Optimized ontology based internet search systems and methods

• US 9,262,527 B2
• Filed: 06/22/2012
• Issued: 02/16/2016
• Est. Priority Date: 06/22/2011
• Status: Expired due to Fees

• Current Assignee: New Jersey Institute of Technology
• Original Assignee: New Jersey Institute of Technology
• Inventors: Geller, James, Ochs, Christopher
• Primary Examiner(s): Ly, Cheyne D
• Application Number:
  US13/530,456
• Publication Number:
  US 20130013580A1
• Time in Patent Office: 1,334 Days
• Field of Search: 707/731, 707/739
• US Class Current: 1/1
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/90324 using system suggestions
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Methods and apparatus for providing suggested completions to search results are disclosed which employ locating nodes in a stored ontology, wherein nodes consist of classes or instances of information, that correspond to the at least one search term, retrieving neighbors of located nodes in the stored ontology, and generating suggested completions in a format which separates homonymic terms. Methods and apparatus are disclosed for building ontologies and dynamically expanding ontologies.

Claims

1. A computer-implemented method of providing suggested completions to search engine results, comprising executing on a processor the steps of:

receiving at a server at least one search term;

locating nodes in an ontology stored in a memory associated with the server, wherein nodes consist of classes or instances of information, that correspond to the at least one search term;

retrieving at least one neighbor and/or extended neighbor of each of two or more located nodes in the stored ontology, wherein neighbors of a class comprise parent classes, child classes and classes that are reachable from a class by traversing a semantic relationship, and wherein neighbors of an instance are the class to which the instance belongs, object properties and data type properties of the instance;

wherein extended neighbors of a class comprise ancestor classes, descendant classes and classes that are reachable from a class by traversing a path of at least one semantic relationship followed by another semantic relationship and wherein extended neighbors of an instance are classes which are reachable as parent or ancestors from the class to which the instance belongs, class properties and data type properties of the class to which the instance belongs;

retrieving first a parent or ancestor for each node which is a class;

for each node which is an instance, retrieving first the class of which it is an instance or at least one of its ancestors;

optionally, limiting the retrieved information to a subset; and

generating suggested completions in a format using at least one visually distinguishing feature that may be presented to a user, wherein

the method comprises using an ontology built according to the steps of querying a search engine with common terms;

extracting at least one term of interest from results generated by the search engine;

assigning the at least one term of interest to a top value category;

querying a separate database with the at least one term;

saving type and relationship data for term(s) found in the separate database;

removing all terms not correlated to the selected term type;

creating mappings for disambiguation tags;

assigning terms to an ontology type,

analyzing the types of relationships for each type, and

retaining the most common relationships for each type.

2. The method according to claim 1 wherein the step of optionally limiting the retrieved information to a subset comprises eliminating neighbors that are common to more than one sense of a search term.

3. The method according to claim 1 comprising providing a suggested completion using information of the node when only a single node is located.

4. The method according to claim 1 wherein the format of suggested completions comprises separating different homonymic terms by at least one visually distinguishing feature selected from horizontal or vertical lines, zig-zag lines, dashes, dots, and/or background colors.

5. The method according to claim 1 comprising showing a maximum of six to eighteen suggestions for a maximum of two to six homonymic terms.

6. The method according to claim 1 comprising employing selection criteria to limit the number of homonymic terms to four.

7. The method according to claim 6 wherein a selection criteria is a hit count estimate from a search engine.

8. The method according to claim 1 comprising using neighbors of a node that are used as positive search terms for one homonymic term as a negative search terms for another homonymic term.

9. An apparatus, including a processor operating to perform actions in response to executing computer program instructions, the actions comprising:

providing suggested completions to search engine results comprising receiving at a server at least one search term;

locating nodes in a stored ontology, wherein nodes consist of classes or instances of information, that correspond to the at least one search term;

retrieving at least one neighbor and/or extended neighbor of each of two or more located nodes in the stored ontology, wherein neighbors of a class comprise parent classes, child classes and classes that are reachable from a class by traversing a semantic relationship, and wherein neighbors of an instance are the class to which the instance belongs, object properties and data type properties of the instance;

wherein extended neighbors of a class comprise ancestor classes, descendant classes and classes that are reachable from a class by traversing a path of at least one semantic relationship followed by another semantic relationship and wherein extended neighbors of an instance are classes which are reachable as parent or ancestors from the class to which the instance belongs, class properties and data type properties of the class to which the instance belongs;

retrieving first a parent or ancestor for each node which is a class;

for each node which is an instance, retrieving first the class of which it is an instance or at least one of its ancestors;

optionally, limiting the retrieved information to a subset; and

generating suggested completions in a format using at least one visually distinguishing feature that may be presented to a user;

wherein the apparatus uses an ontology built according to the steps of querying a search engine with common terms;

extracting at least one term of interest from results generated by the search engine;

assigning the at least one term of interest to a top value category;

querying a separate database with the at least one term;

saving type and relationship data for term(s) found in the separate database;

removing all terms not correlated to the selected term type;

creating mappings for disambiguation tags;

assigning terms to an ontology type,

analyzing the types of relationships for each type, and

retaining the most common relationships for each type.

10. The apparatus according to claim 9 comprising using neighbors of a node that are used as positive search terms for one homonymic term as a negative search terms for another homonymic term.

11. A non-transitory, computer readable storage medium containing a computer program, which when executed by a computer processor causes the computer processor to perform actions, the actions comprising:

providing suggested completions to search engine results comprising receiving at a server at least one search term;

locating nodes in a stored ontology, wherein nodes consist of classes or instances of information, that correspond to the at least one search term;

retrieving at least one neighbor and/or extended neighbor of each of two or more located nodes in the stored ontology, wherein neighbors of a class comprise parent classes, child classes and classes that are reachable from a class by traversing a semantic relationship, and wherein neighbors of an instance are the class to which the instance belongs, object properties and data type properties of the instance;

wherein extended neighbors of a class comprise ancestor classes, descendant classes and classes that are reachable from a class by traversing a path of at least one semantic relationship followed by another semantic relationship and wherein extended neighbors of an instance are classes which are reachable as parent or ancestors from the class to which the instance belongs, class properties and data type properties of the class to which the instance belongs;

retrieving first a parent or ancestor for each node which is a class;

for each node which is an instance, retrieving first the class of which it is an instance or at least one of its ancestors;

optionally, limiting the retrieved information to a subset; and

generating suggested completions in a format using at least one visually distinguishing feature that may be presented to a user,

wherein the non-transitory, computer readable storage medium uses an ontology built according to the steps of querying a search engine with common terms;

extracting at least one term of interest from results generated by the search engine;

assigning the al least one term of interest to a top value category;

querying a separate database with the at least one term;

saving type and relationship data for term(s) found in the separate database;

removing all terms not correlated to the selected term type;

creating mappings for disambiguation tags;

assigning terms to an ontology type,

analyzing the types of relationships for each type, and

retaining the most common relationships for each type.

12. The apparatus according to claim 11 comprising using neighbors of a node that are used as positive search terms for one homonymic term as a negative search terms for another homonymic term.