Machine learning based intent resolution for a digital assistant
AssigneeSap SE
InventorJulian Seibel
Life of this patent
Estimating the term…
Abstract
Disclosed herein are system, method, and computer program product embodiments for training a machine learning model and using the machine learning model to determine an intent associated with an utterance. An embodiment operates by receiving an utterance, determining a first vector based on the received utterance, and determining a second vector based on the received utterance. A third vector is generated by concatenating the first vector and the second vector. The third vector is used to perform at least one of classifying, using a machine learning model, the utterance to determine the intent associated with the utterance or training the machine-learning model.
Claims15
A computer implemented method comprising:
- receiving, by an intent resolution system, an utterance that comprises two words;
- determining, by the intent resolution system, a first vector, based on the received utterance, for a first word of the two words using a term frequency-inverse document frequency (TF-IDF);
- determining, by the intent resolution system, a second vector, based on the received utterance, for the first word of the two words using a Global Vectors (GloVe) model;
- generating, by the intent resolution system, a third vector for the first word of the two words by concatenating the first vector and the second vector;
- determining, by the intent resolution system, a fourth vector for a second word of the two words using the TF-IDF;
- determining, by the intent resolution system, a fifth vector for the second word of the two words using the GloVe model;
- generating, by the intent resolution system, a sixth vector for the second word of the two words by combining the fourth vector and the fifth vector;
- generating, by the intent resolution system, an average vector by averaging the third vector and the sixth vector;
- and
- using the average vector to classify, using a machine learning model, the utterance to determine an intent associated with the utterance.
A system comprising:
- a memory;
- and
- at least one processor coupled to the memory and configured to:
- determine, based on a received utterance that comprises two words, a first vector for a first word of the two words using a term frequency-inverse document frequency (TF-IDF);
- determine, based on the received utterance, a second vector for the first word of the two words using a Global Vectors (GloVe) model;
- generate a third vector for the first word of the two words by combining the first vector and the second vector;
- determine a fourth vector for a second word of the two words using the TF-IDF;
- determine a fifth vector for the second word of the two words using the GloVe model;
- generate a sixth vector for the second word by combining the fourth vector and the fifth vector;
- generate an average vector by averaging the third vector and the sixth vector;
- and
- use the average vector to classify, using a machine learning model, the utterance to determine an intent associated with the utterance.
A non-transitory computer-readable device having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations comprising:
- determining, based on a received utterance that comprises two words, a first vector for a first word of the two words using a term frequency-inverse document frequency (TF-IDF);
- determining, based on the received utterance, a second vector for the first word of the two words using a Global Vectors (GloVe) model;
- generating a third vector for the first word of the two words by combining the first vector and the second vector;
- determining a fourth vector for a second word of the two words using the TF-IDF;
- determining a fifth vector for the second word of the two words using the GloVe model;
- generating a sixth vector for the second word by combining the fourth vector and the fifth vector;
- generating an average vector by averaging the third vector and the sixth vector;
- and
- using the average vector to perform at least one of:
- classifying, using a machine learning model, the utterance to determine an intent associated with the utterance;
- or
- training the machine learning model, wherein the machine learning model is used to determine the intent associated with the utterance,
- wherein the machine learning model comprises a support-vector machine classifier.
Dependencies are inferred from the claim text.