PUblicationMAnagement  PumaLogo 
TitoloAutomatic Creation of quality multi-word Lexica from noisy text data

Autori Frontini F., Quochi V., Rubino F.

In AND '12 (Mumbai, India, 01 Gennaio 2012).

This paper describes the design of a tool for the automatic creation of multi-word lexica that is deployed as a web service and runs on automatically web-crawled data within the framework of the PANACEA platform. The main purpose of our task is to provide a (computationally "light") tool that creates a full high quality lexical resource of multi-word items. Within the platform, this tool is typically inserted in a work flow whose first step is automatic web-crawling. Therefore, the input data of our lexical extractor is intrinsically noisy. The paper evaluates the capacity of the tool to deal with noisy data, and in particular with texts containing a significant amount of duplicated paragraphs. The accuracy of the extraction of multi-word expressions from the original crawled corpus is compared to the accuracy of the extraction from a later "de-duplicated" version of the corpus. The paper shows how our method can extract with sufficiently good precision also from the original, noisy crawled data. The output of our tool is a multi-word lexicon formatted and encoded in XML according to the Lexical Mark-up Framework.

Soggetti Lexical induction, multi-word extraction, web-based distri- buted platform, noisy data

DownloadIcona documento      2012-A3-008.pdf

Icona documento Open access Icona documento Restricted Icona documento No access