Dialogue processing system using speech act control and operation method thereof
Disclosed is a dialogue processing system using speech act control, the dialogue processing system comprising: a main speech act unit which processes a free speech act and performs speech act control such that the free speech act returns to a main speech act, thereby processing a multi-turn dialogue in a consistent manner, and which processes a purposed utterance having a set purpose, for reaching a final dialogue objective; and a free speech act unit which processes a free utterance deviating from the purposed utterance and performs control such that the free utterance returns to the main speech act unit by searching for a node capable of returning to the purposed utterance.
1. A dialogue processing system using speech act control, comprising:
a main speech act unit which processes a purposed utterance having a set purpose, and
a free speech act unit which (i) finds candidate return inducing paths which connect between the free utterance and the purposed utterance, and (ii) selects an optimal path among the candidate return inducing paths, and (iii) processes the free utterance using the optimal path,
wherein the optimal path is determined by using (i) the number of nodes placed between the free utterance and the purposed utterance, (ii) similarity between the free utterance and the purposed utterance, (iii) correlation between the response utterance and the last node of the optimal path, or (iv) a combination thereof,
wherein the main speech act unit comprises:
a first main speech act unit which processes the purposed utterance, wherein the purposed utterance is a multi-turn, and
a second main speech act unit which identifies whether or not the purposed utterance is deviated and, when a deviation of the purposed utterance occurs, changes authority of speech act control to the free speech act unit,
wherein the second main speech act unit checks whether or not an utterance of a user is deviated from the purposed utterance and changes the authority of the speech act control to the free speech act unit when the utterance of the user is determined as the free utterance.
2. The system of claim 1 ,
wherein the free utterance is a multi-turn.
3. The system of claim 2 ,
wherein the first free speech act unit processes the free utterance via a chatbot pool consisting of one or more chatbots or a dialogue flow graph and classifies each free speech act to select a chatbot or a response utterance.
4. The system of claim 1 ,
wherein the second free speech act unit searches for the return inducing paths by back-tracking nodes between the free utterance and the utterance.
5. The system of claim 4 ,
wherein the first free speech act unit searches for a response utterance corresponding to a last output in which a topic movement or a topic deviation from the main speech act unit occurred.
6. An operation method of a dialogue processing system using speech act control, comprising the steps of:
upon receipt of a purposed utterance having a set purpose, performing a main speech act,
upon receipt of a free utterance deviating from the purposed utterance, performing a free speech act which includes,
(i) finding candidate return inducing paths which connect between the free utterance and the purposed utterance,
(ii) selecting an optimal path among the candidate return inducing path, and
(iii) processing the free utterance using the optimal path,
wherein the optimal path is determined by using (i) the number of nodes placed between the free utterance and the purposed utterance, (ii) similarity between the free utterance and the purposed utterance, (iii) correlation between the response utterance and the last node of the optimal path, or (iv) a combination thereof,
wherein the step of performing the main speech act comprises:
processing the purposed utterance, wherein the purposed utterance is a multi-turn, and
identifying whether or not the purposed utterance is deviated and
when a deviation of the purposed utterance occurs, changing authority of speech act control to perform the free speech act,
wherein the step of identifying whether or not the purposed utterance is deviated includes: checking whether or not an utterance of a user is deviated from the purposed utterance and changing the authority of the speech act control to perform the free speech act unit when the utterance of the user is determined as the free utterance.