BBN's PLUM Probabilistic Language Understanding System

Three key design features distinguish PLUM from other approaches: statistical language modeling, learning algorithms and partial understanding. The first key feature is the use of statistical modeling to guide processing. For the version of PLUM used in MUC-5, part of speech information was determined by using well-known Markov modeling techniques embodied in BBN's part-of-speech tagger POST [5]. We also used a correction model, AMED [3], for improving Japanese segmentation and part-ofspeech tags assigned by JUMAN. For the microelectronics domain, we used a probabilistic model to help identify the role of a company in a capability (whether it is a developer, user, etc.). Statistical modeling in PLUM contributes to portability, robustness, and trainability. algorithms. We feel the key to portability of a data extraction system is automating the acquisition of the knowledge bases that need to change for a particular language or application. For the MUC-5 applications we used learning algorithms to train POST, AMED, and the template-filler model mentioned above. We also used a statistical learning algorithm to learn case frames for verbs from examples (the algorithm and empirical results are in [4]).