Natural language interfaces to data services will be a key technology to guarantee access to huge data repositories in an effortless way. This involves solving the complex problem of recognizing a relevant service or service composition given an ambiguous, potentially ungrammatical natural language question. As a first step toward this goal, we study methods for identifying the salient terms (or foci) in natural language questions, classifying the latter according to a taxonomy of services and extracting additional relevant information in order to route them to suitable data services. While current approaches deal with single-focus (and therefore single-domain) questions, we investigate multi-focus questions in the aim of supporting conjunctive queries over the data services they refer to. Since such complex queries have seldom been studied in the literature, we have collected an ad-hoc dataset, SeCo-600, containing 600 multi-domain queries annotated with a number of linguistic and pragmatic features. Our experiments with the dataset have allowed us to reach very high accuracy in different phases of query analysis, especially when adopting machine learning methods.

Complex Question Answering is a discipline that involves a deep understanding of question/answer relations, such as those characterizing definition and procedural questions and their answers. To contribute to the improvement of this technology, we deliver two question and answer corpora for complex questions, WEB-QA and TREC-QA, extracted by the same Question Answering system, YourQA, from the Web and from the AQUAINT-6 data collection respectively. We believe that such corpora can be useful resources to address a type of QA that is far from being efficiently solved. WEB-QA and TREC-QA are available in two formats: judgment files and training/testing files. Judgment files contain a ranked list of candidate answers to TREC-10 complex questions, extracted using YourQA as a baseline system and manually labelled according to a Likert scale from 1 (completely incorrect) to 5 (totally correct). Training and testing files contain learning instances compatible with SVM-light; these are useful for experimenting with shallow and complex structural features such as parse trees and semantic role labels. Our experiments with the above corpora have allowed to prove that structured information representation is useful to improve the accuracy of complex QA systems and to re-rank answers.