Intelligent Information Systems for Web Product Search Defended on Friday, 10 February 2017
Over the last few years, online shopping has become very popular among consumers. However, this rapid growth of e-commerce has also introduced some issues. Users can get confused or are overwhelmed by the information they get presented while searching online for products. In an attempt to lighten this burden on consumers, several product search engines have been introduced that aggregate product descriptions and price information from the Web and allow the user to easily query this information. However, because it is difficult for systems to understand all the different ways online shops represent their production information, most product search engines expect to receive the data from the participating Web shops in a custom format. In this thesis, we investigate how to design Web product search engines that automatically aggregate product information from different sources and allow users to perform effective and efficient queries on this data. We first focus on how to classify products into an existing taxonomy using only their textual descriptions. Next, we focus on the problem of finding duplicates among product descriptions found on the Web. We also investigate how one can effectively populate ontologies from semi-structured product data using lexico-syntactic mappings and how to design an approach that automatically maps one product taxonomy into another using only the category names. Last, we perform two studies where we investigate how we can reduce the consumer search effort for product search engines that rely on faceted user interfaces.
e-commerce, product category classification, product entity resolution (deduplication), ontology, population, semantic web, product taxonomy mapping, word sense disambiguation, product facet optimization, fuzzy product search