engineering, to effectively disambiguate these named entities within geographical contexts.
Our methodology aims to demonstrate how leveraging prompt engineering from general large language models (LLMs) can be effectively employed for less commonly addressed topics, such as toponym resolution in the field of Natural Language Processing (NLP). We have evaluated the few-shot chain of thought (COT) prompting approach combining the knowledge base (KB) as a retriever to provide the fewshots required for the reasoning process of LLM. This technique illustrates the efficacy of these advanced approaches in accurately identifying and resolving toponyms in complex textual datasets, thereby contributing valuable insights to the field of geographic information systems and digital humanities.
Paper Type: Long
Research Area: Information Extraction
Research Area Keywords: named entity recognition and relation extraction, knowledge base construction; entity linking/disambiguation, zero/few-shot extraction
Contribution Types: NLP engineering experiment, Approaches to low-resource settings, Data analysis
Languages Studied: English
Submission Number: 4925
Loading