KomParse Project
News
- Photos von der sehr erfolgreichen KomParse Abschlusspräsentation
- Die Abschlusspräsentation am Donnerstag, 27. Oktober 2011, 15:00-17:00 im THESEUS Innovationszentrum war ein voller Erfolg! (Very successful Project Wrap Up Meeting on Thursday, October 27, 2011!)
- Project presentation at the KI-11 conference at TU Berlin on Thursday, October 6, 2011.
Project Description
Welcome to the homepage of the KomParse project, a three year language technology and linguistic theory research project, funded by a grant from the ProFIT programe of the Investitionsbank Berlin (IBB; the subsidy bank of the Federal State of Berlin) and cofunded by the EFRE funds of the European Union. The main contributers to the project are the German Research Center for Artificial Intelligence (DFKI) as well as the Center for General Linguistics (ZAS).
Through partnerships with companies in the Berlin computer game industry, it is guaranteed that the solutions, evaluation criteria and technological results comply with the requirements and demands of commercial applications. Please visit the „partners“ page for more information.
Research Objectives
The KomParse project aims to integrate advanced natural language processing technologies into MMOG (Massive Multiplayer Online Games). A central unsolved problem for such games is the naturalness of NPCs (Non Player Characters), characters which are controlled by the game and not by a human player.
To achieve a more natural behaviour, dialogue ability has to become a core competence of such NPCs. This means that it is necessary to equip such characters with limited, but nevertheless adequate and robust linguistic capabilities which allow for naturalness in interactions and whose efficiency satisfies the real time requirements of interactive applications.
The project thus focusses on the following issues:- Integrating a hybrid approach to Language Understanding into traditional NPC implementations. A shallow text processing baseline guarantees fast processing and robust behaviour, which nevertheless can be incrementally enhanced at many levels with increasingly sophisticated, linguistically motivated components. This enables more flexible behaviour and robust processing of unknown input.
- Modeling of pragmatic phenomena such as the generation of optimal answers and the interpretation and generation of implicatures. This is not only of interest with regard to the primary target application but also for the development of more intuitive interfaces for human-machine communication.
- Combining the linguistic processing with domain and world knowledge. The integration of game and world knowledge will allow us to adapt the linguistic capabilities of an NPC to specific virtual contexts and to ensure data driven flexible language processing.
If you want to get an impression about how the system looks like please visit our Videos section.