US 11,947,919 B2
Interactive dialogue system and a method for facilitating human machine conversation
Sridhar Gadi, Maharashtra (IN); Manish Kumar, Maharashtra (IN); Pavan Jakati, Maharashtra (IN); and Neeraj Pandey, Maharashtra (IN)
Assigned to Zensar Technologies Limited, Maharashtra (IN)
Filed by Zensar Technologies Limited, Maharashtra (IN)
Filed on Mar. 19, 2021, as Appl. No. 17/206,828.
Claims priority of application No. 202021031942 (IN), filed on Jul. 25, 2020.
Prior Publication US 2022/0027573 A1, Jan. 27, 2022
Int. Cl. G06F 40/35 (2020.01); G06F 16/33 (2019.01); G06F 16/903 (2019.01); G06F 16/906 (2019.01); G06F 16/951 (2019.01); G06N 20/20 (2019.01)
CPC G06F 40/35 (2020.01) [G06F 16/3344 (2019.01); G06F 16/90335 (2019.01); G06F 16/906 (2019.01); G06F 16/951 (2019.01); G06N 20/20 (2019.01)] 8 Claims
OG exemplary drawing
 
1. A method for facilitating a human-machine conversation for responding to a user query received from a user, the method comprising:
determining at least one of a category and a context of the user query, wherein the category and the context relates to a domain in which the human-machine conversation is provided;
implementing a first model and a second model for responding to the user query, wherein the first model is associated with at least one dataset comprising one or more pre-stored queries and corresponding one or more responses, and wherein the first model is further configured to implement one or more machine learning techniques comprising random forest classifier, stochastic gradient descent classifier, and AdaBoost classifier upon the at least one dataset, and wherein the first model further comprising generating a first model score comprising a summation of an accuracy score, a first document score, and a first user score, wherein
the first model score is generated when the first model is capable of providing the relevant information in response to the user query, and wherein:
the accuracy score is generated based on the implementation of the at least one of the machine leaning techniques upon the at least one dataset,
the first document score is generated by:

OG Complex Work Unit Math
wherein, DS1 represents the first document score,
IOQ represents index of words for the user query, and
IOR represents index of the relevant information provided for the user query, and
the user score is generated based on user feedback upon receiving the relevant information provided by the first model, and
wherein the second model is triggered when the first model is unable to respond to the user query, and wherein, upon triggering the second model, the method further comprising:
configuring one or more crawlers for accessing corresponding one or more databases of a plurality of databases, based on the at least one of the category and the context of the user query, for retrieving the relevant information;
generating a second model score for the relevant information retrieved by the second model; and
enabling the second model to present the relevant information to the user based on the second model score, wherein the second model updates the first model with the relevant information and its corresponding second model score such that the first model is capable of responding to a future query similar to the user query using the updated relevant information.