JAVA Information Retrieval System (JIRS) is an Information Retrieval system specially suited for Question Answering. The traditional search engines are based on question keywords in order to obtain relevants documents from a query. JIRS, otherwise, tries to obtain pieces (passages), with higher probability to contain the answer. To fulfill this goal the system performs a search based on question n-grams -uttered in natural language- using three possible models. The n-gram models developed are language-independent, resulting JIRS in a suitable system to work in multilingual environments.
In the other hand, JIRS incorporates a powerful kernel that allows an unprecedented adaptation and scalability in modern search engines. From the begining, it was designed in order to be a powerful tool that allows an adaptation without dificulties to very diferent functions. These features allow to expand or modify JIRS aspects easily and intiutively, so that the final user does not need to know the third-party code. Moreover, it allows to generate new applications with a client/server and distribuited structures, by just modifying the configuration file.
This work presents the state of the art in Information Retrieval focused on the multilingual Question Answering. Also, a detailed description of JIRS together with its search models are described. Finally, the obtained results by this system in the CLEF competitions are presented.