Ontology for use with a system, method, and computer readable medium for retrieving information and response to a query

• US 9,747,390 B2
• Filed: 11/15/2013
• Issued: 08/29/2017
• Est. Priority Date: 04/07/2004
• Status: Active Grant

• Current Assignee: Oracle OTC Subsidiary LLC (Oracle Corporation)
• Original Assignee: Oracle OTC Subsidiary LLC (Oracle Corporation)
• Inventors: Cooper, Edwin Riley, Bierner, Gann, Yuret, Deniz, Williams, James Charles, Beghelli, Filippo, Graham, Laurel Kathleen
• Primary Examiner(s): Adesanya, Olujimi
• Application Number:
  US14/081,746
• Publication Number:
  US 20140074826A1
• Time in Patent Office: 1,383 Days
• Field of Search: 704 9, 704 10, 704257
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: G06F 16/367 Ontology
• CPC Class Codes: G06F 16/90335 Query processing
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A system, method, and computer readable medium for applying one or more information retrieval technologies is disclosed for resolving a query. In one embodiment, an exemplary system generates a response using a language analysis module configured to parse a query into elements. This system can also include a rules engine configured to compare a condition of a rule against the elements, where the rule is configured to perform an action to retrieve information. Further, a response generator coupled to said rules engine receives the rule and retrieves the information for presentation to a querier in a portion of a display that adjusts proportionately to the degree of importance of said information.

Claims

1. A method comprising:

configuring, by a computer system, a data structure according to an ontology, wherein the data structure configured according to the ontology comprise a plurality of concepts that are organized in hierarchical layers and that are linked by a set of relationships;

generating, by the computer system, an index representing stored content based on the data structure configured according to the ontology;

receiving, by the computer system, a query comprising a natural language expression or a natural language question;

parsing, by the computer system, the query into elements;

annotating, by the computer system, the elements so that each element has an associated annotation, wherein the associated annotation defines the respective element as one or both of conceptual and canonical based on the ontology;

comparing, by the computer system, a first condition of one or more rules from a rules dictionary against the elements and the associated annotations;

comparing, by the computer system, a second condition of the one or more rules from the rules dictionary against peripheral information attributed to a querier, wherein the peripheral information is distinct from the query and is retrieved from a peripheral information repository;

generating, by the computer system, an action list corresponding to one or more actions of the one or more rules based on whether the first condition and the second condition are satisfied, the action list comprising first search operator criteria that specifies a first scoring scope of search and second search operator criteria that specifies a second scoring scope of search, wherein the first scoring scope of search is different from the second scoring scope of search in order of priority and scoring;

selecting, by the computer system and based on one or both of the first search operator criteria and the second search operator criteria in the action list, one or more retrieval technologies from a plurality of retrieval technologies, the plurality of retrieval technologies comprising a keyword searching technology, a document-level relevance-based searching technology, and an ontology-based searching technology;

performing, by the computer system, the one or more actions, the one or more actions comprising using the selected one or more retrieval technologies to retrieve stored content represented in the index based on the elements and the associated annotations; and

generating, by the computer system, a response to the query using the stored content retrieved based on the index.

2. The method of claim 1, wherein the ontology comprises a hierarchical representation of the concepts and the set of relationships wherein each concept is defined as an expression and is identified as being in either a conceptual form or a canonical form.

3. The method of claim 2, wherein the generating the index comprises:

parsing the stored content into one or more tokens;

annotating each token with one or more associated annotations; and

storing the one or more tokens and associated annotations in the index.

4. The method of claim 3, wherein the annotating each token with the one or more associated annotations comprises defining each token as being conceptual, canonical, or both conceptual and canonical based on the ontology.

5. The method of claim 4, wherein:

the parsing comprises parsing the received query into one or more tokens;

the annotating comprises annotating each token of the parsed query with one or more associated annotations defining the token as being conceptual, canonical, or both conceptual and canonical based on the ontology; and

the comparing the first condition is based on the annotated tokens.

6. The method of claim 5, wherein the second condition defines a context of an application in which the one or more actions are performed, the context of the application comprising:

session information corresponding to a session in which the query was initiated; and

a web page and content with which the query coincides.

7. The method of claim 5, wherein the generating the response comprises:

identifying a set of results from the stored content satisfying the query based on performing the one or more actions;

ordering the identified set of results; and

presenting the ordered, identified set of results.

8. A system comprising:

a processor; and

a memory coupled with and readable by the processor and storing instructions which, when executed by the processor, causes the processor to;

configure a data structure according to an ontology, wherein the data structure configured according to the ontology comprise a plurality of concepts that are organized in hierarchical layers and that are linked by a set of relationships;

generate an index representing stored content based on the data structure configured according to the ontology;

receive a query comprising a natural language expression or a natural language question;

parse the query into elements;

annotate the elements so that each element has an associated annotation, wherein the associated annotation defines the respective element as one or both of conceptual and canonical based on the ontology;

compare a first condition of one or more rules from a rules dictionary against the elements and the associated annotations;

compare a second condition of the one or more rules from the rules dictionary against peripheral information attributed to a querier, wherein the peripheral information is distinct from the query and is retrieved from a peripheral information repository;

generate an action list corresponding to one or more actions of the one or more rules based on whether the first condition and the second condition are satisfied, the action list comprising first search operator criteria that specifies a first scoring scope of search and second search operator criteria that specifies a second scoring scope of search, wherein the first scoring scope of search is different from the second scoring scope of search in order of priority and scoring;

select, based on one or both of the first search operator criteria and the second search operator criteria in the action list, one or more retrieval technologies from a plurality of retrieval technologies, the plurality of retrieval technologies comprising a keyword searching technology, a document-level relevance-based searching technology, and an ontology-based searching technology;

perform the one or more actions, the one or more actions comprising using the selected one or more retrieval technologies to retrieve stored content represented in the index based on the elements and the associated annotations; and

generate a response to the query using the stored content retrieved based on the index.

9. The system of claim 8, wherein the ontology comprises a hierarchical representation of the concepts and the set of relationships wherein each concept is defined as an expression and is identified as being in either a conceptual form for or a canonical form.

10. The system of claim 9, wherein the generating the index comprises:

parsing the stored content into one or more tokens;

annotating each token with one or more associated annotations; and

storing the one or more tokens and associated annotations in the index.

11. The system of claim 10, wherein the annotating each token with the one or more associated annotations comprises defining each token as being conceptual, canonical, or both conceptual and canonical based on the ontology.

12. The system of claim 11, wherein

the parsing comprises parsing the received query into one or more tokens;

the annotating comprises annotating each token of the parsed query with one or more associated annotations defining the token as being conceptual, canonical, or both conceptual and canonical based on the ontology; and

the comparing the first condition is based on the annotated tokens.

13. The system of claim 12, wherein the second condition-defines a context of an application in which the one or more actions are performed, the context comprising:

session information corresponding to a session in which the query was initiated; and

a web page and content with which the query coincides.

14. The system of claim 12, wherein the generating the response comprises:

identifying a set of results from the stored content satisfying the query based on performing the one or more actions;

ordering the identified set of results; and

presenting the ordered, identified set of results.

15. The system of claim 8, further comprising:

a plurality of processors that comprises the processor;

a plurality of memories that comprises the memory;

the plurality of processors and the plurality of memories facilitate an information retrieval system that comprises;

a language analysis module communicatively coupled to content storage, a multi-layered concept repository, a rules engine, and a semantic indexing engine;

a response engine communicatively coupled to the rules engine, the semantic indexing engine, and the multi-layered concept repository;

the plurality of memories store a set of instructions that comprises the instructions further causing the plurality of processors to;

interface with a remote computing device via a plurality of networks, the remote computing device programmed to automatically form and submit queries to the plurality of processors via the plurality of networks;

provide an administrator interface that allows configuration of operations of the system based on generating instructions to preprocess content from which information will be indexed and used to process questions and to retrieve answers;

cause the system to operate in two modes, the two modes comprising an indexing mode and a query mode, where;

in the indexing mode, the system extracts text from a plurality of textual content sources and indexes the text for retrieval; and

in the query mode, the system analyzes one or more questions, received from the remote computing device, against a set of one or more rules, and then forms a response using the text indexed for retrieval;

encode the data structure according to a hierarchical representation of the plurality of concepts and the set of relationships so that the data structure corresponds to the multi-layered concept repository, the encoding comprising defining each concept of the plurality of concepts as an expression and identifying each concept of the plurality of concepts as being in either a conceptual form or a canonical form, wherein;

the ontology comprises the hierarchical representation; and

the configuring comprises the encoding and results in an ontology-based concept repository configured to link the plurality of concepts at multiple different concept domain layers, so that;

the concepts of the plurality of concepts are represented by one or more natural language words associated with a first concept domain layer;

at least some of the plurality of concepts are for higher concept domain layers and are represented by one or more natural language words having more specific terms than the one or more natural language words representing the concepts associated with the lower domain layers, the one or more natural language words having more specific terms comprising;

 

one or more industry-specific terms associated with a second concept domain layer higher than the first concept domain layer; and

 

one or more organization-specific terms associated with a third concept domain layer higher than the second concept domain layer;

at least some of the plurality of concepts are further associated with a part of speech and are defined as a rigid phrase that is matched by any punctuation and inflectional variant of the rigid phrase, or are further defined as a compositional phrase that includes a set of multiple terms that are matched by other concepts that contain all of the multiple terms contained in the set of multiple terms;

a first concept of the plurality of concepts includes;

 

a first part of speech identifier that identifies a first type of word class;

 

a first domain identifier identifying one of multiple different concept levels assigned to the first concept; and

 

a first headword that identifies a first collection of synonyms and other relationships that define the first concept;

a second concept of the plurality of concepts includes;

 

a second part of speech identifier that identifies a second type of word class;

 

a second domain identifier identifying one of multiple different concept levels assigned to the second concept; and

 

a second headword that identifies a second collection of synonyms and other relationships that define the second concept;

import structured content and unstructured content from the plurality of textual content sources into the content storage, the importing comprising crawling a plurality of remote content sources for web page content and acquiring the web page at least in part by recursively following discovered links within the plurality of remote content sources, wherein the plurality of textual content sources comprises the plurality of remote content sources;

storing managed answers;

operate in the indexing mode and perform the generating the index at least in part by;

transforming, with the language analysis module using an elementizer of the language analysis module to implement a change-based tokenization process, the text from the structured content and the unstructured content stored in the content storage into tokens annotated with additional information to enhance retrieval, the transforming comprising;

parsing the structured content and the unstructured content into a first set of elements used to form a first set of tokens based on the change-based tokenization process that recognizes degrees of character-to-character change in the text;

processing, with an annotator of the language analysis module, each token of the first set of tokens;

annotating, with the annotator, each token of the first set of tokens with one or more associated annotations, wherein the annotating each token with the one or more associated annotations comprises defining each token as being conceptual, canonical, or both conceptual and canonical based on the ontology so that;

when the one or more associated annotations define the respective token as conceptual, the one or more associated annotations comprise a respective concept specification corresponding to the respective token; and

when the one or more associated annotations define the respective token as canonical, the one or more associated annotations comprise a respective canonical form corresponding to the respective token; and

configuring, with the semantic indexing engine, the index as a semantic index to retrieve from the structured content and the unstructured content and the managed answers stored based on storing the first set of tokens and the one or more associated token annotations in the index;

operate in the query mode and use the language analysis module to parse the query into the elements and to annotate the elements so that each element has the associated annotation so that;

the parsing the query into the elements comprises forming a second set of tokens based on the elements;

the annotating each token of the second set of tokens with the one or more associated annotations defining each token as one or both of conceptual and canonical based on the ontology;

the annotating the elements is based on matching ones of the elements against concepts stored in the multi-layered concept repository to produce conceptual annotations;

perform the comparing the first condition at least in part by;

matching, via a regular expression language specifying the first condition of the one or more rules, the second set of tokens and the one or more associated annotations with a particular concept in the multi-layered concept repository at least in part by determining if at least part of the second set of tokens and the one or more associated annotations shares a common ancestor in the multi-layered concept repository with a question example from the rules dictionary;

using the particular concept matched as a key to search the semantic index;

determining range operators that specify a query scope or amount of surrounding text in the query that is compared with specified expressions of the one or more rules; and

comparing the query with the first part of speech identifier, the first domain identifier, the first headword, the second part of speech identifier, the second domain identifier, and the second headword; and

defining, with the second condition, a context of an application in which the one or more actions are performed, the context of the application comprising;

session information corresponding to a session in which the query was initiated; and

a web page and content with which the query coincides; and

specifying the peripheral information to include data from a particular user profile that is preset prior to generation of the query, the particular user profile specifying prior queries submitted by a respective user of the particular user profile;

specifying the peripheral information to further include a time that the query was initiated and session information or web page information;

selectively firing at least one action of at least one of the one or more rules from the rules dictionary based on results of the comparisons, wherein the selected one or more retrieval technologies comprises at least a first retrieval technology and a second retrieval technology, the selectively firing the at least one action comprises selectively firing a plurality of actions based on;

determining a respective relevancy of each firing action of the plurality of firing actions based on;

the elements and the annotations that contributed to respective results of the comparisons; and

determining relevance to the elements and the annotations of individual documents according to weightings specified via the at least one action;

selectively performing each firing action of the plurality of firing actions based upon the respective relevancy;

in response to firing a first action of the one or more actions, operating the first retrieval technology;

in response to a second action of the one or more actions, operating the second retrieval technology that is configured to supply a particular managed answer to the query from the managed answers stored, wherein the managed answer is specified via the one of the rules having the second action;

the first retrieval technology and the second retrieval technology being selected from a group comprising a keyword searching technology, a document-level relevance-based searching technology, and an ontology-based searching technology;

the rules engine is configured to perform the comparing the second condition prior to performing the comparing the first condition;

the generating the response is performed at least partially by a response generator communicatively coupled to the rules engine, and the generating comprises;

identifying a set of results from the stored content satisfying the query based on the semantic index based on performing the one or more actions using the selected one or more retrieval technologies;

determining respective importance of units of information corresponding to the set of results based on a quantifiable measure distinguishable from other actions of other rules, where the quantifiable measure is determined by one or more of an accumulator relevancy determinator, a recency module, a scope-based scorer, and a relative weight determinator,

the accumulator relevancy determinator effectuating a degree of document-level relevance based on occurrences of a combination of tokens, stems, and concepts,

the recency module providing a relevancy contribution for a particular document with respect to recency of particular information contained in the particular document,

the scope-based scorer evaluating one or more of the one or more actions as potential answers to the query at least in part by assigning a respective score to respective actions that is representative of an ability of a scope of a corresponding potential answer to answer the query,

the relative weight determinator effectuating a degree of various searching techniques when the query is of general nature and generating results based on weightings of results the accumulator relevancy determinator, the recency module, and the scope-based scorer;

ordering the set of results identified to create an ordered set of results based on the determining each respective importance of the units of the information;

generating a plurality of portions of a display based on the ordered set of results, each portion of the plurality of portions presenting one or more respective units of information based on the respective importance of the one or more respective units of the information,

the respective importance of at least one of the corresponding units being based on the relevancy of an identified concept to a document, and

the generating the plurality of portions comprising scoring various scopes of the units of the information;

generating, with a response formatter, each portion of the plurality of portions of the display at least in part by adjusting the portion based on a respective scope of the corresponding unit of information, wherein the respective scope is determined by one of a word, a phrase, a sentence, and a document in conjunction with identifiers of at least some of the concepts including the first part of speech identifier, the first domain identifier, the first headword, the second part of speech identifier, the second domain identifier, and the second headword;

presenting at least a first part of a query answer in a portion of a display that adjusts proportionately to a degree of importance of information corresponding to the at least part of the query answer;

selectively displaying at least a second part of the query answer based on a bias requirement, wherein a third action of the one or more actions, when enabled, provides the bias requirement;

displaying the particular managed answer to the query based on the second retrieval technology;

employing anaphora resolution and co-reference resolution for noun phrases at in part by generating a salience list from a previous question, the generating comprising;

retrieving synsets from the previous question, where each synset is a respective set of terms that denotes a similar concept;

discarding entities not compliant with syntactic constraints;

discarding entities not relevant to a content domain or topic by requiring an ontology score to meet a threshold;

partially ordering the salience list using search component scores;

further ordering the salience list using ontology entity scores; and

further ordering the salience list using at least some of the synsets in the salience list located furthest left first while moving right and using at least some of the synsets in the salience list having longer ranges first followed by smaller ranges;

generating a follow-up question to the previous question at least in part by;

recognizing certain expressions identifying a follow-up query intent;

annotate a first question portion that will be used to replace a first portion of the previous question;

annotate a second question portion that will be used to supplement the previous question;

determine a portion of the previous question to be subtracted;

identifying, using the ontology, a common ancestor and a cousin concept of at least one concept in the previous question;

replacing or supplementing the at least one concept in the previous question with the cousin concept; and

subtracting the portion of the previous question to be subtracted.

16. A computer-readable memory, which is non-transitory and comprises a set of instructions stored therein which, when executed by a processor, causes the processor to:

configure a data structure according to an ontology, wherein the data structure configured according to the ontology comprise a plurality of concepts that are organized in hierarchical layers and that are linked by a set of relationships;

generate an index representing stored content based on the data structure configured according to the ontology;

receive a query comprising a natural language expression or a natural language question;

parse the query into elements;

annotate the elements so that each element has an associated annotation, wherein the associated annotation defines the respective element as one or both of conceptual and canonical based on the ontology;

compare a first condition of one or more rules from a rules dictionary against the elements and the associated annotations;

compare a second condition of the one or more rules from the rules dictionary against peripheral information attributed to a querier, wherein the peripheral information is distinct from the query and is retrieved from a peripheral information repository;

generate an action list corresponding to one or more actions of the one or more rules based on whether the first condition and the second condition are satisfied, the action list comprising first search operator criteria that specifies a first scoring scope of search and second search operator criteria that specifies a second scoring scope of search, wherein the first scoring scope of search is different from the second scoring scope of search in order of priority and scoring;

select, based on one or both of the first search operator criteria and the second search operator criteria in the action list, one or more retrieval technologies from a plurality of retrieval technologies, the plurality of retrieval technologies comprising a keyword searching technology, a document-level relevance-based searching technology, and an ontology-based searching technology;

perform the one or more actions, the one or more actions comprising using the selected one or more retrieval technologies to retrieve stored content represented in the index based on the elements and the associated annotations; and

generate a response to the query using the stored content retrieved based on the index.

17. The computer-readable memory of claim 16, wherein the ontology comprises a hierarchical representation of the concepts and the set of relationships wherein each concept is defined as an expression and is identified as being in either a conceptual form for or a canonical form.

18. The computer-readable memory of claim 17, wherein the generating the index comprises:

parsing the stored content into one or more tokens;

annotating each token with one or more associated annotations; and

storing the one or more tokens and associated annotations in the index.

19. The computer-readable memory of claim 18, wherein the annotating each token with the one or more associated annotations comprises defining each token as being conceptual, canonical, or both conceptual and canonical based on the ontology.

20. The computer-readable memory of claim 19, wherein:

the parsing comprises parsing the received query into one or more tokens;

the annotating comprises annotating each token of the parsed query with one or more associated annotations defining the token as being conceptual, canonical, or both conceptual and canonical based on the ontology; and

the comparing the first condition is based on the annotated tokens.