Theme detection for object-recognition-based notifications
AssigneeCapital One Services, LLC
InventorsJoshua Edwards, Michael Mossoba, Abdelkader Benkreira
Life of this patent
Estimating the term…
Abstract
In certain embodiments, speech is converted to text for theme identification by natural language processing. Notification data is generated based on detected themes and the notification data may include rules for notification presentation on a client device. The notification data may include parameters for processing image data captured by an augmented reality device to detect one or more objects. The objects may be associated with the theme and detection thereof within captured image data, and in accordance with other rules, may cause the augmented reality device to present a notification with contextual relevance to a current environment of a user utilizing the augmented reality device.
Claims20
A system for generating notifications based on themes detected by natural language processing, comprising:
- one or more processors programmed with computer program instructions that, when executed, cause the system to:
- obtain a natural language processing model configured to provide an output indicative of one or more themes identified in a set of chunks of input natural language text, wherein:
- the output indicative of the one or more themes is associated with theme information for at least one candidate theme based on scores of candidate themes identified in the set of chunks of input natural language text,
- the candidate themes are identified by the natural language processing model, based on the set of chunks of input natural language text, from a set of possible themes represented in training records identifying themes within corresponding ones of, or sets of, training chunks of natural language text, and
- at least some of the sets of training chunks of natural language text respectively include encoded links between training chunks in the set that are indicative of a shared theme in combination and not individually indicative of the shared theme;
- receive a plurality of requests, each request including at least one segment of audio data;
- convert, for each request, one or more segments of audio data into corresponding chunks of natural language text;
- determine, for at least some requests, one or more links between two or more of the corresponding chunks of natural language text;
- and
- process, with the natural language processing model, for a given one of the at least some requests, the two or more corresponding chunks of natural language text to identify at least one theme for the given one of the requests for presenting associated theme information to a user, wherein the at least one theme is selected responsive to scores of candidate themes that include at least some shared theme candidates.
A non-transitory computer readable medium comprising computer-program instructions for generating notifications based on themes detected by natural language processing, the instructions when executed by one or more processors causing the one or more processors to perform operations to:
- obtain a natural language processing model configured to provide an output indicative of one or more themes identified in a set of chunks of input natural language text, wherein:
- the output indicative of the one or more themes is associated with theme information for at least one candidate theme based on scores of candidate themes identified in the set of chunks of input natural language text,
- the candidate themes are identified by the natural language processing model, based on the set of chunks of input natural language text, from a set of possible themes represented in training records identifying themes within corresponding ones of, or sets of, training chunks of natural language text, and
- at least some of the sets of training chunks of natural language text respectively include encoded links between training chunks in the set that are indicative of a shared theme in combination and not individually indicative of the shared theme;
- receive a plurality of requests, each request including at least one segment of audio data;
- convert, for each request, one or more segments of audio data into corresponding chunks of natural language text;
- determine, for at least some requests, one or more links between two or more of the corresponding chunks of natural language text;
- and
- process, with the natural language processing model, for a given one of the at least some requests, the two or more corresponding chunks of natural language text to identify at least one theme for the given one of the requests for presenting associated theme information to a user, wherein the at least one theme is selected responsive to scores of candidate themes that include at least some shared theme candidates.
A non-transitory computer readable medium comprising computer-program instructions that when executed by one or more processors cause the one or more processors to perform operations to:
- obtain a training set comprising training chunks of natural language text and training records identifying themes within corresponding training chunks of natural language text, wherein:
- at least some of the training chunks of natural language text are linked training chunks of natural language text;
- and
- at least some sets of linked training chunks of natural language text correspond to a respective conversation having a central theme identified by corresponding training records encoding at least one link for each linked training chunk and at least two links for at least some of the linked training chunks to other ones of the linked training chunks;
- train on the training set with one or more natural language processing algorithms to generate a natural language processing model configured to provide an output indicative of whether a theme appears in a set of chunks of natural language text;
- and
- transmit, to a user device, the natural language processing model to cause the user device to:
- classify segments of audio data as likely including speech;
- convert the segments of audio data into a corresponding plurality of chunks of natural language text;
- process, with the natural language processing model, a plurality of chunks of natural language text to identify at least one theme;
- and
- generate a notification based on the at least one theme.
Dependencies are inferred from the claim text.