Book chapter · Routledge eBooks · April 22, 2025

Parallel corpus-based approach to crime event extraction

Orken J. Mamyrbayev, Nina Khairova, Waldemar Wójcik

Abstract

Crime-related event extraction from published texts enables the monitoring, analysis, and comparison of police and criminal activities across different countries or regions. This chapter presents a parallel corpus-based approach that employs a closed-domain event extraction methodology for extracting events from web news articles in low-resource languages. To identify events, their arguments, and the roles of those arguments in the source-language corpus, an enhanced pattern-based method was utilized. This method incorporates a multilingual synonyms dictionary, which represents knowledge of crime-related concepts, as well as logic-linguistic equations. The event extraction from the target-language part of the corpus employs a cross-lingual crime-related event extraction transfer technique based on supplementary knowledge of the semantic similarity patterns between the two languages. This approach does not require a pre-annotated corpus for training, making it particularly advantageous for low-resource languages. It allows for the simultaneous extraction of TRANSFER, CRIME, and POLICE event types, along with their seven subtypes from various news topics. The implementation of this approach for the Russian-Kazakh parallel corpus of news portal articles yielded an F1 measure of over 82% for crime-related event extraction in the source language and 63% in the target language.

Natural Language Processing TechniquesTopic ModelingAdvanced Text Analysis Techniquesparallel corpuslow-resource languagescross-lingual transfer

Parallel corpus-based approach to crime event extraction

Abstract

More by Orken Mamyrbayev