Patent Issued for Natural language processing platform for automated event analysis, translation, and transcription verification (USPTO 11947872): Allstate Insurance Company

2024 APR 23 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- According to news reporting originating from Alexandria, Virginia, by NewsRx journalists, a patent by the inventors Cluck, Matthew (Schaumburg, IL, US), Colgan, Geoff (Arlington Heights, IL, US), Eley, Ross Kirk (Belfast, IE), Mahler-Haug, Olivia (Chicago, IL, US), Walsh, Connor (Lake Forest, IL, US), Yeddula, Harideep (Chicago, IL, US), filed on May 14, 2020, was published online on April 2, 2024.

The assignee for this patent, patent number 11947872, is Allstate Insurance Company (Northbrook, Illinois, United States).

Reporters obtained the following quote from the background information supplied by the inventors: “Aspects of the disclosure relate to enhanced processing systems for performing natural language processing. Many organizations and individuals rely on claim processing services to determine fault and provide compensation accordingly. In many instances, however, claim processing may be inefficient due to extensive manual review of the claims. There remains an ever-present need to develop improved methods of improving claim analysis using natural language processing.

“In addition, many organizations and individuals rely on call processing services from a variety of call center settings. For example, they may rely on claim processing services to determine fault and provide compensation accordingly. In many instances, however, claim processing may be inefficient due to the use of foreign languages during a claim processing call. There remains an ever-present need to develop improved methods of handling the use of foreign languages during such calls.”

In addition to obtaining background information on this patent, NewsRx editors also obtained the inventors’ summary information for this patent: “Aspects of the disclosure provide effective, efficient, scalable, and convenient technical solutions that address and overcome the technical problems associated with the processing of calls (e.g., such as claims processing calls). In accordance with one or more arrangements discussed herein, a computing platform having at least one processor, a communication interface, and memory may determine one or more utterance segments based on a received audio file. The computing platform may send one or more audio chunks based on the one or more utterance segments to an audio transcription system. The computing platform may receive one or more text segments in response to the one or more audio chunks. Based on the one or more text segments, the computing platform may generate an audio transcription file. The computing platform may assign a category to each word in the audio transcription file. The computing platform may send user interface information, generated based on the category assigned to each word in the audio transcription file, to a user device along with one or more commands directing the user device to generate a user interface based on the user interface information, which may cause the user device to generate and display the user interface.

“In one or more instances, the computing platform may identify, in the audio transcription file, personal information, which may be credit card information, bank account information, a social security number, driver’s license information, or a tax identifier. The computing platform may redact the personal information from the audio transcription file and the one or more audio chunks, which may be linked to the audio transcription file.

“In one or more instances, the computing platform may identify a claim number in the audio transcription file by: 1) determining that a number in the audio transcription file matches a pattern corresponding to a claim number based on one or more of: a number of digits, a digit pattern, or a character pattern, 2) extracting a segment from the audio transcription file corresponding to a predetermined character length window, 3) analyzing the extracted segment to identify that a length of numbers included in the extracted segment is greater than or equal to an amount of numbers in a claim number, and 4) eliminating identified repetition in the extracted segment until the length of numbers is reduced to the amount of numbers in the claim number.

“In one or more instances, the claim number may be broken up by one or more words and the computing platform may identify the claim number by determining, based on the length of numbers, that the extracted segment contains the claim number. In one or more instances, the computing platform may be configured to implement logic to interpret number variations in the audio transcription file. In these instances, the number variations may be leading zeros or varied number pronunciations.

“In one or more instances, the received audio file may include a first channel and a second channel. In these instances, the first channel may include speech from a first individual and the second channel may include speech from a second individual. In one or more instances, the generated one or more audio chunks may include a first subset and a second subset. In these instances, the first subset of the one or more audio chunks may include speech and the second subset of the one or more audio chunks may include silence. Further, in these instances, the computing platform may send the first subset of the one or more audio chunks without sending the second subset of the one or more audio chunks.”

The claims supplied by the inventors are:

“1. A computing platform, comprising: at least one processor; a communication interface communicatively coupled to the at least one processor; and memory storing computer-readable instructions that, when executed by the at least one processor, cause the computing platform to: determine one or more utterance segments based on a received audio file; generate one or more audio chunks based upon the one or more utterance segments, wherein the generated one or more audio chunks comprise a first subset and a second subset, wherein the first subset of the one or more audio chunks includes speech and the second subset of the one or more audio chunks includes silence; send the first subset of the one or more audio chunks without sending the second subset of the one or more audio chunks to an audio transcription system; receive one or more text segments in response to the one or more audio chunks; generate an audio transcription file based on the one or more text segments; assign a category to each word in the audio transcription file; determine, based on the audio transcription file, identifying information indicative of a user account; request additional information associated with the user account; send user interface information, generated based on the category assigned to each word in the audio transcription file and the additional information, to a user device along with one or more commands directing the user device to generate a user interface based on the user interface information, wherein sending the one or more commands directing the user device to generate the user interface based on the user interface information causes the user device to generate and display the user interface; and receive a selection of text from the audio transcription file, via the user interface, and a request to play a portion of the audio file corresponding to the selected text.

“2. The computing platform of claim 1, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to: identify, in the audio transcription file, personal information, wherein the personal information comprises one or more of: credit card information, bank account information, a social security number, driver’s license information, or a tax identifier; and redact the personal information from the audio transcription file and the one or more audio chunks, wherein the one or more audio chunks are linked to the audio transcription file.

“3. The computing platform of claim 1, wherein wherein determining, based on the audio transcription file, identifying information indicative of a user account in the audio transcription file comprises: determining that a number in the audio transcription file matches a pattern corresponding to a claim number based on one or more of: a number of digits, a digit pattern, or a character pattern, extracting a segment from the audio transcription file corresponding to a predetermined character length window, analyzing the extracted segment to identify that a length of numbers included in the extracted segment is greater than or equal to an amount of numbers in a claim number, eliminating identified repetition in the extracted segment until the length of numbers is reduced to the amount of numbers in the claim number.

“4. The computing platform of claim 3, wherein the claim number is broken up by one or more words and wherein identifying the claim number comprises determining, based on the length of numbers, that the extracted segment contains the claim number.

“5. The computing platform of claim 3, wherein the computing platform is configured to implement logic to interpret number variations in the audio transcription file, wherein the number variations comprise leading zeros or varied number pronunciations.

“6. The computing platform of claim 1, wherein the received audio file comprises a first channel and a second channel, and the first channel includes speech from a first individual and the second channel includes speech from a second individual.

“7. The computing platform of claim 1, wherein: each of the one or more text segments contains text corresponding to at least one of the one or more audio chunks, the one or more text segments are generated by simultaneous processing of the one or more audio chunks, each of the one or more text segments includes one or more embedded timestamps or speaker identifiers, and the one or more text segments are received in real time as they are generated.

“8. The computing platform of claim 7, wherein generating the audio transcription file comprises: stitching together the one or more text segments using the one or more embedded timestamps or speaker identifiers, and linking each of the one or more text segments to one or more corresponding audio chunks, wherein a user input received at a portion of the audio transcription file corresponding to a particular text segment may result in audio output of one or more particular audio chunks used to generate the particular text segment.

“9. The computing platform of claim 1, wherein assigning the categories to each word in the audio transcription file comprises, for each word: computing, using a transfer learning model, a probability that the word corresponds to each of a plurality of categories, wherein the plurality of categories comprises: preamble, loss details, injury, damages, contact details, next steps, sign off, or other: generating a vector including the computed probabilities that the word corresponds to each of the plurality of categories, wherein the plurality of categories comprises: preamble, loss details, injury, damages, contact details, next steps, sign off, or other; generating a vector including the computed probabilities that the word corresponds to each of the plurality of categories; assigning, to the word, a category corresponding to a largest computed probability included in the vector; and tagging the word with an indication of the assigned category.

“10. The computing platform of claim 9, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to tag, within the audio transcription file, one or more sections, wherein each of the one or more tagged sections includes words corresponding to a common category.

“11. The computing platform of claim 9, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to modify an assigned category of a word based on a category of one or more adjacent words.

“12. The computing platform of claim 9, wherein each of the plurality of categories corresponds to a priority value, and wherein assigning the category to the word is based, at least in part, on the priority value.

“13. The computing platform of claim 9, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to generate, for each of the plurality of categories, a count, wherein the counts correspond to a number of times in the audio transcription file a word in the corresponding category appears.

“14. The computing platform of claim 1, wherein the memory stores additional computer-readable instructions that, when executed by the at least one processor, further cause the computing platform to: identify, using the audio transcription file, one or more of: a call type, a loss location, or additional participants corresponding to the audio transcription file.

“15. The computing platform of claim 1, wherein sending the one or more commands directing the user device to generate the user interface based on the user interface information causes the user device to generate and display an interface that includes: a line by line script indicating words spoken by a particular speaker, an indication of the particular speaker, and a time at which the words were spoken, and a series of selectable objects each corresponding to one of a plurality of categories, wherein selection of one of the selectable objects causes words in the line by line script, which correspond to a category of the one of the selectable objects, to be emphasized.

“16. The computing platform of claim 15, wherein the received audio file comprises a live audio stream, and wherein the user interface is displayed in real time during the live audio stream.”

There are additional claims. Please visit full patent to read further.

For more information, see this patent: Cluck, Matthew. Natural language processing platform for automated event analysis, translation, and transcription verification. U.S. Patent Number 11947872, filed May 14, 2020, and published online on April 2, 2024. Patent URL (for desktop use only): https://ppubs.uspto.gov/pubwebapp/external.html?q=(11947872)&db=USPAT&type=ids

(Our reports deliver fact-based news of research and discoveries from around the world.)