interaction_model

Interaction model.

The interaction model is responsible for defining the allowed transitions between dialogue acts based on their intents and updating the agenda.

Attributes

logger

Classes

InteractionModel

Initializes the interaction model.

Module Contents

interaction_model.logger

class interaction_model.InteractionModel(config_file: str, domain: usersimcrs.core.simulation_domain.SimulationDomain, annotated_conversations: List[dialoguekit.core.dialogue.Dialogue])

Initializes the interaction model.

Parameters:

• config_file – Path to configuration file.

• domain – Simulation domain.

• annotated_conversations – Annotated conversations.

REQUIRED_INTENTS

initialize_transition_matrices(annotated_conversations: List[dialoguekit.core.dialogue.Dialogue]) → Tuple[pandas.DataFrame, pandas.DataFrame]

Initializes transition matrices from annotated conversations.

We consider two transition matrices. The first one uses single intents as states, while the second one uses compound intents. For example: Dialogue acts in utt. 1: [GREETING(), REQUEST(genre=?, year=?)] Dialogue acts in utt. 2: [INFORM(genre=action, year=2024)]

The single intent transition matrix will be: GREETING -> INFORM : 1 REQUEST -> INFORM : 1 The compound intent transition matrix will be: GREETING_REQUEST -> INFORM : 1

Note that the compound intent may also include single intents, in case an utterance has a single dialogue act.

Parameters:

annotated_conversations – Annotated conversations.

Returns:

Transition matrices.

initialize_agenda(information_need: usersimcrs.core.information_need.InformationNeed)

Initializes user agenda.

Parameters:

information_need – Information need.

is_agent_intent_elicit(agent_intent: dialoguekit.core.intent.Intent) → bool

Checks if the given agent intent is elicitation.

Parameters:

agent_intent – Agent’s intent.

Returns:

True if it is an elicitation intent.

is_agent_intent_set_retrieval(agent_intent: dialoguekit.core.intent.Intent) → bool

Checks if the given agent intent is set retrieval.

Parameters:

agent_intent – Agent’s intent.

Returns:

True if it is a set retrieval intent.

is_agent_intent_inquire(agent_intent: dialoguekit.core.intent.Intent) → bool

Checks if the given agent intent is inquiry.

Parameters:

agent_intent – Agent’s intent.

Returns:

True if it is an inquiry intent.

get_next_dialogue_acts(n: int = 1) → List[dialoguekit.core.dialogue_act.DialogueAct]

Returns the next n dialogue acts from the stack.

Parameters:

n – Number of dialogue acts to return. Defaults to 1.

Returns:

List of dialogue acts.

update_agenda(information_need: usersimcrs.core.information_need.InformationNeed, preference_model: usersimcrs.user_modeling.preference_model.PreferenceModel, item_collection: usersimcrs.items.item_collection.ItemCollection) → None

Updates the agenda based on the last agent dialogue acts and state.

Each agent dialogue act results in a push operation on the stack. We consider four cases: the agent elicits, recommends, inquires, or neither. Once the push operations are done, we clean the stack.

Parameters:

information_need – Information need.