Toward Automatic Processing of English Metalanguage

International Joint Conference on Natural Language Processing, pages 760–766, Nagoya, Japan, 14-18 October 2013. Toward Automatic Processing of English Metalanguage Shomir Wilson* School of Informatics University of Edinburgh 10 Crichton Street Edinburgh EH8 9AB, UK School of Computer Science Carnegie Mellon University 5000 Forbes Avenue Pittsburgh, PA 15213 USA shomir@cs.cmu.edu Abstract The metalinguistic facilities of natural lan- guage are crucial to our ability to communi- cate, but the patterns behind the appearance of metalanguage—and thus the clues for how we may instruct computers to detect it—have re- mained relatively unknown. This paper de- scribes the first results on the feasibility of au- tomatically identifying metalanguage in Eng- lish text. A core metalinguistic vocabulary has been identified, supporting intuitions about the phenomenon and aiding in its detection and delineation. These results open the door to ap- plications that can extract the direct, salient in- formation that metalanguage encodes. 1 Introduction In linguistic communication it is sometimes nec- essary to refer to features of language, such as orthography, vocabulary, structure, pragmatics, or meaning. Metalanguage enables a speaker to select a linguistically-relevant referent over (or in addition to) other typical referents (Audi, 1995). Metalanguage is illustrated in sentences such as (1) Graupel refers to a kind of precipitation. (2) The name is actually Rolla. (3) Keep tabs on is a colloquial phrase. (4) He wrote “All gone” and nothing more. The roles of the bold substrings in the above sen- tences contrast with those in (5)-(8) below: (5) Graupel fell on the weary hikers. (6) Rolla is a small town. (7) Keep tabs on him, will you? (8) They were all gone when I returned. * This research was performed during a prior affiliation with the University of Maryland at College Park. Conventional stylistic cues, such as italics in (1), (2), and (3) and quotation marks in (4), some- times help the audience to recognize metalinguis- tic statements in written language. In spoken lan- guage or in written contexts where stylistic cues are not used, the audience is expected to identify metalinguistic statements using paralinguistic cues (such as intonation, when speaking) or con- text and meaning. Metalanguage is both pervasive and, paradox- ically, the subject of limited attention in research on language technologies. The ability to produce and understand metalanguage is a core linguistic competence that allows humans to converse flex- ibly, unrestricted by domain (Anderson et al., 2002). Humans use it to establish grounding, ver- ify audience understanding, and maintain com- munication channels (Anderson et al., 2004). Metalanguage encodes direct and salient infor- mation about language, but many typical exam- ples thwart parsers with novel word usage or ar- rangement (Wilson, 2011a). Metalanguage is difficult to classify through the interpretive lens of word senses, given that conventional word senses have little relevance when a word appears chiefly “as a word”. The roles of metalanguage in L2 language acquisition (Hu, 2010), expres- sion of sentiment toward others’ utterances (Jaworski et al., 2004), and irony (Sperber and Wilson, 1981) have also been noted. This …

Content truncated to 500 words