Recognizing similar or close meaning on different surface form is a common challenge in various Natural Language Processing and Information Access applications. However, we identified multiple limitations in existing resources that can be used for solving the vocabulary mismatch problem. To this end, we will propose the Diversifiable Bootstrapping algorithm that can learn paraphrase patterns with a high lexical coverage. The algorithm works in a lightly-supervised iterative fashion, where instance and pattern acquisition are interleaved, each using information provided by the other. By tweaking a parameter in the algorithm, resulting patterns can be diversifiable with a specific degree one can control.

Multilingual Question Answering systems are generally very complex, integrating several sub-modules to achieve their result. Global metrics (such as average precision and recall) are insufficient when evaluating the performance of individual sub-modules and their influence on each other. In this paper, we present a modular approach to error analysis and evaluation; we use manually-constructed, gold-standard input for each module to obtain an upper-bound for the (local) performance of that module. This approach enables us to identify existing problem areas quickly, and to target improvements accordingly.