Intent recognition model creation from randomized intent vector proximities

A computer system for automating intent recognition model (IRM) creation comprising:

• a processor;
• and
• a computer-readable storage media coupled to the processor, wherein the computer-readable storage media contains program instructions executing a computer-implemented method comprising the steps of:
• encoding, by the processor, sample text into a sample intent vector comprising a multi-dimensional matrix wherein each word of the sample text corresponds to a dimensional vector;
• generating, by the processor, a Gaussian noise vector based upon the sample intent vector, the Gaussian noise vector comprising a sequence of data elements representing a Gaussian noise distribution relative to respective data elements that form the sample intent vector;
• configuring, by the processor, long short-term memory (LSTM) parameters of a LSTM model;
• parallel processing, by the processor, the sample intent vector and the Gaussian noise vector using the LSTM model to generate a set of candidate intent vectors;
• determining, by the processor, a measurement of a vector distance and confidence level of a candidate intent vector, of the set of candidate intent vectors, relative to the sample intent vector;
• validating, by the processor, the candidate intent vector as valid using filter-based processing, wherein the candidate intent vector is considered valid upon meeting or exceeding a threshold level of confidence and the candidate intent vector is close to the sample intent vector in multidimensional intent vector space;
• and
• in response to validating the candidate intent vector as valid, assigning the candidate intent vector as a final IRM vector by adding the candidate intent vector to the intent recognition model.

A computer program product for automating intent recognition model (IRM) creation comprising:

• one or more computer readable storage media having computer-readable program instructions stored on the one or more computer readable storage media, said program instructions executes a computer-implemented method comprising:
• encoding, by a processor, sample text into a sample intent vector comprising a multi-dimensional matrix wherein each word of the sample text corresponds to a dimensional vector;
• generating, by the processor, a Gaussian noise vector based upon the sample intent vector, the Gaussian noise vector comprising a sequence of data elements representing a Gaussian noise distribution relative to respective data elements that form the sample intent vector;
• configuring, by the processor, long short-term memory (LSTM) parameters of a LSTM model;
• parallel processing, by the processor, the sample intent vector and the Gaussian noise vector using the LSTM model to generate a set of candidate intent vectors;
• determining, by the processor, a measurement of a vector distance and confidence level of a candidate intent vector, of the set of candidate intent vectors, relative to the sample intent vector;
• validating, by the processor, the candidate intent vector as valid using filter-based processing, wherein the candidate intent vector is considered valid upon meeting or exceeding a threshold level of confidence and the candidate intent vector is close to the sample intent vector in multidimensional intent vector space;
• and
• in response to validating the candidate intent vector as valid, assigning the candidate intent vector as a final IRM vector by adding the candidate intent vector to the intent recognition model.