From Cohen Courses
Jump to navigationJump to search

These datasets were created by MUC, the Message Understanding Conference.

The conference was not as important as the work needed to get into the conference. It was novel in that participants had to complete in the task in order to gain admission. The performance measure used in the tasks is F1, the geometric mean of precision and recall.

Table partially taken from [1]

Conference Year Text Source Topic (Domain) Task
MUC-1 1987 Mil. reports Fleet Operations Open ended
MUC-2 1989 Mil. reports Fleet Operations Template extraction (1 template; 10 slots)
MUC-3 1991 News reports Terrorist activities in Latin America Template extraction (1 template; 18 slots)
MUC-4 1992 News reports Terrorist activities in Latin America
Additional details can be found here
Template extraction (1 template; 24 slots)
MUC-5 1993 News reports Corporate Joint Ventures, Microelectronic production Template extraction (11 templates; 47 slots total) Some templates were nested. Note: Only multilingual MUC; also in Japanese.
MUC-6 1995 News reports Negotiation of Labor Disputes and Corporate Management Succession Named entity extraction, Coreference, Template element (this was mainly person and organization), Scenario template
MUC-7 1997 News reports Airplane crashes, and Rocket/Missile Launches Named entity extraction, Coreference, Template elements, Facts or Template relations, Events

MUC-7: overview of results


General information on MUC A history of the MUC conference