Message Understanding Conference
The Message Understanding Conferences (MUC) were a series of research competitions initiated and funded by the US agency DARPA to promote the development of new and better methods of information extraction . The competitive nature - several competing research teams competed against each other - required the development of reliable standards for evaluation , e.g. B. the dimensions Recall and Precision developed for the MUC-2 .
Subject areas and tasks
Only at the first conference (MUC-1) could the participants determine the output format for the extracted information themselves. From the second conference on, the output format was specified and the participants' systems were evaluated. For each subject area, individual fields were given that had to be filled with information from the texts. Typical fields were e.g. B. the event, the actors, time and place of the event, the consequences etc. The number of fields increased steadily from conference to conference.
With the sixth conference (MUC-6), the identification of proper names ( Named Entity Recognition ) was added as an additional task . In doing so, all phrases in a text should be marked that designate people, places, organizations, times and dimensions.
The subject areas and types of text that had to be dealt with show a continuous transition from military to civil subjects, which reflects the increasing economic importance of information extraction .
conference | year | Text genre | Subject area (domain) |
MUC-1 | 1987 | mil. messages | Fleet operations |
MUC-2 | 1989 | mil. messages | Fleet operations |
MUC-3 | 1991 | news | terrorist activities |
MUC-4 | 1992 | news | terrorist activities |
MUC-5 | 1993 | news | Joint ventures, chip production |
MUC-6 | 1995 | news | Change of leadership in business |
MUC-7 | 1997 | news | Airplane crashes, spacecraft, rocket launches |
literature
- Ralph Grishman, Beth Sundheim: Message Understanding Conference - 6: A Brief History. (PDF; 580 kB) In: Proceedings of the 16th International Conference on Computational Linguistics (COLING), I, Copenhagen, 1996, 466-471.