System and method for determining intentions of messages

Patent metadata

Field	Value
Publication number	US-2023222291-A1
Application number	US-202217571766-A
Country	US
Kind code	A1
Filing date	Jan 10, 2022
Priority date	Jan 10, 2022
Publication date	Jul 13, 2023
Grant date	—

Abstract

One or more computing devices, systems, and/or methods for determining an intention of a message and/or performing one or more operations based upon the intention are provided. In an example, a first set of text may be received, from a client device, via a messaging interface. Based upon the first set of text, one or more sets of text semantically similar to the first set of text may be determined. Based upon the first set of text and the one or more sets of text a plurality of intention predictions may be determined. A first intention of the first set of text may be determined based upon the plurality of intention predictions. One or more operations may be performed based upon the first intention.

First claim

What is claimed is:

1 . A method comprising:
receiving, from a client device, a first set of text via a messaging interface; determining, based upon the first set of text, one or more sets of text semantically similar to the first set of text; determining, based upon the first set of text and the one or more sets of text, a plurality of intention predictions comprising:
a first intention prediction based upon the first set of text; and
one or more second intention predictions generated based upon the one or more sets of text, wherein a second intention prediction of the one or more second intention predictions is generated based upon a second set of text of the one or more sets of text;
determining a first intention, of the first set of text, based upon the plurality of intention predictions; and performing one or more operations based upon the first intention.

2 . The method of claim 1 , wherein determining the one or more sets of text comprises:
determining, based upon the first set of text, a plurality of sets of text; determining, based upon the first set of text and the plurality of sets of text, a plurality of similarity scores associated with the plurality of sets of text, wherein each similarity score of the plurality of similarity scores corresponds to a measure of semantic similarity between the first set of text and a set of text of the plurality of sets of text; and selecting, based upon the plurality of similarity scores, the one or more sets of text from the plurality of sets of text.

3 . The method of claim 2 , wherein:
selecting the one or more sets of text from the plurality of sets of text is performed based upon a determination that similarity scores, of the plurality of similarity scores, associated with the one or more sets of text meet a threshold similarity score.

4 . The method of claim 2 , wherein:
a quantity of sets of text of the one or more sets of text is a defined number k; and selecting the one or more sets of text from the plurality of sets of text is performed based upon a determination that the one or more sets of text are associated with k highest similarity scores of the plurality of similarity scores.

5 . The method of claim 1 , wherein:
determining the one or more sets of text is performed based upon information comprising at least one of:
historical information comprising text received from client devices;
text of an encyclopedia;
text of news articles; or
text associated with a field associated with the messaging interface.

6 . The method of claim 1 , wherein:
determining the one or more sets of text is performed using a first machine learning model.

7 . The method of claim 6 , further comprising:
training a machine learning model using training information to generate the first machine learning model, wherein the training information comprises at least one of:
historical information comprising text received from client devices;
text of an encyclopedia;
text of news articles; or
text associated with a field associated with the messaging interface.

8 . The method of claim 1 , wherein each intention prediction of the plurality of intention predictions comprises:
an indication of an intention; and a confidence score associated with the intention prediction.

9 . The method of claim 1 , wherein:
each intention prediction of the plurality of intention predictions comprises an indication of an intention; the method comprises determining one or more intention prediction counts associated with one or more intentions comprising the first intention; each intention prediction count of the one or more intention prediction counts indicates a quantity of one or more intention predictions, of the plurality of intention predictions, indicative of an intention of the one or more intentions; and determining the first intention of the first set of text is based upon the one or more intention prediction counts.

10 . The method of claim 9 , wherein:
determining the first intention of the first set of text is based upon a determination that a first intention prediction count, of the one or more intention prediction counts, associated with the first intention corresponds to a majority of the plurality of intention predictions.

11 . The method of claim 9 , wherein:
determining the first intention of the first set of text is based upon a determination that a first intention prediction count, of the one or more intention prediction counts, associated with the first intention is the highest intention prediction count of the one or more intention prediction counts.

12 . The method of claim 8 , wherein:
first intention predictions, of the plurality of intention predictions, are indicative of the first intention; and determining the first intention of the first set of text is based upon confidence scores of the first intention predictions.

13 . The method of claim 9 , wherein:
each intention prediction of the plurality of intention predictions comprises a confidence score associated with the intention prediction; first intention predictions, of the plurality of intention predictions, are indicative of the first intention; determining the first intention of the first set of text is based upon confidence scores of the first intention predictions based upon at least one of:
a determination that a first intention prediction count, of the one or more intention prediction counts, associated with the first intention does not correspond to a majority of the plurality of intention predictions; or
a determination that the first intention prediction count is equal to a second intention prediction count, of the one or more intention prediction counts, associated with a second intention of the one or more intentions.

14 . The method of claim 9 , wherein:
each intention prediction of the plurality of intention predictions comprises a confidence score associated with the intention prediction; the one or more intention prediction counts comprise a first intention prediction count associated with the first intention and a second intention prediction count associated with a second intention of the one or more intentions; the first intention prediction count is equal to the second intention prediction count; first intention predictions, of the plurality of intention predictions, are indicative of the first intention; second intention predictions, of the plurality of intention predictions, are indicative of the second intention; and determining the first intention of the first set of text comprises:
determining a first score associated with the first intention based upon confidence scores of the first intention predictions;
determining a second score associated with the second intention based upon confidence scores of the second intention predictions; and
selecting the first intention based upon the first score and the second score.

15 . The method of claim 1 , wherein:
the first intention corresponds to a request for information; and the one or more operations comprise transmitting the information to the client device.

16 . The method of claim 1 , wherein:
the first intention corresponds to a request to perform an action; and the one or more operations comprise performing the action.

17 . A non-transitory computer-readable medium storing instructions that when executed perform operations comprising:
receiving, from a client device, a first set of text via a messaging interface; determining, based upon the first set of text, one or more sets of text semantic

Assignees

• Verizon Patent & Licensing Inc

Inventors

• Ranganathan Prakash

Classifications

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title