Automatic semantic rating and abstraction of literature

• US 9,535,901 B2
• Filed: 09/13/2014
• Issued: 01/03/2017
• Est. Priority Date: 02/20/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Allen, Corville Orain, Carrier, Scott Robert, Woods, Eric
• Primary Examiner(s): PATEL, SHREYANS A
• Application Number:
  US14/485,692
• Publication Number:
  US 20140379331A1
• Time in Patent Office: 843 Days
• Field of Search: 704/9
• US Class Current: 1/1
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Deep semantic analysis is performed on an electronic literary work in order to detect plot elements and optional other storyline elements such as characters within the work. Multiple levels of abstract are generated into a model representing the literary work, wherein each element in each abstraction level may be independently rated for preference by a user. Through comparison of multiple abstraction models and one or more user rating preferences, one or more alternative literary works may be automatically recommended to the user.

Claims

1. A method for automatic semantic rating and abstraction of literature comprising:

extracting by inferring by a computer system a plurality of plot details and character relationships using deep semantic analysis of a first electronic text, wherein the extracting comprises using deep semantic analysis of a first electronic text comprises program instructions for performing at least one semantic parsing process selected from the group consisting of;

parsing the electronic text to extract relationships between information entities in their associated contexts and how they relate to each other thereby producing triplets including triple store information items, pairs of information items and an ontological relationship between each pair, or both; and

parsing the electronic text to extract specific relationships between two information entities, thereby producing triplets of pairs of entities and their relationship;

introducing by a computer a first level in a first electronic hierarchical model representing the electronic text, wherein the first level comprises semantic relationships including the extracted plot details and character relationships from the first electronic text;

abstracting by a computer using deep semantic analysis a plurality of plot details and character relationships from the first level of the first electronic hierarchical model;

introducing by a computer a second level of semantic relationships to the first electronic hierarchical model, wherein the second level comprises the plot details and character relationships abstracted from the first level such that the second level in the hierarchical model is more abstract and less detailed than the first level;

comparing by a computer system two or more levels in the first electronic model to two or more levels of a second hierarchical electronic model representing a second electronic text, wherein the second electronic text has an associated user preference indicator;

responsive to the comparing finding a match above a pre-determined degree of similarity, declaring by a computer system to a user via a user interface device the first electronic text to have an expected user preference indicator as the user preference indicator associated with the second electronic text; and

responsive to the comparing finding a match above pre-determined degree of dissimilarity, declaring by a computer system to a user via a user interface device the first electronic text to have an expected user preference indicator complementary with the user preference indicator associated with the second electronic text;

wherein the declaration via a user interface device comprises at least a tangible recommendation output from the computer system.

2. The method as set forth in claim 1 wherein a highest level of semantic relationships of the first electronic model comprises a literary element category selected from the group consisting of genre, setting, central storyline event, and mood.

3. The method as set forth in claim 1 further comprising detecting deep semantic relationships of a consumption history of the user, and correlating the consumption history relationships to the identified events of the first electronic text identify one or more other electronic texts of likely interest to the user.

4. The method as set forth in claim 1 wherein the first electronic text comprises a work of literature selected from the group consisting of an electronic book, an electronic play, an electronic novel, a transcription of a video or movie, an electronic magazine, an electronic reference book, and electronic research paper.

5. The method as set forth in claim 1 wherein the extracting, identifying and abstracting are repeated for a plurality of different electronic texts, and wherein the method further comprises:

aggregating by a computer user ratings from the same user across three or more different electronic texts;

cross-referencing by a computer one or more elements across the three or more different electronic texts to infer commonalities and distinctions; and

inferring by a computer one or more literary elements the user is inclined to enjoy according to the aggregated user ratings according to the commonalities, or according to the distinctions, or according to both the commonalities and the distinctions; and

designating one or more of the three or more different electronic texts with a likely user rating.

6. The method as set forth in claim 5 wherein the extracting, identifying and abstracting are repeated for a plurality of different electronic texts forms an electronic model for each different electronic text, each model having a plurality of abstraction levels, and wherein the cross-referencing comprises overlaying cross-sections within topographical nodes of two or more of the electronic models.

7. The method as set forth in claim 5 wherein the designating comprises recommending a different electronic text that the user is likely to prefer.

8. The method as set forth in claim 5 wherein the designating comprises dissuading against a different electronic text that the user is unlikely to prefer.

9. The method as set forth in claim 8 wherein extracting of a plurality of plot details and character relationships further comprises using

sequence matching with parsing of the electronic text thereby identifying specific ontologies having similarities.

10. The method as set forth in claim 9 wherein the sequence matching with parsing of the electronic text to identify specific ontologies having similarities comprises a Smith-Waterman process.