System and method for text categorization based on ontologies

• US 8,782,051 B2
• Filed: 04/26/2013
• Issued: 07/15/2014
• Est. Priority Date: 02/07/2012
• Status: Expired due to Fees

• Current Assignee: South Eastern Publishers Inc.
• Original Assignee: South Eastern Publishers Inc.
• Inventors: Chashchin, Kirill, Anshukov, Sergey, Bardin, Valery, Kordonsky, Simon
• Primary Examiner(s): Btit, Jacob F
• Assistant Examiner(s): TORGRIMSON, TYLER J
• Application Number:
  US13/872,022
• Publication Number:
  US 20130212111A1
• Time in Patent Office: 445 Days
• US Class Current: 707/740
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates

Abstract

A system for text categorization based on ontologies comprising data collector software modules; a categorizer software module; and a database comprising an indexed database of documents and their categorizations, and further comprising a plurality of ontologies, each ontology comprising a plurality of hierarchical taxonomies and each hierarchical taxonomy comprising a plurality of taxons. The data collector software modules receive a document to be classified and submit them to the categorizer software module; and the categorizer performs the following steps to categorize each document: splitting the document into sentences; selecting words or phrases that are present in ontologies stored in the database server; selecting a plurality of subtrees from the ontologies based on the presence of specific subcategories in the document; determining a weight for each subcategory; pruning subcategories having a weight below a threshold; and for each of the plurality of modified subtrees, computing a conditionality coefficient.

Claims

1. A system for text categorization based on ontologies, the system comprising:

a plurality of data collector software modules stored and operating on a plurality of network-attached computers;

a categorizer software module stored and operating on a network-attached server computer; and

a database server comprising an indexed database of documents and their categorizations, and further comprising a plurality of ontologies, each ontology comprising a plurality of hierarchical taxonomies and each hierarchical taxonomy comprising a plurality of taxons;

wherein the data collector software modules receive a text to be classified and submits the received text to the categorizer software module; and

further wherein the categorizer performs the following steps to categorize the received text;

splitting the text into sentences;

selecting words or phrases from the sentences of the received text that are present in one or more of the plurality of ontologies stored in the database server;

determining one or more specific subcategories that the sentence corresponds to in view of pattern analysis of the selected words or phrases;

selecting a plurality of subtrees from the plurality of ontologies based on the determination of the selected words or phrases belonging to one or more specific subcategories;

determining a weight for each subcategory within the one or more of the specific subcategories;

using the selected plurality of subtrees to create at least one modified subtree by eliminating from the selected plurality of subtrees subcategories having a category weight below a threshold;

for each of the at least one modified subtree, computing a conditionality coefficient to make a Boolean determination of whether to consider the respective modified subtree or not in categorization of the text; and

using any modified subtree that has been determined to be considered in categorization of the text to categorize the text;

wherein the conditionality coefficient is determined at least in part by user-defined rules or the presence, or absence, of diagnostics in nodes of at least one neighboring subtree.

2. A method for text categorization based on ontologies, the method comprising the steps of:

receiving, via a plurality of data collector software modules stored and operating on a plurality of network-attached computers, a text to be classified;

submitting the received text to a categorizer software module stored and operating on a network-attached server computer;

performing the following using the categorizer software module;

splitting the text into sentences;

selecting words or phrases from the sentences of the received text that are present in one or more the plurality of ontologies stored in the database server;

determining one or more specific subcategories that the sentence corresponds to in view of pattern analysis of the selected words or phrases;

selecting a plurality of subtrees from the plurality of ontologies based on the determination of the selected words or phrases belonging to one or more specific subcategories;

determining a weight for each subcategory within the one or more of the specific subcategories;

using the selected plurality of subtrees to create at least one modified subtree by eliminating from the selected plurality of subtrees subcategories having a category weight below a threshold;

for each of the at least one of modified subtree, computing a conditionality coefficient;

using the conditionality coefficient to make a Boolean determination of whether to consider the respective modified subtree or not in categorization of the text; and

using any modified subtree that has been determined to be considered to store a resulting document categorization in a database server comprising an indexed database of texts and each of their respective categorizations, and further comprising a plurality of ontologies, each ontology comprising a plurality of hierarchical taxonomies and each hierarchical taxonomy comprising a plurality of taxons;

wherein the conditionality coefficient is determined at least in part by user-defined rules or the presence, or absence, of diagnostics in nodes of at least one neighboring subtree.