Pattern Matching in a Linguistically-Motivated Text Understanding System

Pattern Matching in a Linguistically- Motivated Text Understanding System Damaris M. Ayuso and the PLUM Research Group BBN Systems and Technologies 70 Fawcett St. Cambridge, MA 02138 dayuso@bbn.com ABSTRACT An ongoing debate in text understanding efforts centers on the use of pattern-matching techniques, which some have characterized as "designed to ignore as much text as possible," versus approaches which primarily employ rules that are domain-independent and linguisticaUy-motivated. For instance, in the message-processing community, there has been a noticeable pulling back from large- coverage grammars to the point where, in some systems, traditional models of syntax and semantics have been completely replaced by domain-specific finite-state approximations. In this paper we report on a hybrid approach which uses such domain-specific patterns as a supplement to domain-independent grammar rules, domain-independent semantic rules, and automati- cally hypothesized domain-specific semantic rules. The surprising result, as measured on TIPSTER test data, is that domain-specific pattern matching improved performance, but only slightly, over more general linguistically-motivated techniques. 1. Introduction Virtually all systems which participated in the Fifth Message Un- derstanding Conference, MUC-5 [1], used finite-state (FS) pattem matching to some extent. Two useful tasks that this approach is well suited for are: 1. treating application-specific simple constructions that may not belong in a general grammar of the language, and 2. detecting constructions which, though grammatical, may be found more reliably using domain-specific patterns. For example, special-purpose FS subgrammars were used widely to efficiently and reliably recognize equipment names and company names. This illustrates one (1) above. An illustration of (2) appears in the complex sentence below: Daio Paper Corp. said it will set up a cardboard factory in Ho Chi Minh City, Vietnam, jointly with state-run Cogido Paper Manufacturing Company. It is easy for any parser to err in not attaching the modifier "jointly" to "set up," and thereby miss the fact that a joint venture is being reported. One might argue that the sentence includes a discontigu- ous constituent ("set up ... jointly"). Nevertheless, it is easy to write a general pattern to deal with the discontiguous constituent correctly for this domain. Finite-state parsers perform simple operations, and they are fast. In data-extraction applications, where much of the input can be safely ignored, they provide an easy means to skip text without deep analysis. Some of the best-performing systems in MUC-5 relied heavily on the use of finite-state pattern-matching in crucial system components. However, there are several advantages in maintaining broad linguistically-based coverage of a language, even in a data- extraction task. First, it allows for well-defined linguistic struc- tures to be recognized and represented in a domain independent way. This provides a level of linguistic representation that can be used by other general linguistic components such as a domain- independent discourse processor. In fact, this is a representational level which will probably be evaluated in the next Message Un- derstanding Conference, MUC-6. Secondly, general linguistic coverage provides application inde- pendence. Different applications, such as data detection (infor- mation retrieval) can use the linguistic representations for …

Content truncated to 500 words