Speech Acts in Electronic Communication With Special Reference to KQML and ANSI X12

This paper examines the encoding of speech acts in KQML at length and in ANSI X12 briefly. KQML is a speech-act-based language developed with ARPA funding, and X12 is the American standard for electronic data interchange (EDI) message formats. I conclude that although speech act theory is highly relevant to electronic communication, the needs of computers are different from those of humans. Computers need to perform concisely speech acts that are clumsy in human speech, such as arranging communication paths. They also need to recognize speech act types as immediately as possible, whereas human language gets along with clumsy encodings of speech acts into grammar. 1. Speech acts in electronic communication Speech act theory — the study of how utterances function as statements, questions, commands, and so on — is no longer just an area of theoretical linguistics; it is finding increasing applications in software engineering. Several groups of researchers are experimenting with knowledge interchange languages based explicitly on speech act theory [4, 13, 9, 5]. Further, at last year’s HICSS conference, Scott Moore [14] opened an important line of investigation by comparing the repertoire of speech acts used in electronic communications with those used in human speech. Moore analyzed the illocutionary force of X12 EDI transactions, S.W.I.F.T. securities transactions, and Apple Events in the Macintosh operating system. He found, perhaps surprisingly, that all of these electronic messages display much the same variety of speech act types as human speech. There are a few gaps; for example, computers do not normally express condolences to each other. But the applicability of human-language speech act theory to electronic messaging, even the internal messages used within an operating system, is impressive. Nonetheless, electronic communication is not human speech. It is time to look more deeply at speech act theory from the viewpoint of software engineering as well as linguistic description. In this paper I will raise some methodological points, then examine the usage of speech acts in KQML, a new speech-act-based knowledge interchange language, and (briefly) in ANSI X12 EDI forms.1 2. The central claim of speech act theory 2.1. The F P hypothesis The central claim of speech act theory is that people do not just utter propositions; they perform ILLOCUTIONARY ACTS such as stating, requesting, commanding, and so forth. Every speech act consists of an ILLOCUTIONARY FORCE F applied to a proposition P . This is known as the F P hypothesis. The importance of illocutionary force was first made explicit by Austin [1] but was foreshadowed by the semantic theories of the ancient Stoics.2 Moore summarizes theF P hypothesis as claiming that “the outermost [logical] operator of every utterance (everything we could possibly say) is not Boolean, not temporal, not even defeasible — it is an illocutionary force” [14]. Further, this outermost operator is never vacuous; that is, F P P . Even when stating a fact, you are making a statement, not just voicing a fact. 2.2. Some distinctions The illocutionary force of an utterance is distinct from both its grammar and its meaning. The encoding of speech acts in English grammar is notoriously non-uniform. Some speech acts are encoded by particular syntactic structures (statements, questions, exclamations); others are encoded 1I am indebted to Steve Kimbrough and Roggie Boone for helpful conversations and encouragement relating to this work. 2Diogenes Laertius (Lives of the Philosophers VII.65–68) divides utterances into statements, yes/no questions, questions seeking information, commands, oaths, acclamations, and exclamations. 1060-3425/97 $10.00 (c) 1997 IEEE Proceedings of The Thirtieth Annual Hawwaii International Conference on System Sciences ISBN 0-8186-7862-3/97 $17.00 © 1997 IEEE Communicative speech acts Constatives (statements of fact) Assertives, predictives, retrodictives, responsives, suggestives Directives Requestives, questions, requirements, prohibitives, permissives, advisories Commissives Promises, offers Acknowledgments Apologize, condole, congratulate, greet, thank, bid, accept, reject Conventional speech acts (declarations) Effectives Appoint, nominate, suspend, demote, resign, abdicate, arrest Verdictives Acquit, certify, disqualify, clear, rule, adjudicate Figure 1. Speech acts as classified by Bach and Harnish. by particular verbs (promise, accept, offer); and still others are performed by asserting that one is performing them, such as “I hereby dub thee knight.” Further, meaning is not illocutionary force. Questions about the weather are no different, as far as speech acts are concerned, from questions about dogs and cats. Whenever a classification of speech acts becomes excessively finegrained, one suspects that the classification is picking up distinctions of meaning as well as illocution. Bierwisch [3] points out that this mistake is especially easy to make when the information content of the utterance refers to a speech act – that is, when P contains another F . Some speech acts refer to others; for example, “Please tell me your name” is a request for a statement. Nonetheless, the utterance itself is one speech act, not a nested combination of them — it is a request referring to a statement, not a request combined with a statement. 3. Classifying speech acts The study of speech acts begins with classifying them. The classification that Moore used in his study is that of Bach and Harnish [2], summarized in Fig. 1. For Moore’s purposes this classification is ideal because it makes as many distinctions as possible, thereby enumerating the whole range of human speech acts. But the Bachask one content PRICE IBM price receiver stock server