On Representing Governed Prepositions and Handling "Incorrect" and Novel Prepositions

NLP systems, in order to be robust, must handle novel and i l l formed input . One common type of error involves the use of non-s tandard preposit ions to mark arguments . In this paper, we argue that such errors can be handled in a systematic fashion, and that a system designed to handle them offers other advantages. We o f f e r a c lass i f i ca t ion scheme for preposi t ion usage errors. Fur ther , we show how the k n o w l e d g e r e p r e s e n t a t i o n employed in the SRA NLP system faci l i tates handl ing these data. 1.0 INTRODUCTION It is well known that NLP systems, in order to be robust, must handle illfo rmed input . One common type of error involves the use o f n o n s t a n d a r d preposit ions to mark arguments. In this paper, we argue that such errors can be handled in a systematic fashion, and that a system designed to handle them offers other advantages. The examples of non-s tandard preposit ions we present in the paper are taken f rom colloquial language, both wri t ten and oral. The type of error these examples represent is qui te f requen t in colloquial wr i t ten language. The f requency of such examples rises sharply in evolving sub-languages and in oral colloquial language. In developing an NLP system to be used by various U.S. government customers, we have been sensitized to the need to handle var ia t ion and innovat ion in preposit ion usage. Handl ing this type of variat ion or innovat ion is part of our overall capabi l i ty to handle novel predicates, which arc f requent in sublanguage. Novel predicates created for sublanguages arc less "stable" in how they mark arguments (ARGUMENT MAPPING) than general English "core" predicates which speakers learn as chi ldren. It can be expected that the eventual advent of successful speech unde r s t and ing systems will f u r t he r emphasize the need to handle this and other var iat ion. The NLP system under deve lopment at SRA incorporates a Natura l Language Knowledge Base (NLKB), a major part of which consists of objects represent ing SEMANTIC PREDICATE CLASSES. The system uses h ierarchical knowledge sources; all general "class-level" character is t ics of a semant ic predicate class, inc lud ing the number , type, and mark ing of their arguments , are put in the NLKB. This leads to increased e f f i c iency in a number of system aspects, e.g., the lexicon is more compact and easier to mod i fy since it only contains id iosyncrat ic in format ion . This representa t ion allows us to d is t inguish be tween Icxical ly and seman t i ca l ly de te rmined ARGUIVIENT MAPPING and to fo rmula te general class-level const ra int re laxat ion mechanisms. I . I CLASSIFYING P R E P O S I T I O N USAGE Preposi t ion usage in English in posit ions governed by predica t ing elements, whether adject ival , verbal, or nominal , may be classified as (I) lexically de te rmined , (2) s y n t a c t i c a l l y d e t e r m i n e d , o r (3) semant ical ly de te rmined . Examples are: LEXICALLY DETERMINED: laugh at, afraid of SYNTACTICALLY DETERMINED: by in passive sentences SEMANTICALLY DETERMINED: