Ontology-Based Natural Language Parser for E-Marketplaces S. Coppi 1 , T. Di Noia 1 , E. Di Sciascio 1 , F. M. Donini 2 , A. Pinto 1 1 Politecnico di Bari, Via Re David, 200, I-70125, Bari, Italy {s.coppi,t.dinoia,disciascio,agnese.pinto}@poliba.it 2 Università della Tuscia, via San Carlo, 32, I-01100, Viterbo, Italy donini@unitus.it Abstract. We propose an approach to Natural Language Processing exploiting knowledge domain in an e-commerce scenario. Based on such modeling an NLP parser is presented, aimed at translating demand/supply advertisements into structured Description Logic expressions, automati- cally mapping sentences with concept expressions related to a reference ontology. 1 Introduction We focus on an approach specifically aimed at translating demand / supply de- scriptions expressed in Natural Language (NL) into structured Description Logic (DL) expressions, mapping in an automated way NL sentences with concepts and roles of a DL-based ontology. Motivation for this work comes from the observa- tion that one of the major obstacles to the full exploitation of semantic-based e-marketplaces, particularly B2C and P2P ones, lies in the difficulties average users have in translating their advertisements into cumbersome expressions or in filling several form-based web pages. Yet constraining a user to completely fill in forms is in sharp contrast with the inherent Open World Assumption typical of Knowledge Representation systems. We report here how we faced this issue in the framework of MAMAS demand/supply semantic-matchmaking service [11]. Distinguishing characteristics of our NL parser include the direct use of DLs to express the semantic meaning, without intermediate stages in First Order Logic Form or Lambda calculus. This has been possible because of the strong con- textualization of the approach, oriented to e-commerce advertisements, which possess an ontological pattern that expresses their semantics and affects gram- mar creation. Such pattern is reflected both in the structure of the ontologies we built for e-commerce tasks and in the creation of the grammars. Two separate lexical category sets are taken into account; the first one for goods, the second one for their description. This choice allows to embed the problem domain into the parser grammar. Furthermore we designed the grammar in two separate levels. In this way we achieve more flexibility: the first level only depends on the ontology terminology, while the second one only on the particular DL used. Finally, our parser performs automatic disambiguation of the parsed sentences, interacting with the reasoner.