Systems and methods for analyzing veracity of statements
View Patent ↗The present embodiments may relate to secondary systems that verify potential fraud or the absence thereof. Artificial intelligence and/or chatbots may be employed to verify veracity of statements used in connection with insurance or loan applications, and/or insurance claims. For instance, a veracity analyzer (VA) computing device includes a processor in communication with a memory device, and may be configured to: (1) generate at least one model by analyzing a plurality of historical statements to identify a plurality of reference indicators correlating to inaccuracy of a historical statement; (2) receive a data stream corresponding to a current statement; (3) parse the data stream using the at least one model to identify at least one candidate indicator included in the current statement matching at least one of the plurality of reference indicators; and/or (4) flag, in response to identifying the at least one candidate indicator, the current statement as potentially false.
1 . A veracity analyzer (VA) computing device comprising at least one processor in communication with a memory device, the at least one processor configured to:
using machine learning or artificial intelligence (AI) tools and a plurality of historical text, video, and/or audio statements some of which include inaccurate aspects as a model training input, generate at least one machine learning model, each of the at least one machine learning model being a text analysis model, a body language analysis model, or an inflection of voice analysis model, each machine learning model generated to (i) analyze a plurality of text data, a plurality of voice data, or a plurality of audio data, and (ii) output one of a plurality of reference indicators associated with at least one inaccurate aspect of a user statement made via text, audio, or video, the plurality of reference indicators generated based upon an analysis of a plurality of historical statements using the machine learning or AI tools;
receive other text, video, and/or other audio statements;
identify one or more inaccurate aspects of the other text, video, and/or audio statements;
further train the at least one machine learning model using the machine learning or AI tools and the other text, video, and/or audio statements including the identified one or more inaccurate aspects;
receive a data stream including one or more types of data including text data, video data, and/or audio data corresponding to a current statement made by a first user;
input at least a portion of data from the data stream into one of the at least one trained machine learning model configured to analyze of one of the one or more types of data of the portion of data;
output, from the one trained machine learning model, at least one candidate indicator included in the portion of data from the data stream matching at least one of the plurality of reference indicators, the at least one candidate indicator indicating that the portion of data includes one or more inaccurate aspects of the at least one inaccurate aspect included in the plurality of historical text, video, and/or audio statements;
generate, in real-time and in response to outputting the at least one candidate indicator, a flag for each segment of the portion of data including the at least one candidate indicator, wherein each flagged portion includes the one or more inaccurate aspects;
display, on a user device of a second user, an alert message including each flagged segment of the portion of data including the one or more inaccurate aspects; and
display, in conjunction with the alert message, a prompt to the second user to perform at least one action to address the one or more inaccurate aspects of each flagged segment.
2 . The VA computing device of claim 1 , wherein the at least one processor is further configured to:
retrieve the plurality of historical text, video, and/or audio statements for generating the at least one machine learning model; and
store the at least one generated machine learning model in the memory device.
3 . The VA computing device of claim 1 , wherein the at least one processor is further configured to:
output, using the machine learning or AI tools and the plurality of historical text, video, and/or audio statements, the plurality of reference indicators including at least one of audio reference indicators and visual reference indicators, wherein the audio reference indicators include voice inflections and tones that are indicators of an inaccuracy, wherein the visual reference indicators include body language that is an indicator of an inaccuracy, and wherein the least one candidate indicator matches at least one of (i) the body language of the visual reference indicators, or (ii) at least one of the voice inflections or the tones of the audio reference indicators.
4 . The VA computing device of claim 1 , wherein the at least one processor is further configured to display, on the user device of the second user, the alert message including a respective candidate indicator associated with each flagged portion.
5 . The VA computing device of claim 1 , wherein the at least one action includes asking the first user follow-up questions related to at least one of forensic evidence or differences in statements between the first user.
6 . The VA computing device of claim 1 , wherein the at least one processor is further configured to:
identify a conflict by comparing the data stream and a previous data stream;
in response to identifying the conflict, generate a conflict flag for the data stream; and
display a conflict alert message including the conflict flag of the data stream and the previous data stream.
7 . The VA computing device of claim 1 , wherein the at least one processor is further configured to:
receive additional video data and/or additional audio statements;
re-train the at least one trained machine learning model by inputting the additional video data and/or additional audio data into the trained machine learning model; and
output, from the at least one re-trained machine learning model, one or more candidate indicators included in a received data stream matching at least one of the plurality of reference indicators, wherein the one or more candidate indicators indicate one or more inaccurate aspects included in the received data stream.
8 . A computer-implemented method using a veracity analyzer (VA) computing device including at least one processor in communication with a memory device, the method comprising:
using machine learning or artificial intelligence (AI) tools and a plurality of historical text, video, and/or audio statements some of which include inaccurate aspects as a model training input, generating at least one machine learning model, each of the at least one machine learning model being a text analysis model, a body language analysis model, or an inflection of voice analysis model, each machine learning model generated to (i) analyze a plurality of text data, a plurality of voice data, or a plurality of audio data, and (ii) output one of a plurality of reference indicators associated with at least one inaccurate aspect of a user statement made via text, audio, or video, the plurality of reference indicators generated based upon an analysis of a plurality of historical statements using the machine learning or AI tools;
receiving other text, video, and/or other audio statements;
identifying one or more inaccurate aspects of the other text, video, and/or audio statements;
further training the at least one machine learning model using the machine learning or AI tools and the other text, video, and/or audio statements including the identified one or more inaccurate aspects;
receiving a data stream including one or more types of data including text data, video data, and/or audio data corresponding to a current statement made by a first user;
inputting at least a portion of data from the data stream into one of the at least one trained machine learning model configured to analyze of one of the one or more types of data of the portion of data;
outputting, from the one trained machine learning model, at least one candidate indicator included in the portion of data from the data stream matching at least one of the plurality of reference indicators, the at least one candidate indicator indicating that the portion of data includes one or more inaccurate aspects of the at least one inaccurate aspect included in the plurality of historical text, video, and/or audio statements;
generating, in real-time and in response to outputting the at least one candidate indicator, a flag for each segment of the portion of data including the at least one candidate indicator, wherein each flagged portion includes the one or more inaccurate aspects;
displaying, on a user device of a second user, an alert message including each flagged segment of the portion of data including the one or more inaccurate aspects; and
displaying, in conjunction with the alert message, a prompt to the second user to perform at least one action to address the one or more inaccurate aspects of each flagged segment.
9 . The computer-implemented method of claim 8 further comprising:
retrieving the plurality of historical text, video, and/or audio statements for generating the at least one machine learning model; and
storing the at least one generated machine learning model in the memory device.
10 . The computer-implemented method of claim 8 further comprising:
outputting, using the machine learning or AI tools and the plurality of historical text, video, and/or audio statements, the plurality of reference indicators including at least one of audio reference indicators and visual reference indicators, wherein the audio reference indicators include voice inflections and tones that are indicators of an inaccuracy, and wherein the visual reference indicators include body language that is an indicator of an inaccuracy.
11 . The computer-implemented method of claim 10 , wherein the at least one candidate indicator matches at least one of (i) a body language of the visual reference indicators, or (ii) at least one of a voice inflection or a tone of the audio reference indicators.
12 . The computer-implemented method of claim 8 further comprising displaying, on the user device of the second user, the alert message including a respective candidate indicator associated with each flagged portion.
13 . The computer-implemented method of claim 8 , wherein the at least one action includes asking the first user follow-up questions related to at least one of forensic evidence or differences in statements between the first user.
14 . The computer-implemented method of claim 8 further comprising:
identifying a conflict by comparing the data stream and a previous data stream;
in response to identifying the conflict, generating a conflict flag for the data stream; and
displaying a conflict alert message including the conflict flag of the data stream and the previous data stream.
15 . At least one non-transitory computer-readable storage medium having computer-executable instructions embodied thereon, wherein when executed by a veracity analyzer (VA) computing device including at least one processor in communication with a memory device, the computer-executable instructions cause the at least one processor to:
using machine learning or artificial intelligence (AI) tools and a plurality of historical text, video, and/or audio statements some of which include inaccurate aspects as a model training input, generate at least one machine learning model, each of the at least one machine learning model being a text analysis model, a body language analysis model, or an inflection of voice analysis model, each machine learning model generated to (i) analyze a plurality of text data, a plurality of voice data, or a plurality of audio data, and (ii) output one of a plurality of reference indicators associated with at least one inaccurate aspect of a user statement made via text, audio, or video, the plurality of reference indicators generated based upon an analysis of a plurality of historical statements using the machine learning or AI tools;
receive other text, video, and/or other audio statements;
identify one or more inaccurate aspects of the other text, video, and/or audio statements;
further train the at least one machine learning model using the machine learning or AI tools and the other text, video, and/or audio statements including the identified one or more inaccurate aspects;
receive a data stream including one or more types of data including text data, video data, and/or audio data corresponding to a current statement made by a first user;
input at least a portion of data from the data stream into one of the at least one trained machine learning model configured to analyze of one of the one or more types of data of the portion of data;
output, from the one trained machine learning model, at least one candidate indicator included in the portion of data from the data stream matching at least one of the plurality of reference indicators, the at least one candidate indicator indicating that the portion of data includes one or more inaccurate aspects of the at least one inaccurate aspect included in the plurality of historical text, video, and/or audio statements;
generate, in real-time and in response to outputting the at least one candidate indicator, a flag for each portion segment of the portion of data including the at least one candidate indicator, wherein each flagged portion includes the one or more inaccurate aspects;
display, on a user device of a second user, an alert message including each flagged segment of the portion of data including the one or more inaccurate aspects; and
display, in conjunction with the alert message, a prompt to the second user to perform at least one action to address the one or more inaccurate aspects of each flagged segment.
16 . The computer-readable storage medium of claim 15 , wherein the computer-executable instructions further cause the at least one processor to:
retrieve the plurality of historical text, video, and/or audio statements for generating the at least one machine learning model; and
store the at least one generated machine learning model in the memory device.
17 . The computer-readable storage medium of claim 15 , wherein the computer-executable instructions further cause the at least one processor to:
output, using the machine learning or AI tools and the plurality of historical text, video, and/or audio statements, the plurality of reference indicators including at least one of audio reference indicators and visual reference indicators, wherein the audio reference indicators include voice inflections and tones that are indicators of an inaccuracy, and wherein the visual reference indicators include body language that is an indicator of an inaccuracy.
18 . The computer-readable storage medium of claim 17 , wherein the at least one candidate indicator matches at least one of (i) a body language of the visual reference indicators, or (ii) at least one of a voice inflection or a tone of the audio reference indicators.
19 . The computer-readable storage medium of claim 15 , wherein the computer-executable instructions further cause the at least one processor to display, on the user device of the second user, the alert message including a respective candidate indicator associated with each flagged portion.
20 . The computer-readable storage medium of claim 15 , wherein the computer-executable instructions further cause the at least one processor to:
identify a conflict by comparing the data stream and a previous data stream;
in response to identifying the conflict, generate a conflict flag for the data stream; and
display a conflict alert message including the conflict flag of the data stream and the previous data stream.