Utilizing a gated self-attention memory network model for predicting a candidate answer match to a query

A system comprising:

• one or more memory devices comprising:
• a gated-self attention mechanism (GSAM);
• a memory network;
• a query;
• and
• a plurality of candidate answers to the query;
• and
• at least one server configured to cause the system to:
• generate a context vector for a plurality of context vector states based on contextual information associated with one or more utterances;
• generate input vectors based on combinations of respective candidate answers of the plurality of candidate answers and the query;
• for each combination of a respective candidate answer and the query:
• generate a next cell state of the memory network based on the GSAM, a current context vector state, and a current cell state of the memory network;
• generate a next context vector state based on a current gated self-attention output vector state of the GSAM, the current context vector state, and the next cell state of the memory network;
• and
• determine, based on the next context vector state, a matching probability between the query and the respective candidate answer;
• and
• determine, based on the matching probabilities for the combinations of the respective candidate answers and the query, an answer to the query from the plurality of candidate answers.

A computer-implemented method comprising:

• generating a context vector for a plurality of context vector states based on contextual information associated with one or more utterances;
• generating input vectors based on combinations of respective candidate answers of a plurality of candidate answers and a query;
• for each combination of a respective candidate answer and the query:
• generating, based on a gated self-attention mechanism (GSAM) and the context vector, a gated self-attention output vector;
• generating, based on the gated self-attention output vector, a next cell state of a memory network;
• generating, based on a current gated self-attention output vector state of the GSAM and the next cell state of the memory network, a next context vector state;
• and
• determining, based on the next context vector state, a matching probability between the query and the respective candidate answer;
• and
• determining, based on the matching probabilities for the combinations of the respective candidate answers and the query, an answer to the query from the plurality of candidate answers.