Information processing method and system

• US 20040167875A1
• Filed: 02/20/2003
• Published: 08/26/2004
• Est. Priority Date: 02/20/2003
• Status: Abandoned Application

• Current Assignee: Askology Hb
• Original Assignee: Askology Hb
• Inventors: Sneiders, Eriks
• Application Number:
  US10/368,452
• Publication Number:
  US 20040167875A1
• Time in Patent Office: Days
• US Class Current: 707/3
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries

Abstract

The present invention relates to processing of textual information in order to automatically answer user-submitted questions pertaining to a given knowledge domain. A proposed system (500) includes a user input interface (510) adapted to receive a user-formulated question (Q_U) on a natural language format. A transforming module (520), on one hand, converts the user formulated question (Q_U) into a question template format (Q_UT) including at least one entity term indicative of a respective main concept embodied in the user formulated question (Q_U), and on the other hand, converts the user formulated question (Q_U) into a data instance matching format (Q_UD) adapted to query a structured database (570), which represents at least a part of a conceptual model of the knowledge domain. A lexicon database (540) contains a multitude of concepts (C) which each is related to a specific question template. A template-matching engine (550) matches the question-template formatted user question (Q_UT) against the lexicon database (540) to retrieve a matching template cluster (C_TQ^m) associated with a matching query template (T_Y^m) and a matching answer template (α_T^m), and including at least one matching question template. This template, in turn, has at least one entity slot which is linked to the structured database (570). A data instance-matching engine (560) matches the data-instance-matching formatted user question (Q_UD) against the structured database (570) to identify at least one matching data instance (D_i). A central processing unit (530) fills at least one particular entity slot of the at least one matching question template with the identified at least one matching data instance (D_i). A search engine (580) queries the structured database (570) with the matching query template (T_Y^m) to retrieve information ([i]) to complete the matching answer template (α_T^m). A presentation interface (590) presents an answer (α) based on the retrieved information ([i]) and the matching answer template (α_T^m).

Claims

1. A method of processing textual information to automatically answer user-submitted questions pertaining to a knowledge domain, comprising the steps of:

receiving a user formulated question on a natural language format,

representing the user formulated question on a question template format including at least one entity term indicative of a respective main concept embodied in the user-formulated question,

representing the user formulated question on a data instance-matching format adapted to query a structured database,

template matching the question-template formatted user formulated question against a lexicon database to retrieve a matching template cluster including at least one matching question template, the lexicon database containing a multitude of concepts which each is related to a specific question template,

testing if at least one of the at least one matching question template includes at least one entity slot which is linked to a structured database representing at least a part of a conceptual model of the knowledge domain, the matching template cluster being associated with a matching query template and an answer template, and if so

data instance matching the data instance-matching formatted user formulated question against the structured database to identify at least one matching data instance to fill at least one particular entity slot of the at least one matching question template,

querying the structured database with the matching query template to obtain information completing the matching answer template, and

presenting an answer to be perceived by a user, the answer being based on the retrieved information and the matching answer template.

2. A method according to claim 1, wherein if none of the at least one matching question template includes any entity slots, comprising the step of:

presenting an answer to be perceived by a user, the answer being based on static data in a matching answer template being associated with the at least one matching question template.

3. A method according to claim 1, wherein before the querying step, comprising the steps of:

presenting the at least one matching question template filled with at least one matching data instance in at least one particular entity slot to be perceived by a user, and

receiving a user selection command in respect of one particular matching question template to represent a preferred interpretation of the user formulated question.

4. A method according to claim 3, wherein producing the answer exclusively on basis of the preferred interpretation of the user formulated question.

5. A method according to claim 1, wherein the lexicon database is a multiple-lexicon database containing at least one autonomous unit, and each autonomous unit represents at least one concept related to one question template.

6. A method according to claim 1, wherein the data instance matching process involving searching in an index over data instances to produce a reduced set of candidate data instances.

7. A method according to claim 1, wherein the data instance matching process involving an entity-specific matching of matching data instances belonging to at least one concept/entity.

8. A method according to claim 1, wherein the data instance matching process involving a data-instance-specific matching of matching alternative representations of data instances.

9. A computer program product directly loadable into the internal memory of a computer, comprising software for controlling the steps of claim 1 when said program product is run on the computer.

10. A computer readable medium, having a program recorded thereon, where the program is to make a computer control the steps of claim 1.

11. A textual information processing system for automatically answering user-submitted questions pertaining to a knowledge domain, comprising:

a user input interface adapted to receive a user-formulated question on a natural language format,

a transforming module adapted to;

convert the user formulated question into a question template format including at least one entity term indicative of a respective main concept embodied in the user formulated question, and

convert the user-formulated question into a data instance-matching format adapted to match data instances,

a lexicon database containing a multitude of concepts which each is related to a specific question template,

a structured database representing at least a part of conceptual model of the knowledge domain,

a template matching engine adapted to match the question-template formatted user formulated question against the lexicon database to retrieve a matching template cluster including at least one matching question template, the at least one matching question template having at least one entity slot which is linked to the structured database, the matching template cluster being associated with a matching query template and an answer template.

a data instance-matching engine adapted to match the data instance-matching formatted user formulated question against the structured database to identify at least one matching data instance,

a central processing unit adapted to fill at least one particular entity slot of the at least one matching question template with the identified at least one matching data instance,

a search engine adapted to query the structured database with the matching query template to retrieve information to complete the matching answer template, and

a presentation interface adapted to present an answer to be perceived by a user, the answer being based on the retrieved information and the matching answer template.

12. A textual information processing system according to claim 11, wherein before triggering the search engine to query the structured database:

the presentation interface is adapted to present the at least one matching question template to be perceived by a user, and

the user input interface is adapted to receive a user selection command in respect of one particular matching question template and the at least one matching data instance to represent a preferred interpretation of the user formulated question.

13. A textual information processing system according to claim 12, comprising a central processing unit adapted to produce the answer exclusively on basis of the preferred interpretation of the user submitted question.

14. A textual information processing system according to claim 11, wherein the lexicon database is a multiple-lexicon database containing at least one autonomous unit, and each autonomous unit represents at least one concept related to one question template.

15. A textual information processing system according to claim 11, wherein the data instance matching engine is adapted to search an index over data instances to produce a reduced set of candidate data instances.

16. A textual information processing system according to claim 11, wherein the data instance matching engine is adapted to perform an entity-specific matching of matching data instances belonging to at least one concept/entity.

17. A textual information processing system according to claim 11, wherein the data instance-matching engine is adapted to perform a data-instance-specific matching of alternative representations of data instances.