TOMATE: A heuristic-based approach to extract data from HTML tables

Extracting data from user-friendly HTML tables is difficult because of their different lay outs, formats, and encoding problems. In this article, we present a new proposal that first applies several pre-processing heuristics to clean the tables, then performs functional anal ysis, and finally applie...

Descripción completa

Detalles Bibliográficos
Autores: Roldán Salvador, Juan Carlos, Jiménez Aguirre, Patricia, Szekely, Pedro, Corchuelo Gil, Rafael
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2021
País:España
Institución:Universidad de Sevilla (US)
Repositorio:idUS. Depósito de Investigación de la Universidad de Sevilla
OAI Identifier:oai:idus.us.es:11441/131986
Acceso en línea:https://hdl.handle.net/11441/131986
https://doi.org/10.1016/j.ins.2021.04.087
Access Level:acceso abierto
Palabra clave:HTML tables
Data extraction
Descripción
Sumario:Extracting data from user-friendly HTML tables is difficult because of their different lay outs, formats, and encoding problems. In this article, we present a new proposal that first applies several pre-processing heuristics to clean the tables, then performs functional anal ysis, and finally applies some post-processing heuristics to produce the output. Our most important contribution is regarding functional analysis, which we address by projecting the cells onto a high-dimensional feature space in which a standard clustering technique is used to make the meta-data cells apart from the data cells. We experimented with two large repositories of real-world HTML tables and our results confirm that our proposal can extract data from them with an F1 score of 89:50% in just 0:09 CPU seconds per table. We confronted our proposal with several competitors and the statistical analysis confirmed its superiority in terms of effectiveness, while it keeps very competitive in terms of efficiency.