Granted Patent B2

US 12,236,497 · App. 15/494,377 · Granted Feb 25, 2025

Inventors: Brian Grom (Lake Worth, FL); Vladimir Eidelman (Chevy Chase, MD); Daniel Argyle (Alexandria, VA); Jervis Pinto (Md, IN)

Assignee: FiscalNote, Inc.

G06Q50/26G06F3/0482G06F3/04847G06F16/288G06F16/3331G06F40/205G06F40/263G06F40/30G06N5/022G06N5/04G06Q10/06375G06Q50/18G06F16/95G06N20/00G06Q2230/00G06T11/206G06T2200/24

Abstract

Claims (54)

1. A text analytics system for predicting whether a policy will be adopted, the system comprising:

at least one processor configured to:

scrape information from a plurality of sources on the Internet by a web crawler and an extraction bot, wherein the web crawler is configured to perform functions of finding, indexing, and fetching information from the plurality of sources on the Internet, and wherein the extraction bot is configured to perform processing on the information from the plurality of sources to generate text data, the text data being associated with comments authored by a plurality of individuals about a proposed policy;

generate a policy feature vector based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the proposed policy;

generate a plurality of comment feature vectors based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the text data, wherein each of the comment feature vectors is associated with at least one of the comments;

train, using a training set of the text data and an association between each comment feature vector and an output sentiment, a sentiment model for determining comment sentiments, wherein the sentiment model comprises weights computed for one or more input features of the plurality of comment feature vectors using machine learning, each weight reflecting an importance of the-one or more input features;

determine, based on application of the sentiment model to the text data, a sentiment of each comment;

train, using a training set of at least one of the plurality of comment feature vectors, the policy feature vector, and a policy outcome, a policy adoption model to determine the likelihood of adoption of the policy, wherein the policy adoption model comprises weights computed for the one or more input features using machine learning, each weight reflecting an importance of the one or more input features;

determine, based on application of an influence filter to the text data, an influence metric associated with each comment, the influence metric comprising a value indicating a level of influence the author of each comment has on whether the policy is adopted, wherein applying the influence filter includes accessing a database of individual terms associated with a heightened degree of influence, wherein inclusion of one or more of the individual terms in a comment indicates the comment has a higher impact on whether the policy is adopted than if the one or more individual terms were not included;

weigh each comment based on the level of influence using the influence metric;

apply the policy adoption model to the plurality of comment feature vectors to determine a likelihood of adoption of the policy;

determine based on an aggregate of the weighted comments, a first indicator associated with the likelihood of adoption of the policy and a second indicator associated with the weighted comments; and

transmit the indicators to a system user; and

generate a graphical user interface associated with the policy, wherein the graphical user interface includes a comment summary region including representations of each of the comments authored by the plurality of individuals about the policy, wherein each of the representations include a name of an individual author of the comment and an indicator of the sentiment of the comment.

2. The text analytics system of claim 1 , wherein the individual terms include the author and author related characteristics of the comment.

3. The text analytics system of claim 1 , wherein the at least one processor is further configured to apply an association analysis filter to the text data in order to correlate at least a portion of each comment with a particular section of the proposed policy.

4. The text analytics system of claim 1 , wherein applying a text analytics filter to the text data in order to determine a sentiment includes matching at least one piece of text data from the comment to at least one other piece of text data stored in a database.

5. The text analytics system of claim 1 , wherein the policy includes at least one regulation.

6. The text analytics system of claim 1 , wherein the policy includes at least one legislative bill.

7. The text analytics system of claim 1 , wherein the second indicator includes arguments associated with the weighted comments.

8. The text analytics system of claim 1 , wherein the second indicator includes stances taken within the weighted comments.

9. The text analytics system of claim 1 , wherein at least one individual term of the individual terms is included in the database based on a user input.

10. The text analytics system of claim 1 , wherein at least one individual term of the individual terms is included in the database independent of input from a user.

11. A method for predicting whether a policy will be adopted, the method comprising:

scraping information from a plurality of sources on the Internet by a web crawler and an extraction bot, wherein the web crawler is configured to perform functions of finding, indexing, and fetching information from the plurality of sources on the Internet, and wherein the extraction bot is configured to perform processing on the information from the plurality of sources to generate text data, the text data being associated with comments authored by a plurality of individuals about a proposed policy;

generating policy feature vector based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the proposed policy;

generating a plurality of comment feature vectors based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the text data, wherein each of the comment feature vectors is associated with at least one of the comments;

training, using a training set of the text data and an association between each comment feature vector and an output sentiment, a sentiment model for determining comment sentiments, classifier in order to determine a sentiment of each comment wherein the sentiment model classifier comprises weights computed for one or more input features of the plurality of comment feature vectors using machine learning, each weight reflecting an importance of the one or more input features;

determining, based on application of the sentiment model to the text data, a sentiment of each comment;

training, using a training set of at least one of the plurality of comment feature vectors, the policy feature vector, and a policy outcome, a policy adoption model to determine the likelihood of adoption of the policy, wherein the policy adoption model comprises weights computed for the one or more input features using machine learning, each weight reflecting an importance of the one or more input features;

determining, based on application of an influence filter to the text data, an influence metric associated with each comment, the influence metric comprising a value indicating a level of influence the author of each comment has on whether the policy is adopted, wherein applying the influence filter includes accessing a database of individual terms associated with a heightened degree of influence, wherein inclusion of one or more of the individual terms in a comment indicates the comment has a higher impact on whether the policy is adopted than if the one or more individual terms were not included;

weighing each comment based on the level of influence using the influence metric;

applying the policy adoption model to the plurality of comment feature vectors to determine a likelihood of adoption of the policy;

determining based on an aggregate of the weighted comments, a first indicator associated with the likelihood of adoption of the policy and a second indicator associated with the weighted comments; and

transmitting the indicators to a system user; and

generating a graphical user interface associated with the policy, wherein the graphical user interface includes a comment summary region including representations of each of the comments authored by the plurality of individuals about the policy, wherein each of the representations include a name of an individual author of the comment and an indicator of the sentiment of the comment.

12. The method of claim 11 , wherein the individual terms include the author and author related characteristics of the comment.

13. The method of claim 11 , wherein the method further comprises: applying an association analysis filter to the text data; and correlating at least a portion of each comment with a particular section of the proposed policy.

14. The method of claim 11 , wherein applying a text analytics filter to the text data in order to determine a sentiment further comprises: matching at least one piece of text data from the comment to at least one other piece of text data stored in a database.

15. The method of claim 11 , wherein the policy includes at least one regulation.

16. The method of claim 11 , wherein the policy includes at least one legislative bill.

17. A non-transitory computer readable medium storing a program causing a computer to execute a process, the process comprising:

scraping information from a plurality of sources on the Internet by a web crawler and an extraction bot, wherein the web crawler is configured to perform functions of finding, indexing, and fetching information from the plurality of sources on the Internet, and wherein the extraction bot is configured to perform processing on the information from the plurality of sources to generate text data, the text data being associated with comments authored by a plurality of individuals about a proposed policy;

generating policy feature vector based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the proposed policy;

generating a plurality of comment feature vectors based on at least one of words, phrases, sentences, paragraphs, pages, metadata, linguistic patterns, syntactic parsing, or tone associated with the text data, wherein each of the comment feature vectors is associated with at least one of the comments;

training, using a training set of the text data and an association between each comment feature vector and an output sentiment, a sentiment model for determining comment sentiments, wherein the sentiment model comprises weights computed for one or more input features of the plurality of comment feature vectors using machine learning, each weight reflecting an importance of the one or more input features;

determining, based on application of the sentiment model to the text data, a sentiment of each comment;

training, using a training set of at least one of the plurality of comment feature vectors, the policy feature vector, and a policy outcome, a policy adoption model to determine the likelihood of adoption of the policy, wherein the policy adoption model comprises weights computed for the one or more input features using machine learning, each weight reflecting an importance of the one or more input features;

determining, based on application of an influence filter to the text data, an influence metric associated with each comment, the influence metric comprising a value indicating a level of influence the author of each comment has on whether the policy is adopted, wherein applying the influence filter includes accessing a database of individual terms associated with a heightened degree of influence, wherein inclusion of one or more of the individual terms in a comment indicates the comment has a higher impact on whether the policy is adopted than if the one or more individual terms were not included;

weighing each comment based on the level of influence using the influence metric;

applying the policy adoption model to the plurality of comment feature vectors to determine a likelihood of adoption of the policy;

determining based on an aggregate of the weighted comments, a first indicator associated with the likelihood of adoption of the policy and a second indicator associated with the weighted comments; and

transmitting the indicators to a system user; and

generating a graphical user interface associated with the policy, wherein the graphical user interface includes a comment summary region including representations of each of the comments authored by the plurality of individuals about the policy, wherein each of the representations include a name of an individual author of the comment and an indicator of the sentiment of the comment.

Assignments (6)

Continuity (2)

References Cited (77)

Cited By (1)

US 12,627,569