System and method for generating automated response to an input query received from a user in a human-machine interaction environment

A system for generating automated response to an input query received from a user in a human-machine interaction environment, the system comprising:

• an external memory, wherein the external memory comprises data stored in a hierarchical structure, and wherein the data is segregated into a plurality of segments in the hierarchical structure;
• a processor in communication with the external memory;
• and
• a memory coupled with the processor, wherein the memory comprises a plurality of modules capable of being executed by the processor, the plurality of modules comprising:
• a segment identification module configured to identify a relevant segment, from the plurality of segments, matching with a user input;
• a relevant data determination module configured to determine relevant data within the relevant segment matching with the user input;
• and
• a response generation module configured to generate a response with respect to the user input based upon the relevant data;
• wherein the data comprises user context data and system knowledge data, wherein the user context data is generated based upon a user conversation history, a user profile and any other user-related information;
• wherein the data is stored in form of a plurality of data vectors, and wherein the plurality of segments is stored in form of a plurality of segment vectors such that each segment vector comprises a subset of data vectors;
• wherein each segment vector is generated based on one or more hashing functions via hashing technique, wherein each hashing function enables segregation of data vectors into multiple buckets, and wherein each hashing function, each bucket and each data vector has a hashing index, a bucket index and a data index, respectively, such that a data vector segregated into a bucket is capable of being retrieved from the external memory using a combination of the hash index, the bucket index and the data index, and wherein each segment vector has a segment index, the segment index being a concatenation of the hashing index and the bucket index;
• wherein the segment identification module identifies the relevant segment by
• encoding the user input into a segment query vector;
• encoding each segment vector into a segment matching vector;
• computing a similarity score for each segment matching vector with respect to the segment query vector using a cosine similarity;
• and
• calculating a probability distribution over the segments in the segment matching vectors based upon the similarity score thereby selecting the relevant segment, wherein the relevant segment is selected based upon random sampling technique;
• wherein the relevant data determination module determines the relevant data within the relevant segment by
• retrieving data vectors corresponding to the relevant segment;
• encoding the user input into a data query vector;
• computing a similarity score for each entry in the data vectors with respect to the data query vector using a cosine similarity;
• calculating a probability distribution over the data in the data vectors based upon the similarity score;
• encoding each data vector into a data output matching vector;
• computing a weighted sum of entries in each data output matching vector based upon the probability distribution calculated over the data;
• and
• obtaining a data output vector based upon the weighted sum of the entries in each data output matching vector.