IP Library › Granted Patent US 12705420
Granted Patent B2
US 12705420 · App. 18/425,657 · Granted Aug 11, 2026

Systems and methods for deviation detection, information extraction and obligation deviation detection

Inventors: Sally Gao (Boston, MA); Hella-Franziska Hoffmann (London, GB); Nina Hristozova (Zurich, CH); Elizabeth Roman (Somerville, MA); Nicolai Pogrebnyakov (Ontario, CA); Yue Feng (Ontario, CA); Masoud Makrehchi (Ontario, CA); Tate Sterling Avery (Ontario, CA); Shohreh Shaghaghian (Ontario, CA); Borna Jafarpour (Ontario, CA)
Assignee: Thomson Reuters Enterprise Centre GmbH
G06F40/194G06F3/0481G06F40/242G06F40/258G06F40/284G06F40/289G06F40/30G06V30/414
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12705420
App. No.
18/425,657
Granted
Aug 11, 2026
Kind
B2
Abstract

The present disclosure is directed towards systems and methods for detecting deviations between documents and portions thereof, extracting information from text and detecting deviations between obligations. Information is extracted by identifying defined terms and their definitions in input text as well as by identifying portions of different input texts relevant to a point of interest and detecting deviations in those portions between the different input texts.

Claims (39)

1 . A method for extracting information, comprising:

receiving a standard text and a review text;

analyzing the standard text and the review text to identify portions of the standard text and the review text relevant to an input point of interest, wherein analyzing the standard text and the review text comprises:

selecting a first set of portions from the review text and a second set of portions from the standard text,

selecting at least one word of the input point of interest as a query word, and

estimating a likelihood of the query word in document language models associated with respective portions of the first set of portions from the review text and respective portions of the second set of portions from the standard text; and

displaying, based on the likelihood of the query word and via a user interface, a portion of text identified in the review text as relevant to the point of interest while also displaying any deviations therein from a portion of text identified in the standard text as being relevant to the point of interest.

2 . The method of claim 1 , wherein the point of interest is one of a question, a topic, a named fact, and a defined term.

3 . The method of claim 1 , wherein the deviations are not displayed until a user interface interaction is performed with respect to the displayed portion of text identified in the review text as being relevant to the point of interest.

4 . The method of claim 1 , wherein the portion of text identified in the standard text as being relevant to the point of interest is displayed via the user interface upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.

5 . The method of claim 1 , wherein multiple portions of text are identified in the standard text as being relevant to the point of interest and are displayed upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.

6 . The method of claim 5 , wherein any deviations between the portion of text identified in the review text as being relevant to the point of interest and each of the multiple portions of text are identified in the standard text as being relevant to the point of interest and are displayed via the user interface.

7 . The method of claim 1 , wherein estimating the likelihood of the query word in document language models comprises executing a Dirichlet language model for estimating the likelihood of the query word.

8 . The method of claim 1 , wherein the displaying further comprises displaying points of interest via the user interface, wherein the portion of text identified in the review text as being relevant to a particular point of interest is not displayed until a user interface interaction is performed with respect to a display of the particular point of interest via the user interface.

9 . A system comprising memory and one or more processors communicatively coupled to the memory, the one or more processors configured to:

receive a standard text and a review text;

analyze the standard text and the review text to identify portions of the standard text and the review text relevant to an input point of interest, wherein the standard text and the review text are analyzed by:

selecting a first set of portions from the review text and a second set of portions from the standard text,

selecting at least one word of the input point of interest as a query word, and

estimating a likelihood of the query word in document language models associated with respective portions of the first set of portions from the review text and respective portions of the second set of portions from the standard text; and

display, based on the likelihood of the query word and via a user interface, a portion of text identified in the review text as relevant to the point of interest while also displaying any deviations therein from a portion of text identified in the standard text as being relevant to the point of interest.

10 . The system of claim 9 , wherein the point of interest is one of a question, a topic, a named fact, and a defined term.

11 . The system of claim 9 , wherein the deviations are not displayed until a user interface interaction is performed with respect to the displayed portion of text identified in the review text as being relevant to the point of interest.

12 . The system of claim 9 , wherein the portion of text identified in the standard text as being relevant to the point of interest is displayed via the user interface upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.

13 . The system of claim 9 , wherein multiple portions of text are identified in the standard text as being relevant to the point of interest and are displayed upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.

14 . The system of claim 13 , wherein any deviations between the portion of text identified in the review text as being relevant to the point of interest and each of the multiple portions of text are identified in the standard text as being relevant to the point of interest and are displayed via the user interface.

15 . The system of claim 9 , wherein the likelihood of the query word is estimated by executing a Dirichlet language model.

16 . The system of claim 9 , wherein the one or more processors are further configured to:

display points of interest via the user interface, wherein the portion of text identified in the review text as being relevant to a particular point of interest is not displayed until a user interface interaction is performed with respect to a display of the particular point of interest via the user interface.

17 . A computer program product, stored on a non-transitory computer readable medium, comprising instructions that when executed by one or more processors cause the one or more processors to:

receive a standard text and a review text;

analyze the standard text and the review text to identify portions of the standard text and the review text relevant to an input point of interest, wherein the standard text and the review text are analyzed by:

selecting a first set of portions from the review text and a second set of portions from the standard text,

selecting at least one word of the input point of interest as a query word, and

estimating a likelihood of the query word in document language models associated with respective portions of the first set of portions from the review text and respective portions of the second set of portions from the standard text; and

display, based on the likelihood of the query word and via a user interface, a portion of text identified in the review text as relevant to the point of interest while also displaying any deviations therein from a portion of text identified in the standard text as being relevant to the point of interest.

18 . The computer program product of claim 17 , wherein the deviations are not displayed until a user interface interaction is performed with respect to the displayed portion of text identified in the review text as being relevant to the point of interest.

19 . The computer program product of claim 17 , wherein the portion of text identified in the standard text as being relevant to the point of interest is displayed via the user interface upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.

20 . The computer program product of claim 17 , wherein multiple portions of text are identified in the standard text as being relevant to the point of interest and are displayed upon a user interface interaction with respect to the portion of text identified in the review text as being relevant to the point of interest.