Closure Properties of Bulgarian Clinical Text

Proceedings of Recent Advances in Natural Language Processing, pages 667–675, Hissar, Bulgaria, 7-13 September 2013. Closure Properties of Bulgarian Clinical Text Irina P. Temnikova Institute of ICT Bulgarian Acad. of Sciences Ivelina Nikolova Institute of ICT Bulgarian Acad. of Sciences William A. Baumgartner Jr. Computational Bioscience Program U. Colorado School of Medicine Galia Angelova Institute of ICT Bulgarian Academy of Sciences K. Bretonnel Cohen Computational Bioscience Program U. Colorado School of Medicine Abstract Sublanguages are specialized genres of language associated with specific domains and document types. When sublanguages can be recognized and adequately charac- terized, they are useful for a variety of types of natural language processing ap- plications. Although there are sublan- guage studies related to languages other than English, all previous work on sub- language recognition has focused on sub- languages related to general English. This paper tests whether a sublanguage detect- ing technique developed for English can be applied to another language. Bulgar- ian clinical documents are an excellent test case, because of a number of unique lin- guistic properties that affect their lexical and morphological characteristics. Bul- garian clinical documents were studied with respect to their closure properties and were found to fit the sublanguage model and exhibit characteristics like those noted for sublanguages related to English. was also confirmed that the clinical sub-It language phenomenon is not a coinciden- tal phenomenon of English, but applies to other languages as well. Implications of this fact for natural language processing are proposed. 1 Introduction and Related Work 1.1 Sublanguages The term sublanguage has various definitions, de- pending on criteria that will be discussed in a mo- ment. However, descriptions of the sublanguage phenomenon generally have two things in com- mon. One is that a sublanguage is the language used to communicate in a specific genre about a specialized domain. The other is that sublan- guages are restricted in some way. Sublanguages have been described for a variety of domains, including space events (Montgomery and Glover, 1986), recipes (Kittredge, 1982), le- gal documents (Charrow et al., 1982), and es- pecially for clinical documents (Hirschman and Sager, 1982; Hiz, 1982; Friedman, 1986; Dun- ham, 1986; Stetson et al., 2002; Friedman et al., 2002). Concomitantly with this domain restriction, sublanguages are typically characterized as being linguistically restricted in some way. For exam- ple, Kittredge (2003) describes sublanguages as having a restricted lexicon, relatively small num- ber of lexical classes, restricted sentence syn- tax, deviant sentence syntax, restricted word co- occurrence patterns, and different frequencies of occurrence of words and syntactic patterns from the normal language. Although sublanguage properties and sublan- guage versus general language differences have been studied in various languages (e.g. (Laippala et al., 2009) and (Wermter and Hahn, 2004), for clinical language), all approaches to sublanguage recognition have been focussed on English. (We consider recognizing the existence of a sublan- guage as a different task from learning the char- acteristics of a sublanguage; this paper is con- cerned with the problem of recognizing the exis- tence of a …

Content truncated to 500 words