Question Answering on the Real Semantic Web Vanessa López 1 , Miriam Fernández 2 , Enrico Motta 1 , Marta Sabou 1, , Victoria Uren 1 1 Knowledge Media Institute, The Open University. United Kingdom. {v.lopez, e.motta, r.m.sabou, v.s.uren}@open.ac.uk 2 Politécnica Superior, Universidad Autónoma de Madrid. Spain.{miriam.fernandez}@uam.es Abstract. Restriction to a predefined set of ontologies, and consequently limitation to specific domain environments is a pervading drawback in Semantic Search technologies. In this work we present PowerAqua [2], a multi-ontology-based Question Answering (QA) platform that exploits multiple distributed ontologies and knowledge bases to answer queries in multi-domain environments. The system interprets the user’s Natural Language (NL) query using the available semantic information, and translates the user terminology into the ontology terminology (triples), retrieving accurate semantic entity values as response to the user’s request. 1 Introduction The goal of Question Answering (QA) systems is to allow users to ask questions, using their own terminology, and receive a concise answer. A new trend on QA is ontology based QA where the power of ontologies as a model of knowledge is and its semantic information is directly exploited for the query analysis and translation (Aqualog [3], ORAKEL [5], GINO [1]). In contrast with traditional NLIDB systems, semantic QA needs very little customization being almost ontology independent. However, they are limited to the knowledge encoded on one, or a set of a priori defined ontologies in the same domain (semantic intranets). As consequence, they are still far away from their successful use as full NL open interfaces to the SW. For instance, neither to involve the user to provide domain specific grammars or vocabulary (ORAKEL), or the use of guided user interfaces (GINO) which generates a dynamic grammar rule for every ontology element, or asking the user every time ambiguity arises (AquaLog) are feasible solutions in the large SW scenario, where portability is not longer enough and openness is required. In this work we present PowerAqua [2], a platform that evolves from the earlier AquaLog system, designed to take advantage of the vast amount of heterogeneous semantic data offered by the SW in order to interpret a query, without making any assumptions of the relevant ontologies to a particular query a priori. wn w2 w1 term e4, 38, e98 e1, e3 e1, e2 Ontology Entity Natural Laguage query Ontology Discovery Ontology index Ontologies 2.- Power-Map 1.- Linguistic Component Linguistic triples Semantic Filtering WordNet On O2 O1 Ontology e33 e24 e12, e14 Entity On O2 O1 Ontology e41 e21 e11, e13 Entity Entity-mapping tables 3.- Triple Similarity Service Keyword1 (k1) Keyword2 (k2) On O2 O1 Ontology ... (e21, r1, e2) (e2, r3, e24) (e11, r1, e12) (e13, r3, e14) Triples Linguistic triples (k1, rel, k2) Answer! Triple-mapping tables Figure1: Power Aqua Flow