CONTEXTUAL ANALYSIS ENGINE

• US 20150106078A1
• Filed: 10/15/2013
• Published: 04/16/2015
• Est. Priority Date: 10/15/2013
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Chang, Walter
• Granted Patent: US 9,990,422 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 16/35 Clustering; Classification

Abstract

A contextual analysis engine systematically extracts, analyzes and organizes digital content stored in an electronic file such as a webpage. Content can be extracted using a text extraction module which is capable of separating the content which is to be analyzed from less meaningful content such as format specifications and programming scripts. The resulting unstructured corpus of plain text can then be passed to a text analytics module capable of generating a structured categorization of topics included within the content. This structured categorization can be organized based on a content topic ontology which may have been previously defined or which may be developed in real-time. The systems disclosed herein optionally include an input/output interface capable of managing workflows of the text extraction module and the text analytics module, administering a cache of previously generated results, and interfacing with other applications that leverage the disclosed contextual analysis services.

Claims

1. A method of analyzing digital content, the method comprising:

receiving a corpus of text;

extracting a plurality of n-grams from the corpus of text;

constructing a multi-dimensional document feature vector based on the plurality of n-grams;

categorizing a plurality of topics derived from the multi-dimensional document feature vector based on a topic ontology; and

generating a hierarchical listing of the plurality of categorized topics that includes a relevancy ranking for at least a portion of the topics included within the hierarchical listing.

2. The method of claim 1, wherein the multi-dimensional document feature vector is further based on a frequency of each of the plurality of extracted n-grams.

3. The method of claim 1, wherein the corpus of text is unstructured.

4. The method of claim 1, wherein the corpus of text is received from a text extraction module that includes a headless browser.

5. The method of claim 1, wherein constructing the multi-dimensional feature vector further comprises ranking the plurality of n-grams according to a relevance factor that is based on a frequency at which a given n-gram appears within the corpus of text.

6. The method of claim 1, wherein the plurality of topics derived from the multi-dimensional feature vector are selected by identifying (a) n-grams within the topic ontology and (b) topics form the topic ontology that have a parent relationship with the n-grams identified in (a).

7. The method of claim 1, wherein the plurality of topics derived from the multi-dimensional feature vector are selected by identifying (a) n-grams within the topic ontology and (b) topics form the topic ontology that have a direct parent relationship with the n-grams identified in (a).

8. The method of claim 1, wherein the plurality of topics derived from the multi-dimensional feature vector are selected by identifying (a) n-grams within the topic ontology and (b) topics form the topic ontology that have an indirect parent relationship with the n-grams identified in (a).

9. The method of claim 1, wherein the hierarchical listing of the plurality of categorized topics further includes a frequency count for at least a portion of the topics included within the hierarchical listing.

10. The method of claim 1, wherein the multi-dimensional document feature vector does not include any n-grams that were not extracted from the corpus of plain text.

11. The method of claim 1, wherein at least one of the plurality of topics is not included within the plurality of n-grams extracted from the corpus of text.

12. A method of delivering targeted content to a consumer, the method comprising:

generating contextual analysis data for a plurality of webpages that are made available to a consumer via a website, wherein the contextual analysis data includes a hierarchical listing of topics with corresponding relevancy rankings for each of the plurality of webpages;

compiling website visitor log information indicating which of the plurality of webpages have been visited by the consumer;

analyzing the contextual analysis data corresponding to the webpages visited by the consumer to identify at least one topic as being relevant to the consumer; and

identifying targeted content to be delivered to the consumer, wherein the targeted content is related to the at least one relevant topic.

13. The method of claim 12, wherein:

at least one of the plurality of webpages includes a content topic that is dynamically generated when the webpage is rendered; and

the dynamically-generated content topic is included within the hierarchical listing of topics.

14. The method of claim 12, wherein the relevancy ranking for each of the plurality of webpages is calculated using a topic frequency/inverse document frequency technique.

15. The method of claim 12, further comprising:

sending a requested webpage to the consumer, the requested webpage including a marketing box; and

sending the targeted content to the consumer, wherein the targeted content is configured to be displayed in the marketing box.

16. A system for analyzing digital content, the system comprising:

an n-gram extractor configured to extract a plurality of n-grams from an unstructured corpus of text;

a topic model generator configured to construct a multi-dimensional feature vector based on the plurality of n-grams;

a topic categorizer configured to categorize a plurality of topics derived from the multi-dimensional document feature vector based on a topic ontology, wherein at least one of the plurality of categorized topics is not included within the plurality of n-grams extracted from the unstructured corpus of text.

17. The system of claim 16, wherein the n-gram extractor is further configured to generate frequency data corresponding to the extracted plurality of n-grams, the frequency data comprising a frequency count.

18. The system of claim 16, wherein the n-gram extractor is further configured to filter one or more stop words from the extracted plurality of n-grams.

19. The system of claim 16, wherein the topic categorizer is configured to categorize the plurality of topics based on a plurality of topic ontologies.

20. The system of claim 16, further comprising a named entity extractor and a natural language parser, wherein:

the named entity extractor is configured to extract additional topics from the unstructured corpus of text based on one or more tagged noun expressions identified by the natural language parser; and

the topic model generator is configured to construct the multi-dimensional feature vector based also on the additional extracted topics.

21. A non-transient computer readable medium having instructions encoded thereon that, when executed by one or more processors, causes a digital content analysis process to be carried out, the process comprising:

receiving an unstructured corpus of text;

extracting a plurality of n-grams from the unstructured corpus of text;

constructing a multi-dimensional document feature vector based on the plurality of n-grams;

categorizing a plurality of topics derived from the multi-dimensional document feature vector based on a topic ontology; and

generating a hierarchical listing of the plurality of categorized topics that includes a relevancy ranking for at least a portion of the topics included within the hierarchical listing.

22. The non-transient computer readable medium of claim 21, wherein constructing the multi-dimensional feature vector comprises ranking the plurality of n-grams according to a relevance factor that is at least partially based on a spatial distribution of a given n-gram within the unstructured corpus of text.

23. The method of claim 21, wherein the plurality of topics derived from the multi-dimensional feature vector are selected by identifying (a) n-grams within the topic ontology and (b) topics form the topic ontology that have a parent relationship with the n-grams identified in (a).

24. The method of claim 21, wherein a quantity of categorized topics in the hierarchical listing is greater than a quantity of n-grams extracted from the corpus of text.