Composite propensity profile detector

Patent metadata

Field	Value
Publication number	US-10733286-B2
Application number	US-201615042651-A
Country	US
Kind code	B2
Filing date	Feb 12, 2016
Priority date	Aug 20, 2013
Publication date	Aug 4, 2020
Grant date	Aug 4, 2020

Abstract

Detecting propensity profile for a person may comprise receiving artifacts associated with the person; detecting profile characteristics for the person based on the artifacts; receiving a plurality of predefined profiles comprising a plurality of characteristics and relationships between the characteristics over time, each of the plurality of predefined profiles specifying an indication of propensity; matching the profile characteristics for the person with one or more of the plurality of predefined profiles; and outputting one or more propensity indicators based on the matching, the propensity indicators comprising at least an expressed strength of a given propensity in the person at a given time.

First claim

We claim:

1. A method for detecting propensity profile of a person, comprising:
receiving artifacts associated with the person;
detecting, by one or more computer processors, profile characteristics for the person based on the artifacts, at least by tokenizing unstructured text data of the artifacts, part-of-speech tagging of the unstructured text data, and parsing the unstructured text data;
receiving a plurality of predefined profiles comprising a plurality of characteristics and relationships between the characteristics over time, each of the plurality of predefined profiles specifying an indication of propensity, wherein at least one of the plurality of predefined profiles includes a machine learning computer model learned based on annotated data, the computer model updated periodically based on feedback;
matching, by one or more of the computer processor, executing a machine learning algorithm, the profile characteristics for the person with one or more of the plurality of predefined profiles;
outputting one or more propensity indicators based on the matching, the propensity indicators comprising at least an expressed strength of a given propensity in the person at a given time; and
adding one or more latent characteristics to the profile characteristics determined based on performing a what-if analysis, for detecting whether combining the one or more latent characteristics with the profile characteristics at the given time would trigger a propensity indication which would not be triggered by the profile characteristics.

2. The method of claim 1 , wherein the artifacts comprise email, instant messaging messages, web site content, or social media content, or combinations thereof.

3. The method of claim 1 , wherein the detecting profile characteristics comprises one or more of machine learning, detecting by a rule-based method, or detecting by dictionary lookup, or combinations thereof.

4. The method of claim 1 , wherein the detecting profile characteristics comprises: extracting features from the artifacts;
applying rules based on a predefined dictionary to the extracted features to detect profile characteristics; and
scoring the profile characteristics to determine a degree to which the profile characteristics are present in the artifacts.

5. The method of claim 4 , wherein the extracting features from the artifacts comprises:
extracting lexical features comprising at least uppercase words, repeated punctuation, and repeated letters;
extracting emoticons;
extracting positive and negative sentiment words; and
performing a sentiment analysis.

6. The method of claim 5 , wherein the performing a sentiment analysis comprises:
performing psycholinguistic analysis; and
a dictionary-based semantic identification.

7. A system for detecting propensity profile for a person, comprising:
a processor comprising at least hardware;
the processor configured to at least:
detect profile characteristics for the person based on a plurality of received artifacts, at least by tokenizing unstructured text data of the artifacts, part-of-speech tagging of the unstructured text data, and parsing the unstructured text data; and
match by executing a machine learning algorithm, the profile characteristics for the person with one or more of a plurality of predefined profiles, the plurality of predefined profiles comprising a plurality of characteristics and relationships between the characteristics over time, each of the plurality of predefined profiles specifying an indication of propensity, wherein at least one of the plurality of predefined profiles includes a machine learning computer model learned based on annotated data, the computer model updated periodically based on feedback, the processor further configured to output one or more propensity indicators based on the matching, the propensity indicators comprising at least an expressed strength of a given propensity in the person at a given time,
wherein one or more latent characteristics determined based on performing a what-if analysis is added to the profile characteristics, for detecting whether combining the one or more latent characteristics with the profile characteristics at the given time would trigger a propensity indication which would not be triggered by the profile characteristics.

8. The system of claim 7 , wherein one or more latent characteristics are added to the profile characteristics for matching with the one or more of a plurality of predefined profiles.

9. The system of claim 7 , wherein the artifacts comprise email, instant messaging messages, web site content, or social media content, or combinations thereof.

10. The system of claim 7 , wherein the processor detects profile characteristics by one or more of machine learning, a rule-based method, or a dictionary lookup, or combinations thereof.

11. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions readable by a processor to cause the processor to perform a method comprising:
receiving artifacts associated with the person;
detecting profile characteristics for the person based on the artifacts, at least by tokenizing unstructured text data of the artifacts, part-of-speech tagging of the unstructured text data, and parsing the unstructured text data;
receiving a plurality of predefined profiles comprising a plurality of characteristics and relationships between the characteristics over time, each of the plurality of predefined profiles specifying an indication of propensity, wherein at least one of the plurality of predefined profiles includes a machine learning computer model learned based on annotated data, the computer model updated periodically based on feedback;
matching by a machine learning algorithm, the profile characteristics for the person with one or more of the plurality of predefined profiles;
outputting one or more propensity indicators based on the matching, the propensity indicators comprising at least an expressed strength of a given propensity in the person at a given time; and
adding one or more latent characteristics to the profile characteristics determined based on performing a what-if analysis, for detecting whether combining the one or more latent characteristics with the profile characteristics at the given time would trigger a propensity indication which would not be triggered by the profile characteristics.

12. The computer program product of claim 11 , wherein the artifacts comprise email, instant messaging messages, web site content, or social media content, or combinations thereof.

13. The computer program product of claim 11 , wherein the detecting profile characteristics comprises one or more of machine learning, detecting by a rule-based method, or detecting by dictionary lookup, or combinations thereof.

14. The computer program product of claim 11 , wherein the detecting profile characteristics comprises:
extracting features from the artifacts;
applying rules based on a predefined dictionary to the extracted features to detect profile characteristics; and
scoring the profile characteristics to determine a degree to which the profile characteristics are present in the artifacts.

15. The computer program product of claim 14 , wherein the extracting features from the artifacts comprises:
extracting lexical features comprising at least uppercase words, repeated punctuation, and repeated letters;
extracting emoticons;
extracting positive and negative sentiment words; and
perfor

Assignees

• IBM

Inventors

• Coden Anni R
• Houck Keith C
• Lin Ching-Yung
• Lin Wanyi
• Malkin Peter K
• Pan Shimei
• Park Youngja
• Weisz Justin D

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• H04L63/1433

  Vulnerability analysis · CPC title

• G06N7/04

  Physical realisation · CPC title