| Adjacency pairs and insertions | What is the appropriate way of responding to a given input? How can one correct what one has said, ask for additional information in order to and before one can provide the appropriate response, etc.?
|
| Argumentation acts | They represent higher goals and are made up of Core speech acts. For example, an Inform core speech act may serve the argumentation act of summarising what was previously said.
|
| Assertives | Assertives:They commit the speaker to something being the case. The different kinds are: suggesting, putting forward, swearing, boasting, concluding. Example: No one makes a better cake than me.
|
| automaton | An abstract machine consisting of a set of states (including the initial state), an input tape (and possibly an output tape), a state transition function, and some designated ?terminal states. When modeling a dialogue with an automaton, the key idea is to think of the states of that automaton as standing for different states of the dialogue, and of its edges as corresponding to things that happen in the dialogue.
|
| Backward Communicative Functions | They account for the relation of the current utterance to the dialogue up to that point.
|
| chart parser | Chart parsing is based on the idea to keep a record (a chart) of all the structures found during parsing. It is a very general idea, and it can be used with just about any kind of parsing (or recognition) strategy (top down, bottom up, left-corner, depth first, breadth first, ...).
|
| Commisives | They commit the speaker to doing something in the future. The different kinds are: promising, planning, vowing, betting, opposing. Example: "I'm going to Paris tomorrow."
|
| Conversation Acts | Special kind of speech acts used in TRAINS
|
| Core speech acts | These are the traditional speech acts like Inform, Request, Promise, wh-questions etc.
|
| DAMSL | The Dialogue Act Markup in Several Layers is different from many previous annotation schemes, in the sense that it was not created to fit the needs of any particular domain or corpus. It was proposed as the standard annotation scheme by the DRI (Discourse Resource Initiative) for dialogue tagging. Therefore, it is domain and task independent.
|
| Declarations | They change the state of the world in an immediate way. Examples: ``You are fired, I swear, I beg you''.
|
| dialogue acts | In order to model dialogue by use of speech acts there is a need to enrich the original notion and do away with the simplifying assumptions that it presupposes. In other words, the multiple function of a single utterance at the speech act level has to be accounted for. Moreover, communication cannot be assumed, understanding has to be established. Finally, speech acts cannot be interpreted outside the context in which they appear as they are commonly only one part of a higher, more complicated goal. This shift from original speech acts is captured in naming acts used in dialogue modeling dialogue acts.
|
| Dialogue context | How do dialogue participants use the contributions and the conclusions previously drawn to interpret the current contribution and decide on their next contribution?
|
| Dialogue phases | They represent stages in the dialogue. Dialogue acts are marked for the phases in which they can legally appear and dialogue phases consist of dialogue acts. Protocols of behaviour can be generated based on the current dialogue phase. There are five phases: Hello: greeting and introduction on both sides. Opening: introduction of the topic for negotiation. Negotiation: everything that the negotiation involves. Closing: recapitulation of the already agreed upon topic. Goodbye: both participants say goodbye.
|
| Directives | They try to make the addressee perform an action. The different kinds are: asking, ordering, requesting, inviting, advising, begging. Example: "Could you close the window?"
|
| Discourse Unit | A Discourse Unit represents the aggregate of acts necessary for establishing a piece of information. It is an approach rooted in the cooperative nature of dialogue. There are four basic kinds of conversation acts:
|
| Edges | Edges (also known as transitions) of a dialogue automaton may for instance be defined to take the user input into account in order to decide what the new system state is, and to activate certain procedures, for instance output or movements. That way whenever a particular input is recognised by the system, a predefined behaviour and the way to realise it can be produced next.
|
| Eliza | Eliza is one of the first dialogue systems. You can regard it as your friend. Tell it your problems and it will try to help you. It cannot really understand what you tell it. It picks on words that it then maps onto specific output.
|
| Ellipsis | How do dialogue participants make sense of fragmentary responses that are often provided in dialogue situations?
|
| Expressives | They express how the speaker feels about the situation. The different kinds are: thanking, apologising, welcoming, deploring. Example: "I am sorry that I lied to you."
|
| Forward Communicative Functions | Forward Communicative Functions: They correspond, by and large, to the traditional speech acts.
|
| FSA | See ↗automaton
|
| general dialogue characteristics | Besides giving insight into how dialogues work and how they're structured, these characteristics serve as beackround for the evaluation and classification of dialogue systems.
|
| Grounding | How does a dialogue participant make sure that their contribution is rightly understood, or that they themselves have understood correctly a previous contribution? How can misinterpretations be corrected?
|
| Grounding acts | Grounding acts are used for establishing information between the participants. They include categories like initiate, continue, acknowledge, repair etc.
|
| HCRC map task | The HCRC map task (see HCRC page) annotated corpus consists of transcripts of spontaneous task-oriented dialogues between two human agents. One agent holds a map with landmarks and a route that he has to communicate to the other. The second participant holds a slightly different map without the route. The task is to reconstruct the route on the second map.
|
| Illocutionary | Illocutionaryi s the act performed in saying something. The illocutionary act is not in one-to-one correspondence with the locution from which it is derived. There are different locutions that express the same illocution and vice-versa. For example, there are indirect speech acts, that is acts with a different force than the obviously deducible one. A typical example is the locution of the utterance Could you pass the salt? uttered at a dinner table. For a speaker of English in the particular situation this means Pass the salt, please and no one would assume that the speaker is indeed interested in whether the addressee would be able to pass the salt.
|
| Locutionary | Locutionary is the act of actually uttering.
|
| Mixed initiative | Turn-taking: When, how, and for how long should each dialogue participant talk?
|
| Performatives | Verbs that name the speech act that they intend to effect are called Performatives. A performative uttered by the right person under the right circumstances has as a result a change in the world. For example, "I pronounce you husband and wife" uttered by a priest, in the church with all the legal and traditional aspects being settled, will have the actual effect of the couple referred to being husband and wife after the performative has taken place.
|
| Perlocutionary | Perlocutionary is the act performed by saying something in a particular context. It represents the change achieved each time, in a particular context. Depending on the kind of perlocution, different conditions have to hold in order for it to be achieved. For example, the addressee in the salt example has to realise that the speaker's intention is to ultimately get hold of the salt.
|
| Planning schemes | These should encompass the top level of how far into the dialogue the system should plan. That means, deciding if planning the speech acts of the next utterance is enough, or if planning further ahead is necessary for better communication. The other level to be considered is that which connects whatever it is to be communicated with the realisation that achieves that aim.
|
| recognition grammar | Set of rules defining what utterances can be recognized by a dialogue system.
|
| Reference resolution | How can participants disambiguate what referring expressions refer to?
|
| speech acts | The idea of speech acts has its roots in the Philosophy of Language. J. A. Austin () was the first one who wanted to capture the fact that there is more in the function of language than semantics. Traditionally, mapping of entities of a proposition onto referents and defining the truth value of a proposition was the major area of interest in language semantics. With Austin, and his follower J. R. Searle, there is a shift towards the events or acts that occur via language, hence the name speech acts. These acts effect changes both in the observable world, as well as in the mental states of dialogue participants.
|
| speech recogniser | The speech recogniser takes speech as input and it outputs word sequences and messages to draw attention on interpretations that might need revising.
|
| states | See ↗automaton. States in a dialogue system are always defined with certain expectations as to what the system can have had as input, and what else can have happened at certain stages of the dialogue. For instance, the initial state of a dialogue automaton is naturally the beginning of the dialogue, and final states will normally correspond to possible end-points of a dialogue.
|
| TRAINS | The basic dialogue scenario that TRAINS deals with is route-planning in a transportation domain. The user sees a display of a map with cities and their train connections. He is asked to solve logistics problems, taking into account environmental factors as well as the standard connections. Both spoken and typed communication is possible.
|
| transitions | Edges (also known as transitions) of a dialogue automaton may for instance be defined to take the user input into account in order to decide what the new system state is, and to activate certain procedures, for instance output or movements. That way whenever a particular input is recognised by the system, a predefined behaviour and the way to realise it can be produced next.
|
| TRIPS | The TRIPS system was built with an emphasis on modularity. It, therefore, consists of a number of modules that are each responsible for different tasks and inter-communicate to share the necessary information for the overall effect. The communication is again managed by a general purpose manager. This architecture allows the extension of the system by plugging in different autonomous off-the-self resources. There are three main components that are responsible for the communication and management of the rest: The Interpretation Manager, the Behavioural Agent and the Generation Manager. They operate asynchronously.
|
| Turn-taking | Turn-taking: When, how, and for how long should each dialogue participant talk?
|
| Turn taking acts | They are keep-turn, release-turn and its sub-variant assign-turn, and take-turn. They define turn taking.
|
| Utterance Features | They capture information both about the content and the form of utterances. For example, they give information on the relation of the utterance to the communication and task management. An utterance is labeled for its communication management function only when there is no task management function present.
|
| Verbmobil | The Verbmobil (see Verbmobil page) corpus comprises transcripts of human-to-human, spontaneous, task-oriented dialogues. The task is to negotiate on appointment scheduling between the participants. One participant is also able to give travel information on demand by the other. The interactions can be in German, English and Japanese.
|