Dynamic message categorization for optimized message targeting

Patent metadata

Field	Value
Publication number	US-10929773-B2
Application number	US-201715431326-A
Country	US
Kind code	B2
Filing date	Feb 13, 2017
Priority date	Dec 14, 2016
Publication date	Feb 23, 2021
Grant date	Feb 23, 2021

Abstract

An automated dynamic message categorization system is provided and includes first, second and third processing units. The first processing unit is configured to generate a user interface (UI) and to present the UI to a user. The second processing unit is configured to pull information from a first textual element which has been entered into the UI, to identify second textual elements that are relevant to the first textual element based on the pulled information and to extract textual element identifiers from the second textual elements. The third processing unit is configured to generate, for each extracted textual element identifier, a confidence score describing a degree of correlation between each extracted textual element identifier and the first textual element. The first processing unit is further configured to present to the user each extracted textual element identifier with a corresponding confidence score as a selectable option via the UI.

First claim

What is claimed is:

1. A method of automated dynamic message categorization, comprising:
presenting a user interface (UI) to a user;
pulling information from a first textual element which has been entered into the UI;
identifying second textual elements that are relevant to the first textual element based on the pulled information;
extracting textual element identifiers from the second textual elements, wherein the extracting comprises:
interpreting user intent behind text of the first textual element and returning a corresponding classification of that text based on the results of that interpretation by a natural language classifier; and
using linguistic analysis to detect emotion, social tendencies and language style from the text, the emotions being anger, fear, joy, sadness and disgust, the social tendencies being openness, conscientiousness, extroversion, agreeableness and emotional range and the language styles being confident, analytical and tentative by a tone analyzer;

generating for each extracted textual element identifier a confidence score describing a degree of correlation between each extracted textual element identifier and the first textual element, wherein the degree of correlation is based on emotive, social tendency and language style overlaps between each extracted textual element identifier and the first textual element; and
presenting to the user each extracted textual element identifier with a corresponding confidence score as a selectable option via the UI, wherein:
respective outputs of the natural language classifier and the tone analyzer are aggregated with timestamps, user identification and historical data as metadata,
the pulled information comprises the metadata and representative portions of the text of the first textual element,
the pulled information is dynamically loaded into and purged from a first database based on characteristics of the pulled information to increase a likelihood of the second textual elements being found,
an output portion of the first database is populated with the second textual elements and at least the extracted textual element identifiers, and
a second database is populated by an updateable and modifiable listing of generic tags and hashtags.

2. The method according to claim 1 , wherein the pulling of information comprises natural language classification and tone analysis of the first textual element.

3. The method according to claim 1 , wherein the first textual element comprises a tweet.

4. The method according to claim 3 , wherein each extracted textual element identifier comprises a tag or a hashtag.

5. The method according to claim 3 , further comprising:
maintaining a database of generic tags and hashtags; and
discarding an extracted textual element identifier comprising any one or more of the generic tags or hashtags included in the database.

6. The method according to claim 1 , further comprising:
identifying emotive, social tendency and language style overlaps between each extracted textual element identifier and the first textual element; and
basing the degree of correlation on the identified emotive, social tendency and language style overlaps.

Assignees

• IBM

Inventors

• Chakra Al
• Harpur Liam
• Patel Sumit
• Sexton Enda

Classifications

• G06F16/38

  Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually · CPC title

• H04L51/52

  for supporting social networking services · CPC title

• H04L51/214

  using selective forwarding · CPC title

• G06Q30/0241

  Advertisements · CPC title

• G06F40/20

  Natural language analysis (semantic analysis of natural language G06F40/30) · CPC title