indirect speech act

4.Conversation cooperation principle。 会话合作的一般原则
Back
Examples - Indirect Speech Acts in use:
A . Mike and Ann are in the living room. Mike asks Ann whether she’d like to eat dinner in the living room or the kitchen. Ann replies: It’s cold in here. -----let’s eat dinner in the kitchen. B . The queen and her butler, James ,are in the drawing room. The window is open. The queen says: It’s cold in here. -----James, shut the window. C.Mike and Ann are in the greenhouse.Mike wonders why his orchids haven't bloomed.Ann replies:It's cold in here. -----The orchids aren't blooming.
“间接言语行为是通过实施另一种施事行
为的方式来间接地实施某一种施事行为”.
Indirect speech act
Secondary illocutionary act(Literal meaning) 次要施事行为 primary illocutionary act(Nonliteral meanign) 首要施事行为 For example: Could you pass the salt? 1.The secondary illocutionary act:The surface form of this utterance is an interrogative one. 2.The primary illocutionary act:In whether the hearer have ability to pass the salt, while expressing his intention which is a request.

4. The conventional indirect speech act
• 六种间接指令（indirect directives） • 1、涉及听话人实施A的能力的句子。 Can you pass the salt? • 2、涉及说话人希望听话人实施A的句子。 I would like you to pass the salt. • 3、涉及听话人实施A的句子。 Will you pass the salt? • 4、涉及听话人实施A的意愿的句子。 Would you be willing to pass the salt?
• 通过另一种施事行为（次要施事行为 secondary illocutionary act）的方式来 间接地实施某一种施事行为（首要施 事行为primary illocutionary act）。 • eg.Can you pass the salt? S: 询问（interrogative）: P:请求（request）：
2. Indirect Speech Acts
• Terms • 字面语力（literal force）：句子本身表达的言语 行为。 • 实事语力（illocutionary force）：在字面语力的 基础上推导出来的语力，即间接语力。
• This kind of speech acts, "in which one illocutionary act is porformed indirectly by way of performing another" (Searle1995b:60),Searle calls indirect speech acts.
• 5、涉及实施A理由的 句子。 • It might help if you pass the salt. • 6、把上述形式中的一种嵌入到另一种中去，以及 在上述的一种形式中嵌入一种显性指令性施事动 词的句子。 • I hope you(2) won't mind (4)if I ask you (IV)if you could(1) pass the salt.

Pragmatics-Teaching Plan (5)IV. Indirect Speech Acts4.1 Primary and Secondary Illocutionary Acts4.2 Inferential Process of Indirect Speech Acts4.3 Searle’s Researches on Indirect Directives4.3.1 Six Groups of Indirect Directives4.3.2 Basic Ideas about Indirect Directives4.3.2.1 Idiomatic, but not Idioms4.3.2.2 Coexistence of the Indirect Directive Act and the Literal Illocutionary Act4.3.3 Interpretation of Indirect Directives in terms of Speech Acts Theory4.4 Some Problems in Searle’s Pattern of Indirect Speech Analysis4.5 An Extended Analysis of Indirect CommissivesReadings:Searle, J. R. Indirect speech acts [A]. P. Cole & J. Morgan (eds.) Syntax and Semantics 3: Speech Acts.New York : Academic Press, 1975.Searle, J. R. Expression and Meaning: Studies in the Theory of Speech Acts [M]. Cambridge: Cambridge University Press, 1979.Huang Yan. 2009: 4.6 Indirect Speech ActsHe Zhaoxiong, 2011: 3.3 Indirect Speech Acts涂靖. Irony言语行为的研究[J]. 四川外语学院学报, 2004, (4).丁风. 汉语请求言语行为中的性别差异[J]. 西外学报, 2002, (1): 46-50.朱晓姝. 中美学生抱怨言语行为的差异研究[J]. 西外学报, 2008, (1): 51-55.黄永红. 对言语行为“道歉”的跨文化研究[J]. 解放军外国语学院学报, 2001, (5): 33-36.吴淑琼et al. 英汉批评言语行为策略对比研究[J]. 外语教学, 2004, (2): 22-25.周启强, 白解红. 原型范畴与间接指令[J]. 外语与外语教学, 2004, (12): 1-4.刘森林. 语用策略与言语行为[J]. 外语教学, 2003, (5): 10-15.赵英玲，李洁芳. 虚假邀请(ostensible invitation)言语行为的语用研究[J]. 哈尔滨工业大学学报(社科版), 2004, (7)..Isaac, E. A. Ostensible Invitation [J]. Language in Society, 1999, (19)Song Yuan. On the Probing Speech Act in English Talk Shows —A Case Study of Oprah Show [D].South-central University for Ethnic Communities, 2010.(英语脱口秀中的打探言语行为研究—以欧普拉脱口秀为个案)IV. Indirect Speech Acts4.1 Primary and Secondary Illocutionary ActsSearle (1975: 60) initiated the notion of indirect speech act to indicate the illocutionary act which is performed indirectly by way of performing another.In some cases, what people say (locutionary act / literal meaning) are in consistence with what they do (illocutionary act), as in Example (1):(1) A: What’s your name?B: My name is Julia.B’s utterance is a direct reply to A’s question about her name. What B says is in complete agreement with what she intends to say. Indirect usage of speech is not involved here.But frequently, what people actually say do not agree with what they intend to say / what they do, as in Example (2):(2) It’s cold in here.If the speaker’s intention is only to inform the hearer of the temperature at the location and time of speaking and nothing else, his speech is direct. If, however, the speaker’s intention is, by virtue of telling the hearer of his own feelings about the temperature, to get the hearer to do something like closing the window, switching on the air-conditioner, his speech is indirect. In latter case, through this utterance, the speaker simultaneously performs two illocutionary acts: an assertive (陈述) and a directive (指令). He indirectly performs the illocutionary act of directive by means of performing the illocutionary act of assertive.Searle (1975) assigned the two terms, primary illocutionary act and secondary illocutionary act, respectively to the two illocutionary acts performed simultaneously in indirect speech acts. Primary illocutionary act, such as the directive act in (2), represents the real intention of the speaker; secondary illocutionary act, such as the assertive act in (2), is another illocutionary act performed in order to perform the primary illocutionary act. The secondary illocutionary act coincides with the literal meaning of sentence whereas the primary illocutionary act means more than the literal meaning (The speaker performs the secondary illocutionary act by way of uttering a sentence, the literal meaning of which is such that its literal utterance constitutes a performance of that illocutionary act).4.2 Inferential Process of Indirect Speech ActsHow is it possible for the speaker to mean the primary illocution when he only utters a sentence that means the secondary illocution? And how does the hearer understand the non-literal primary illocutionary act from understanding the literal secondary illocutionary act?Searle (1975) reconstructs a series of inference steps by means of which the hearer could, from the literal meaning of the sentence, draw the speaker’s real intention, and interprets the inferential apparatuses as follows:“In indirect speech acts the speaker communicates to the hearer more than he actually says by way of relying on their mutually shared background information, both linguistic and nonlinguistic, together with the general powers of rationality and inference on the part of the hearer” (Searle, 1979: 31-32)Specifically, the apparatus necessary to explain the indirect part of indirect speech acts are: (i) a theory of speech acts; (ii) certain general principles of cooperative conversation (not limited to Grice’s principles); (iii) mutually shared factual background information of the speaker and the hearer; (iv) an ability on the part of the hearer to make inferences.In real communication, the hearer will resort to these apparatuses unconsciously in their interpretation of indirect speech acts. Take the following verbal exchange as an example:(3) Student X: Let’s go to the movies tonight.Student Y: I have to study for an exam.X’s utterance constitutes a proposal in virtue of its meaning, in particular because of the meaning of Let’s, and the normal response to a proposal is an acceptance or rejection, but in virtue of the literal meaning, Y’s reply is neither acceptance nor rejection, it is simply a statement about Y. The questions then arise: How does X know that Y’s response is a rejection of his own proposal? And how is it possible for Y to mean his response as a rejection of the proposal?Searle’s (1979: 33-34) interpretation of the question is: the primary illocutionary act performed in Y’s utterance is the rejection of X’s proposal, and Y does that by way of performing a secondary illocutionary act of making a statement to the effect that he has to prepare for the exam. He performs the secondary illocutionary act by way of uttering a sentence, the literal meaning of which is such that its literal utterance constitutes a performance of that illocutionary act. Alternatively, the secondary illocutionary act is literal; the primary illocutionary act is not literal. Then the above questions can be re-expressed as: How does X understand the non-literal primary illocutionary act from understanding the literal secondary illocutionary act? And how is it possible for Y to mean the primary illocution when he only utters a sentence that means the secondary illocution? Searle reconstructed ten inferential steps necessary for X to derive the non-literal primary illocution from the literal secondary illocution as follows:`Step 1: I (X) have made a proposal to Y, and in response he has made a statement to the effect that he has to study for an exam (facts about the conversation).Step 2: I assume that Y is cooperative in the conversation and that therefore his remark is intended to be relevant (principles of conversational cooperation)Step 3: A relevant response must be one of acceptance, rejection, counterproposal, further discussion, etc. (theory of speech acts).Step 4: But his literal utterance was not one of these, and so was not a relevant response (inference from Step 1 and 3).Step 5: Therefore, he probably means more than he says. Assuming that his remark is relevant, his primary illocutionary point must differ from his literal one (inference from Step 2 and 4).(This step is crucial. Unless a hearer has some inferential strategy for finding out when primary illocutionary points differ from literal illocutionary points, he has no way of understanding indirect illocutionary acts.)Step 6: I know that studying for an exam normally takes a large amount of time relative to a single evening, and I know that going to the movies normally takes a large amount of time relative to a single evening (factual background information).Step 7: Therefore, he probably cannot both go to the movies and study for an exam in one evening (inference from Step 6).Step 8: A preparatory condition on the acceptance of a proposal, or on any other commissive, is the ability to perform the act predicted in the propositional content condition (theory of speech acts).Step 9: Therefore, I know that he has said something that has the consequence that he probably cannot consistently accept the proposal (inference from Step 1, 7, and 8).Step 10: Therefore, his primary illocutionary point is probably to reject the proposal (inference from Step 5 and 9).(Searle, 1979: 34-35)Through these steps, any indirect illocutionary points can be drawn by the hearer, no matter what kind of literal forms they take on.4.3 Searle’s Researches on Indirect Directives4.3.1 Six Groups of Indirect DirectivesIn the field of indirect illocutionary acts, Searle shows largest interest in the area of indirect directives because, to him, ordinary conversational requirements of politeness normally make it awkward to issue flat imperative sentences (e.g. Leave the room!) or explicit performatives (e.g.I order you to leave the room!). In directives, politeness is the chief motivation for indirectness. Searle collected six categories of sentences that are conventionally used in the performance of indirect directives and indicated, with bold type,the generality of the syntactic forms of the sentences in question:Group 1: Sentences concerning H’s ability to perform A:Can you reach the salt?Can you pass the salt?Could you be a little more quiet?You could be a little more quiet.You can go now (this may also be a permission=you may go now).Are you able to reach the book on the top shelf?Have you got change for a dollar?Group 2: Sentences concerning S’s wish or want that H will do A:I would like you to go now.I want you to do this for me, Henry.I would/should appreciate it if you would/could do it for me.I would/should be most grateful if you would/could help us out.I’d rather you didn’t do that any more.I’d be very much obliged if you would pay me the money back soon.I hope you’ll do it.I wish you wouldn’t do that.Group 3: Sentences concerning H’s doing A:Officers will henceforth wear ties at dinner.Will you quit making that awful racket?Would you kindly get off my foot?Won’t you stop making that noise soon?Aren’t you going to eat your cereal?Group 4: Sentences concerning H’s desire or willingness to do A:Would you be willing to write a letter of recommendation for me?Do you want to hand me that hammer over there on the table?Would you mind not making so much noise?Would it be convenient for you to come on Wednesday?Would it be too much (trouble) for you to pay me the money next Wednesday?Group 5: Sentences concerning reasons for doing A:You ought to be more polite to you mother.You should leave immediately.Must you continue hammering that way?Ought you to eat quite so much spaghetti?Should you be wearing John’s tie?You had better go now.Hadn’t you better go now?Why not stop here?Why don’t you try it just once?Why don’t you be quiet?It would be better for you (for us all) if you would leave the room.It wouldn’t hurt if you left now.It might help if you shut up.It would be better if you gave me the money now.It would be a good idea if you left town.We’d all be a good off if you’d just pipe down a bit.This class also contains many examples that have no generality of form but obviously, in an appropriate context, would be uttered as indirect requests, e.g.:You’re standing on my foot.I can’t see the movie screen while you have that hat on.Group 6: Sentences embedding one of these elements inside another; also, sentences embedding an explicit directive illocutionary verb inside one of these contexts.Would you mind awfully if I asked you if you could write me a letter of recommendation?Would it be too much if I suggested that you could possibly make a little less noise?Might I ask you to take off your hat?I hope you won’t mind if I ask you if you could leave us alone?I would appreciate it if you could make less noise.(Searle, 1979: 36-39)4.3.2 Basic Ideas about Indirect Directives4.3.2.1 Idiomatic, but not IdiomsTo some speech act theorists, some syntactic structures can be taken as the idioms of conventionally performing certain indirect directives, just like using kick the bucket to meandie. According to them, Can you + V? / Would you please + V? / Would you mind + V-ing? are all idioms for indirect requests. An indirect request on sb. to open the door, for example, can be realized by the idiomatic forms (4), (5) and (6):(4) Can you open the door?(5) Would you please open the door?(6) Would you mind opening the door?Contrastive to this understanding, Searle’s (1979: 40-41) viewpoint is that indirect directives are idiomatic, but not idioms. To him, sentences like (7) are not idioms like kicked the bucket in (8):(7) Can you pass the salt?(8) Jones kicked the bucket.Searle (1979: 41) presents two reasons, indicating why the indirect requests like (7) are not idioms:First, when these sentences are used as requests, they still have their literal meaning, so that literal responses are also appropriate (e.g. Sure, I can. Here it is. or No, sorry, I can’t.It’s down there at the end of the table.) Idioms, however, cannot be understood literally. In response to (8), it’s inappropriate to say (9):(9) Really? Did he hurt his leg?Second, a word-for-word translation of the sentences like (7) into other languages will often produce sentences with the same indirect illocutionary act potential, but a word-for-word translation of Jones kicked the bucket into other languages will not produce a sentence meaning “Jones died.”However, Searle also admits that indirect directives are idiomatic, though they are not idioms. Sentences like (4)-(7) are idiomatic or conventional ways of making requests. In general, their non-idiomatic equivalents or synonyms would not have the same indirect illocutionary act potential. Thus, the idiomatic structure “Do you want to hand me the hammer over there on the table?” can be uttered as a request, but its non-idiomatic equivalent “Is it the case that you at present desire to hand me that hammer over there on the table?”has a formal and unnatural character that would prevent it from becoming a candidate for an indirect request potential.4.3.2.2 Coexistence of the Indirect Directive Act and the Literal Illocutionary ActIn his research on indirect directives, Searle (1979: 42-43) proposes the idea that whenone of the sentences like (4)-(7) is uttered with the primary illocutionary point of a directive, the literal illocutionary act is also performed. In every one of these cases, the speaker issues a directive by way of asking a question or making a statement. The fact that his primary illocutionary point is directive does not alter the fact that he is asking a question or making a statement. In cases where these sentences are uttered as requests, they still have their literal meaning.Evidence that these sentences keep their literal meanings when uttered as indirect requests is that responses that are appropriate to their literal utterances (as demonstrated in B1 and B2) are also appropriate to their indirect speech act utterances. For example:A: Can you pass the salt?B1: No, sorry, I can’t. It’s down there at the end of the table.B2: Yes, I can. (Here it is.)Additional evidence for the existence of the literal meaning in indirect speech is that a subsequent report of the utterances can truly report the literal illocutionary act. For example, the utterance of “I want you to leave now, Bill” can be reported by an utterance of “He told me he wanted me to leave, so I left”, and the utterance of “Can you reach the salt?” can be reported by an utterance of “He asked me whether I could reach the salt”. Similarly, an utterance of “Could you do it for me, Henry; could you do it for me and Cynthia and the children?” can be reported by an utterance of “He asked me whether I could do it for him and Cynthia and the children”4.3.3 Interpretation of Indirect Directives in terms of Speech Acts TheorySearle’s categorization of indirect directives is mainly based on the felicity conditions on performing illocutionary acts, as are mentioned in the Theory of Speech Acts. Searle (1979: 44) illustrates the felicity conditions necessary for the performance of directive class of illocutionary acts as follows:Preparatory condition: H is able to perform A.Sincerity condition: S wants H to do A.Propositional content condition: S predicates a future act A of H.Essential condition: Counts as an attempt by S to get H to do A.Based on these conditions, the six groups of indirect directives collected by Searle can be reduced to three types: (1) those having to do with felicity conditions on the performance of a directive illocutionary act; (2) those having to do with reasons for doing the act; (3) those embedding one element inside another one.All of Groups 1-3 indirect directives concern felicity conditions on directiveillocutionary acts since:Group 1: The ability of H to perform A concerns the preparatory condition;Group 2: The desire of S that H perform A concerns the sincerity condition;Group 3: The prediction of A of H concerns the propositional content condition.Group 4 and 5 both concern reasons for doing A, since desire or willingness to do something (Group 4) is “a reason par excellence for doing it”. (Searle, 1979: 45) Group 6 is a special class only by courtesy, since its elements either are performative verbs or are already contained in the other two categories of felicity conditions and reasons.Group 1-6 sentences of indirect directives demonstrate that the syntactic structures for indirect directives are either statements or questions. Searle (1979: 45) generalizes four strategies of making indirect directives (primary illocutionary acts) by means of the illocutionary acts of making a statement or asking a question:Generalization 1: S can make an indirect request (or other directive) by either asking whether or stating that a preparatory condition concerning H’s ability to do A obtains. (Group 1)Generalization 2: S can make an indirect directive by either asking whether or stating that the propositional content condition obtains. (Group 3)Generalization 3: S can make an indirect directive by stating that the sincerity condition obtains, but not by asking whether it obtains. (Group 2)Generalization 4: S can make an indirect directive by either stating that or asking whether there are good or overriding reasons for doing A, except where the reason is that H wants or wishes, etc., to do A, in which case he can only ask whether H wants, wishes, etc., to do A. (Groups 4 & 5)Why is it a general tendency that people like to issue directives in an indirect way? To this problem, Searle’s (1979: 48) answer is that the chief motivation for using indirect directives is politeness. The indirect directive form Can you (pass the salt), for example, is polite in at least two aspects: First, its speaker does not presume to know about the hearer’s abilities, as he would if he issued an imperative sentence (e.g. Pass the salt, please!); second, the form gives the hearer the option of refusing, since yes-no question allows no as a possible answer. Thus, the hearer’s compliance (obedience) can be made to appear a free act rather than obeying a command.4.4 Some Problems in Searle’s Pattern of Indirect Speech AnalysisSearle (1979: 48) points out that several problems still exist in his own pattern of indirect speech analysis, even though it might be successful in many more cases.Problem 1: Why is it that some syntactical forms work better than others for indirectspeech acts, if the apparatuses by which indirect speech acts are meant and understood are perfectly general—having to do with the theory of speech acts, the principles of cooperative conversation, and shared background information, and having nothing to do with any particular syntactical form? Why can I ask you to do something by saying “Can you hand me that book on the top shelf?”, but not by saying “Is it the case that you at present have the ability to hand me that book on the top shelf?”To the first part of the problem, Searle’s answer is this: The theory of speech acts and the principles of conversational cooperation do, indeed, provide a framework within which indirect illocutionary acts can be meant and understood. However, within this framework certain forms will tend to become conventionally established as the standard idiomatic forms for indirect speech acts. While keeping their literal meanings, they will acquire conventional uses as, e.g., polite forms for requests.Searle (1979: 49) argues that there can be conventions of usage that are not meaning conventions. For instance, can you, could you, I want you to, and numerous other forms are conventional ways of making requests, but at the same time they do not have an imperative meaning. Politeness is the most prominent motivation for indirectness in requests, and certain forms naturally tend to become the conventionally polite ways of making indirect requests.To the second part of the problem, Searle (p50) also gives an explanation: In order to be a plausible candidate for an utterance as an indirect speech act, a sentence has to be idiomatic to start with. The utterance “Can you hand me that book on the top shelf?”will naturally gives the hearer an impression that he is requested to do something whereas the utterance “Is it the case that you at present have the ability to hand me that book on the top shelf?” will fail to do so. Why? This is because in general, if one speaks unidiomatically, the hearer will assume that there must be a special reason for it, and thus various assumptions of normal speech are suspended.Based on this consideration, Searle (1979: 50) suggests that besides the maxims (in Cooperative Principles) proposed by Grice, there seems to be an additional maxim of conversation: “Speak idiomatically unless there is some special reason not to.”Problem 2: Why is there an asymmetry between the sincerity condition and the others such that one can perform an indirect request only by asserting the satisfaction of a sincerity condition, but not by querying it, whereas one can perform indirect directives by either asserting or querying the satisfaction of the propositional content and preparatory conditions? For example, an utterance of “I want you to do it” can be a request, but not an utterance of “Do I want you to do it?”. The former can take please, the latter cannot. This phenomenon may be attributed to the fact that in normal cases, it is odd to ask other people about the existence of one’s own elementary psychological states.4.5 An Extended Analysis of Indirect CommissivesThe general approach to indirect directives will also work for other types of indirect speech acts. Among them, indirect commissives constitute a good type of examples. Consider the following sentences that are conventionally used to perform indirect commissives (offers or promises):Group 1: Sentences concerning the preparatory conditions(A) that S is able to perform the act:Can I help you?I can do that for you.I could get it for you.Could I be of assistance?(B) that H wants S to perform the act:Would you like some help?Do you want me to go now, Sally?Wouldn’t you like me to bring some more next time I come?Would you rather I came on Tuesday?Group 2: Sentences concerning the sincerity condition:I intend to do it for you.I plan on repairing it for you next week.Group 3: Sentences concerning the propositional content condition:I will do it for you.I am going to give it to you next time you stop by.Shall I give you the money now?Group 4: Sentences concerning S’s wish or willingness to do A:I want to be of any help I can.I’d be willing to do it (if you want me to).Group 5: Sentences concerning (other) reasons for S’s doing A:I think I had better leave you alone.Wouldn’t it be better if I gave you some assistance?You need my help, Cynthia.(Searle, 1979: 54-55) The above analysis can be summed up by the following generalizations:Generalization 1: S can make an indirect commissive by either asking whether or stating that the preparatory condition concerning his ability to do A obtains. (Group 1A) Generalization 2: S can make an indirect commissive by asking whether, though not by stating that, the preparatory condition concerning H’s wish or want that S do A obtains. (Group 1B)Generalization 3: S can make an indirect commissive by stating that, and in some forms by asking whether, the propositional content condition obtains. (Group 3)Generalization 4: S can make an indirect commissive by stating that, not by asking whether, the sincerity condition obtains. (Group 2)Generalization 5: S can make an indirect commissive by stating that or by asking whether there are good or overriding reasons for doing A, except where the reason is that S wants or desires to do A, in which case he can only state but not ask whether he wants to do A. (Groups 4 & 5)(Searle, 1979: 56)。

Indirect Speech Acts and Collaborativeness in Human-Machine Dialogue SystemsFrédéric LandraginThales Research and TechnologyRoute départementale 128F-91767 Palaiseau CedexFranceAbstract: In human-machine dialogue systems, indirect and composite speech acts have to be treated in a proper way. First because they appears frequently in human-human dialogue, and then constitute an important aspect of spontaneous communication. Second because they are linked to collaborative aspects. We describe some complex speech acts phenomena and some methods for a system to treat them, with the help of hypotheses on user’s mental states.1. IntroductionA lot of semantic and pragmatic research works (Searle, 1975; Perrault & Allen, 1980, etc.) deal with complex speech acts, particularly indirect speech acts—meaning something instead of something else, for instance the question “can you give me the salt?” instead of the request “please give me the salt”—, and composite speech acts—meaning several things simultaneously, for instance “how long does it take to go by this way which seems to be the shortest?” where a comment (“this way seems to be the shortest”) is added to the question (“how long does it take to go by this way?”). Some other research works (Grice, 1975; Quignard, 2002, etc.) deal with collaborative dialogue and argumentation in dialogue, and are not always linked to speech acts theories.On the other hand, there are only a few computer science works in the area of human-machine dialogue that exploit relevantly the previous research works in order to build on systems with high understanding abilities. Only a very few real systems are able to manage indirect and composite speech acts, and most of them (see for instance the dialogue systems that were designed during the MIAMM and OZONE European projects, and /euprojects/ozone) rely on simple predefined rules that transform pre-identified types of utterances into types of speech acts.Face to these discrepancies between theories and implemented systems, some key questions arise on the way to design more efficient dialogue systems:•Does a human-machine dialogue system need to identify the surface and profound speech acts (for an indirect speech act), and the several speech acts (for a composite one)? Among these identified speech acts, to which must the system react? All of them or only a privileged or optimal one? What are the possible criteria that allow the system to choose between them?•When interpreting a complex speech act, does the system need to make hypotheses on the speaker’s mental states (beliefs, desires, intentions)? Does the system need to manage internal structures reflecting the dialogue state, for instance, following the terms from (Portner, 2004), the ‘common ground’, the ‘questions stack’, and the ‘to-do lists’? How can all these aspects be confronted and managed together? What are the minimal requirements, i.e., the simplest internal structures for interpreting correctly the user’s messages?•What are the links between complex speech acts processing and collaborativeness in dialogue? When interpreting complex speech acts, does a system increase the dialoguecollaborative aspects? Which collaborative characteristics allow the system to resolve indirect speech acts?In this paper we will not be able to answer to all of these questions. But, keeping them in mind, we want to clarify some aspects of the design of human-like understanding systems. In the second section we present a quick state of the art for research works dealing with speech acts. In the third section we address the problems that appear when applying theories to the implementation of systems. Then, we propose in the fourth section the foundations for a computational model to process complex speech acts. Some principles are drawn and illustrated, and we conclude with the future works to increase the communicative abilities of dialogue systems.2. Complex speech actsTwo main manners to apprehend speech acts can be distinguished. Speech acts can first be viewed as semantic units. Types of sentences and utterances are then identified using semantic criteria. (Hamblin, 1987) is an example of such an approach, with the case of imperatives. But semantics is not sufficient for explaining everything, and some pragmatic aspects such as basic conversational principles—the maxims of (Grice, 1975), for instance—are required. Some semanticists try to integrate these aspects into semantic factors, but the resulting theories are not really convincing. In fact, the use of language especially in dialogue situations seems impossible to be included into the semantic content. Second, speech acts can be viewed as pragmatic units. Following Searle (and, for instance, Vanderveken), speech acts intervene at a different level than semantic content. An illocutionary force F is added to a propositional content P: “F (P)”, see (Searle, 1979). This force is not linked to any semantic parameter, that can be seen as a lack. A link can sometimes be useful, for instance between the semantics and the illocutionary force of an expressive attributive adjective such as “damn” (“I have to mow the damn lawn”, see Potts, 2003).More precisely about indirect speech acts, the semantic point of view is well illustrated by (Hamblin, 1987). Proper imperatives (commands, requests, demands, advices), whishes, permissives, and undertakings correspond to speech acts types. There are not linked to particular forms of utterances, but all of them have semantic properties, such as the presence of a conditional clause for permissives. In theory, the speech act type can be identified when analyzing the semantic characteristics of the utterance. But these characteristics are not sufficient and Hamblin emphasizes the importance of pragmatic factors. Concerning the pragmatic point of view, the main fundamental work is the one of (Searle, 1969) and (Searle, 1975). Following his idea of an illocutionary force added to a propositional content, Searle propose five main categories of speech acts: assertives, directives, commissives, expressives, and declarations. For each of them, the formula F (P) can be completed with more precisions on the adjustment direction (from the word to the world, or from the world to the word, or both) and on the sincerity condition (B = belief, D = desire, I = intention). If the adjustment direction has no interest for computational pragmatics and human-machine dialogue (Quignard, 2002), on the other hand taking the user’s mental states (B, D, I) into account clarifies a lot the differences between the categories of speech acts (Searle, 1976): •assertives = B (P),•directives = D (listener does Q),•commissives = I (speaker does Q),•expressives = ∅ (?) (speaker/listener + property),•declarations = ∅ (P).For the treatment of indirect speech acts—so, for the identification of the correct speech act category—, (Searle, 1975) proposed a ten-steps method consisting of a chain of inferences. Among these inferences, one is related to the speaker’s mental states. With the example of “can you give me the salt?”, this inference corresponds to “the speaker probably knows that the answer is yes, so his utterance is perhaps something else than a question”. Even if Searle considered that this inference is not essential, it seems to be of importance for the resolution of indirect speech acts. The role of mental states (four in the Searle’s paper: belief, desire, intention, pleasure) is here emphasized for indirect speech acts processing. Searle exploited these mental states to provide a list of indirect directives due to some conventional principles. Pragmatic aspects are then privileged, and more recent works such as (Potts, 2003) and (Portner, 2004) emphasize their importance. In his Ph.D. dissertation, Potts describes in detail conventional implicatures of two sorts: supplements and expressives. Parentheticals, appositives, or relatives like the “who”-relative in “I spent part of every summer until I was ten with my grandmother, who lived in a working-class suburb of Boston”, are examples of supplementary expressions, as well as speaker-oriented adverbs such as “amazingly” in “After first agreeing to lend me a modem to test, Motorola changed its mind and said that, amazingly, it had none to spare” (Potts, 2003). Epithets, honorifics, or expressive attributive adjectives are examples of expressive expressions. These categories can be considered as typical cases of composite speech acts.What about collaborative aspects? Are they included into pragmatics aspects, or do they intervene at a different level, illustrated by the following formula: “C (F (P))”? Following (Quignard, 2002) and other works dealing with collaboration and argumentation, the collaborative aspects consist of dialectic functions that intervene at another level than the utterance understanding. Speech acts are linked to an informative and communicative level, but dialectic functions are linked to an evaluative level. Nevertheless, we can consider that the use of an indirect speech act by the speaker is a way for him to be less incisive and more polite and more collaborative. As an illustration of the argumentative possibilities, the ‘C’ in the previous formula can be anyone of the dialectic functions proposed by (Quignard, 2002), for instance ‘ARG-PRO-MT’ (the speaker provides an argument in favor of his own thesis) or ‘REQ-TDP-MT’ (the speaker asks his opponent to take a position with respect to his thesis).3. Speech acts processing in dialogue systemsHuman-machine natural language dialogue systems cover an increasing number of phenomena, and exploit more and more the results from linguistic and pragmatic researches. For a long time there existed two main categories of dialogue systems, first the ‘command systems’ where the user uttered only a simple chain of orders, and second the ‘information systems’ where the user could only ask questions to the system, as he did with classical database systems using an artificial query language such as SQL. Then, there was no speech act processing. In fact, each dialogue system was designed for the treatment of one speech act (command or question). Whatever the form of an utterance (assertion, question, or order), its illocutionary force was systematically brought back to the expected one. It was the case for most of industrial systems, and also for research prototypes such as the multimodal dialogue system from the MIAMM European project.With the objective of more natural and less unilateral dialogues, there is a need for different treatments for the various illocutionary forces, and then for a module dedicated to the identification of the most probable illocutionary force of each speaker’s utterance. Thesimplest way to proceed is to specify during the system design phase a ‘hashtable’ linking each possible utterance form to its supposed force. But all possibilities have to be thought by advance, and it is difficult to make the system evolve. The main advantage of this method is to avoid the implementation of a complex algorithm for indirect speech acts resolution. That is why it is used for the design of a lot of dialogue systems, such as the INRIA dialogue system demonstrator of OZONE European project.Concerning the elaboration of a module for indirect speech acts understanding, a lot of computational models have been proposed. Following the idea of conversational postulates from (Gordon & Lakoff, 1975), the hashtable previously mentioned can be replaced by a set of rules that allow the system to identify the profound speech act from the surface one. This method is more flexible, but it presupposes that it is always possible to extract from one utterance form one profound speech act. The problem is that the same utterance may have both interpretations (the surface act interpretation and the profound act one), i.e., some rules must not apply in some situations. A recent work (Xuereb & Caelen, 2004) introduces statistics in order to identify the most probable profound speech act in a dynamic and flexible manner. As a last example of dialogue system, the one from France Telecom R&D (Sadek et al., 1997) follows Searle’s theory by implementing an automatic identification of some mental states of the speaker. The system is then able to make inferences exploiting the speaker’s beliefs and intentions. One problem is here the complexity of the algorithm, which needs to manage logical forms related to propositional contents and to mental states. Imagine for instance something like “Believes (Speaker, (Knows (System, P)))”.4. Managing mental states and dialogue structures to interpret complex speech acts4.1. Segmenting speech actsHow can we design a computational model for speech acts processing? The first point to address is the segmentation of speech acts. Even if an utterance from the user corresponds to one grammatical sentence, several speech acts can be identified. This is the case with examples like “how long does it take to go by this way which seems to be the shortest?”. In fact, it depends on the way we consider discourse structure and relations between discourse segments. With the previous example, a ‘comment’ relation can be identified between the two following discourse segments: “how long does it take to go by this way” and “which seems to be the shortest”. Then, following a theory considering both discourse structure and speech acts, such as (Asher & Lascarides, 2003), this will lead to have two separated speech acts. To the contrary, following an approach where one utterance corresponds to one discourse segment (due to significant acoustic blanks before and after the utterance), this will lead to have one composite speech act. Thus, there is no immediate answer to the segmentation problem. Since composite speech acts might appear even with discourse structure considerations, we can follow the acoustic-based approach and consider that one speech act (simple or composite) is attributed to each dialogue turn.4.2. Interpreting indirect speech actsWhat are the prerequisites for a computational model of indirect speech acts processing? Considering that the profound speech act may be at the origin of the answer content, and that the surface act may be exploited for the answer form, the system must identify both of them. Considering that the same form may lead to different interpretations (due to contextualfactors) in terms of profound acts (Asher & Lascarides, 2001), the identification may rely on the following factors and resources:1.Linguistic and semantic characteristics of the utterance (following Hamblin).For instance: imperative, interrogative, or indicative? what is the semanticcategory of the verb?2.The dialogue history, i.e., the previous utterances and their interpretationsrepresented with logical forms. A particular linguistic form can be usedfrequently by the user with a particular aim that implies the use of an indirectspeech act (we can imagine machine-learning techniques for the managementof such a phenomenon).3. A lexicon of dialogue pairs, with the associated profound speech acts andpossible reactions. For instance, the system may know that a proposition has tobe answered to by an acceptation, a reject, or a counterproposition.4.Classical conventional uses and associated set expressions. This is typicallythe case for “can you give me the salt?” and similar constructions. NB: Thisitem can be seen as a part of the previous one.5.The list of the system abilities and all task constraints. For instance, if thedialogue system helps the user finding a restaurant, a question like “can youlist me the Chinese restaurants near Palaiseau” is of course a request.6.Hypotheses on the speaker’s mental states. For instance, the hypothesis that healready knows the answer to his question. That was the case for the salt, butalso for a lot of less conventional situations: when the user asks the system“can you open this file?”, he may know (or believe) that the system is able toopen the file.Then, a computational model for indirect speech acts processing may take these parameters into account. We can define a priority order that corresponds to the order used for the previous list. One point in this list is the importance of the dialogue history. The system reaction is based on the nature of the user’s utterance as well as on the current state of the dialogue. The notions of ‘common ground’ (CG), ‘questions set’ or ‘questions under discussion’ (QUD), and ‘to-do lists’ (TDL) are here of importance. Following (Portner, 2004), to each of these three notions corresponds a stack of propositions. An assertion from the user increments the CG with the corresponding proposition, a question increments the QUD with the corresponding propositional function, and an order increments the TDL. CG, QUD and TDL are built on during the dialogue, right after each semantic and pragmatic analysis. Following its illocutionary force, an utterance is saved in the right stack with a logical form corresponding to the result of its semantic analysis. Taken together, the three stacks constitute the major part of the dialogue history. In task-oriented human-machine dialogues, the task constraints may be pregnant so that a ‘default’ TDL can be imagined.When producing a message reacting to an indirect speech act, the system first takes into account the profound act in order to determine an answering content in ‘coherence’ with the task and the dialogue history. Then, the system has to choose between ignoring the surface act, or taking it into account for the linguistic form of the answer and its ‘cohesion’ within the dialogue. The only parameter for making this choice seems to be the maintenance of a certain linguistic cohesion. As an example, consider the classical example “do you have time?”. The form is a question but the speaker will of course not be satisfied with a “yes” response. If she has a watch, the hearer can react with “five o’clock” or “yes, five o’clock”. Including a “yes”, the second answer has a better cohesion with the speaker’s utterance than the first one. Then the system will favor this second answer.4.3. Interpreting composite speech actsConcerning composite speech acts, the system has to identify the primary act and the secondary one(s), and to classify them using a salience hierarchy. The identification factors are the same than for indirect speech acts. Moreover, the classification of the different acts relies on the same parameters than their identification:1.Linguistic and semantic characteristics of the utterance. In particular: epithets,evaluative adverbs, appositions, subordinate clauses, etc. (following Potts). In“how long does it take to go by this way which seems to be the shortest?”, thesubordinate clause constitutes a criterion for identifying a secondary act. NB:As we will see with the other items of this list, the parameters categories forindirect speech acts processing and composite speech acts processing are thesame, but the criteria that are exploited are not.2.The dialogue history. When the same composite act is produced again by thespeaker, the same classification has to be made by the system if the first onewas a success (we can also imagine here machine-learning techniques, not onlyfor the identification of the potential primary act, but also for the determinationof the speech act category—simple or composite).3. A lexicon of dialogue pairs, with the associated primary and secondary speechacts and all possible reactions. For instance, the system may know that theassociation of a question and a comment has to be answered to by a response tothe question, a reaction to the comment (confirming, infirming), or both.4.Classical conventional uses and associated set expressions. This is for instancethe case for “who the hell did that?” and similar constructions. In French, thisis above all the case for utterances like “qui a bien pu faire ça ?”, where thepresence of “bien”—a very used adverb with various significations—leads inthis particular construction to the identification of a composite speech act.5.Task constraints. When determining the primary act, the more relevant to thetask the act is, the better it is classified.6.Hypotheses on the speaker’s mental states. When determining the primary act,the best hypothesis is the one that has the most important contextual effects tothe mental states, see (Sperber & Wilson, 1995). With the example “how longdoes it take to go by this way which seems to be the shortest?”, there areseveral possibilities. First possibility: the comment is true, i.e., the way theuser is pointing out is really the shortest one. Then, confirming this belief willhave a limited effect on the speaker’s mental states, whereas uttering theresponse to the question will have a greater effect (adding a new knowledge tothe user’s mind). Second possibility: the comment is false. Then, infirming thisbelief will have an important effect on the speaker’s mental states. Theproblem here is that it is difficult to compare this effect to the effectcorresponding to the new knowledge. Thus, in this case, the system might reactto both acts.A computational model for composite speech acts processing may take these parameters into account, in the same order than presented in the list. The result of this process is an ordered list of speech acts, beginning with the primary act that was identified.When determining the system reaction to a composite speech act, only this primary act may be answered to. Comments can be added for secondary acts. It depends on the level of collaboration that is expected from the system. For instance, with “how long does it take withthis travel, which seems to be the shortest?”, the most salient act is the one of the main proposition (linguistic factor), and the secondary act is the comment. A simple answer from the system can be “this travel will take you 20 minutes” (reaction to the only primary act, corresponding to the expected reaction). A more collaborative answer can be “this travel will take you 20 minutes, and is actually the shortest” (reaction to all acts). If the comment were computed as little more important than the question, the answer should have been “you’re right, this travel is actually the shortest, and it will take you 20 minutes”. And if the comment were considered as the primary act, the question should have been ignored. More precisely, this case can appear when the comment is false. In this case, the system may consider that correcting the comment is more important than answering the question. Thus, the answer may be: “this is not the shortest travel, the shortest is that one”.5. Conclusion and future workWith the aim to design more collaborative and more natural speech-based dialogue systems, indirect and composite speech acts must be taken into account. The most relevant speech act has then to be identified, and this is an important issue for systems with deep understanding abilities. Criteria such as salience or relevance may be exploited during this identification process. Moreover, some of the user’s mental states, and particularly intentions, must also be taken into account. The intention behind an utterance appears to be the main parameter for determining an adequate reaction or answer to this utterance. When this intention is pregnant, the form of the message has sometimes no importance. Dialogue systems must also be able to identify what the user already knows and what he is susceptible to want to know. The purpose is to never repeat what is already known and to focus the dialogue on what might be known by the user. Giving such a capacity to dialogue systems constitutes for the most part a future work.Other future works can be identified concerning the determination and the exploitation of stronger links between complex speech acts and collaborativeness. When he is not familiar with the task, the user often produces utterances where several leads are left opened. This is the case with “I want to go to Paris, no problem?” or “if possible, I would like to buy a train ticket”, where the system may choose between ignoring and reacting to the question and the “if possible” (since they belong to phatic aspects of oral communication). Strict assertions and orders are not frequent in spontaneous communication. They are often accompanied with phatic questions or expressions. Several questions can also be produced together inside one long utterance with a quick rhythm, like “can I have a taxi—uh is it possible?—to go to Palaiseau? uh is it here that I can asked for a taxi? is it possible?”. Face to such an utterance, a collaborative system may be able to identify the main questioning of the user and to answer to it (and to calm him). That can be done with answers like “yes, I’m going to call the taxi company” or “I cannot do that but you can ask to my colleague”. With such a spontaneous example, it really seems that collaborative behaviours rely on a fine treatment of complex speech acts.ReferencesAsher, N. & Lascarides, A. (2001). Indirect Speech Acts. Synthese, 128 (1-2).Asher, N. & Lascarides, A. (2003). Logics of Conversation. Cambridge University Press. Ginzburg, J. & Fernández, R. (2005). Conversational Acts and Non Sentential Utterances in Multilogue. In: Sixth International Workshop on Computational Semantics, Tilburg. Gordon, D. & Lakoff, G. (1975). Conversational Postulates. In: P. Cole & J. Morgan (eds), Syntax and Semantics, Vol. 3: Speech Acts. Academic Press.Grice, H.P. (1975). Logic and Conversation. In: P. Cole & J. Morgan (eds), Syntax and Semantics, Vol. 3: Speech Acts. Academic Press.Hamblin, C.L. (1987). Imperatives. Blackwell.Landragin, F. (2004). Dialogue History Modelling for Multimodal Human-Computer Interaction. In: Eighth Workshop on the Semantics and Pragmatics of Dialogue (CATALOG’04).Perrault, C. & Allen, J. (1980). A Plan-Based Analysis of Indirect Speech Acts. American Journal of Computational Linguistics, 6 (3-4).Portner, P. (2004). The Semantics of Imperatives within a Theory of Clause Types. In: K. Watanabe & R.B. Young (eds), Proceedings of Semantics and Linguistic Theory 14. CLC Publications.Potts, C. (2003). The Logic of Conventional Implicatures. PhD dissertation. University of California.Quignard, M. (2002). A Collaborative Model of Argumentation in Dyadic Problem-Solving Interactions. In: F. van Eemeren, J.A. Blair & C.A. Willard (eds), Proceedings of the Fifth International Conference of the International Society for the Study of Argumentation (ISSA'02). Amsterdam.Sadek, D., Bretier, P. & Panaget, F. (1997). Artimis: Natural Dialogue Meets Rational Agency. IJCAI.Searle, J. (1969). Speech Acts. Cambridge University Press.Searle, J. (1975). Indirect Speech Acts. In: P. Cole & J. Morgan (eds), Syntax and Semantics, Vol. 3: Speech Acts. Academic Press.Searle, J. (1976). A Taxonomy of Illocutionary Acts. In: Gunderson (ed), Language, Mind, and Knowledge. University of Minnesota Press.Sperber, D. & Wilson, D. (1995). Relevance. Communication and Cognition. Blackwell. Xuereb, A. & Caelen, J. (2004). L'interprétation pragmatique en dialogue homme-machine finalisé. Journées scientifiques Sémantique et Modélisation, Lyon.。