AAAI-91 Workshop on Natural Language Text Retrieval
A Three-Tiered Approach to Natural Language Text Retrieval
August 19, 1991
This paper presents a three-tiered approach to text processing in which a novel and quite powerful knowledge-based form of information retrieval plays a central role. We are actively using this approach in our efforts for the NOSC-sponsored Third Meæøge Understanding Conference (MUC-3). Research groups participating in the MUC-3 conference must evaluate the performance of their text processing systems on a black-box template (database record) generation task. To perform this task, a text processing system must extract information about different types of terrorist acts from newspaper articles and radio broadcasts. Relevant data about terrorist acts, including when and where they occurred, who perpetrated them, what weapons were used, who or what were the targets, and so forth, constitute the content of templates used to represent them in a database. The knowledge-based form of information retrieval, which plays a key role in our three-tiered approach, allows us to define an interesting level of text analysis that falls somewhere between what is possible with standard IR techniques and deep linguistic analysis.
Downloads: 21 downloads