Dialogue Representation¶
This page presents the core components of DialogueKit to represent a dialogue and its elements.
Utterance¶
Dialogue participants exchange utterances, which are represented as raw text.
AnnotatedUtterance¶
dialoguekit.core.annotated_utterance
This class enriches Utterance with an intent, annotations, and/or freely definable metadata.
Intent¶
The intent represents the action expressed by a participant in an utterance. For example, an agent asking the question “Do you like pizza?” may have the intent to INQUIRE a preference from the user.
Domain¶
Defines the types of entities and the set of properties (“slots”) for each entity type.
Annotations¶
By default, an annotation has a slot and a value, for example, the slot price has the value $10 in the following utterance “This dish costs $10.”.
DialogueKit is shipped with:
SlotValueAnnotation
dialoguekit.core.slot_value_annotation: slot-value pairs, where a slot refers to an entity or a property in the domain.
Other types of annotation can be implemented by inheriting from dialoguekit.core.annotation.Annotation.
Feedback¶
This object represents user’s feedback. Currently only binary feedback (like/dislike) is supported on the utterance level. Later, it might be extended to graded feedback as well as conversation-level feedback.
Dialogue¶
This object contains all the information related to the dialogue, such as the utterances, the identity of participants, and other metadata.