Entity pages

• US 8,484,015 B1
• Filed: 05/14/2010
• Issued: 07/09/2013
• Est. Priority Date: 05/14/2010
• Status: Active Grant

• Current Assignee: Wolfram Alpha LLC (Wolfram Research Incorporated)
• Original Assignee: Wolfram Alpha LLC (Wolfram Research Incorporated)
• Inventors: Wolfram, Stephen, Williamson, Christopher, Tafjord, Oyvind
• Primary Examiner(s): PULLIAS, JESSE SCOTT
• Application Number:
  US12/780,651
• Time in Patent Office: 1,152 Days
• Field of Search: 704/1, 704/9, 704/10, 707/706, 707/708, 707/769, 707/771
• US Class Current: 704/9
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3349 Reuse of stored results of previous queries

Abstract

An answer engine determines an entity related to a query from a user in a system designed for computation, search, or other activities involving the processing of queries. The system determines whether an entity page for the entity is stored in an answers database and, if so, retrieves and displays the stored entity page. If there is no stored entity page for the entity, the answer engine generates and displays an answer to the query and may optionally store the answer as an entity page. Additionally, the answer engine may pre-generate and store entity pages according to algorithmic variation of received queries and/or according to information received from external sources.

Claims

1. A method for use in a natural language (NL) system that generates answers to NL queries, the method comprising:

generating, using one or more computing devices, one or more predicted queries representing one or more potential queries that may be received subsequently by the NL system;

providing, using one or more computing devices, the one or more predicted queries to an NL query answering system implemented by one or more computing devices;

receiving, from the NL query answering system, one or more respective answers to the one or more predicted queries;

generating, using one or more computing devices, one or more web pages corresponding to the one or more respective answers, wherein each web page is configured to display the respective answer;

storing the one or more web pages in a first database;

receiving a NL query;

determining, using one or more computing devices, whether a pre-generated web page that includes an answer to the NL query is stored in the first database;

when the pre-generated web page that includes the answer to the NL query is stored in the first database, providing the pre-generated web page via a network; and

when no pre-generated web page that includes the answer to the NL query is stored in the first database;

generating, using one or more computing devices, an answer based on data stored in a second database,

generating, using one or more computing devices, a new web page that includes the answer to the NL query; and

providing the new web page via the network in response to the NL query.

2. The method of claim 1, further comprising:

determining, using one or more computing devices, whether to store the new web page to the first database; and

storing in the first database the new web page when it is determined that the new web page should be stored.

3. The method of claim 2, further comprising making the stored new web page available to one or more search engines separate from the NL system.

4. The method of claim 2, wherein determining whether to store the new web page is based on a determined likelihood that the NL query will be received subsequently.

5. The method of claim 4, wherein determining whether to store the new web page is based on at least one of a number of times that the NL query is received or a frequency that the NL query is received.

6. The method of claim 4, wherein determining whether to store the new web page is based on a determined level of interest in an entity corresponding to the NL query measured from data external to the NL system.

7. The method of claim 6, wherein determining whether to store the web page data is based on data from at least one of news feeds, RSS (Really Simple Syndication) feeds, community news websites, search engine trend reports, trend reports from a web-based reference source, news aggregator services, social networking sites, and blogs.

8. The method of claim 1, wherein determining whether the pre-generated web page that includes the answer is stored in the first database comprises:

interpreting, using one or more computing devices, the received NL query;

determining, using one or more computing devices, two or more potentially intended queries corresponding to the received NL query;

making, using one or more computing devices, an assumption to select an intended query from the potentially intended queries; and

determining, using one or more computing devices, whether the pre-generated web page that includes the answer is stored in the first database based on the selected intended query.

9. A method for use in a natural language (NL) system that generates answers to NL queries, the method comprising:

generating, using one or more computing devices, a set of one or more potential alternative NL queries that are related to a first NL query;

providing, using one or more computing devices, the set of one or more potential alternative NL queries to an NL query answering system implemented by one or more computing devices;

receiving, from the NL query answering system, one or more respective answers to the one or more potential alternative NL queries in the set, wherein the one or more respective answers are based on data stored in a first database;

generating, using one or more computing devices, one or more web pages corresponding to the one or more respective answers, wherein each web page is configured to display the respective answer; and

storing in a second database the one or more web pages.

10. The method of claim 9, wherein generating, using one or more computing devices, a set of one or more potential alternative NL queries that are related to a first NL query comprises:

receiving the first NL query;

interpreting, using one or more computing devices, the first NL query to generate a disambiguated second query; and

algorithmically varying, using one or more computing devices, the disambiguated second query to generate one or more third queries.

11. The method of claim 10, wherein interpreting the first NL query to generate a disambiguated second query comprises:

disambiguating, using one or more computing devices, the first NL query to determine a first entity associated with the first NL query; and

determining, using one or more computing devices, a category of the first entity; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting, using one or more computing devices, for the first entity one or more second entities of the same category as the first entity.

12. The method of claim 10, wherein interpreting the first NL query to generate a disambiguated second query comprises disambiguating, using one or more computing devices, the first NL query using one or more first parameters not included in the first NL query, the one or more first parameters including one or more of:

browser type, language, location, internet service provider, time of the query, and date of the query; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting, using one or more computing devices, for the one or more first parameters one or more second parameters.

13. The method of claim 9, wherein interpreting the first NL query to generate a disambiguated second query comprises:

disambiguating, using one or more computing devices, the first NL query to determine a first entity associated with the first NL query; and

determining, using one or more computing devices, a queried attribute of the first entity; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting, using one or more computing devices, for the queried attribute one or more second attributes of the first entity.

14. The method of claim 9, wherein generating, using one or more computing devices, a set of one or more potential NL queries that are related to a first NL query comprises determining, using one or more computing devices, for at least one of the potential NL queries a likelihood of receiving the at least one potential NL query.

15. The method of claim 14, wherein the determined likelihood is based on a plurality of previously received NL queries.

16. The method of claim 15, wherein the plurality of previously received NL queries is received within a predetermined span of time.

17. The method of claim 9, wherein generating, using one or more computing devices, a set of one or more potential alternative NL queries comprises receiving one or more of the following information:

data from a news source;

data from a blogging service;

data from a social networking service;

data from a community-based news article website;

data from a news aggregator;

trend data from a web-based reference source, the trend data describing recent search statistics related to the web-based reference source; and

trend data from a search engine, the trend data describing recent search statistics related to the search engine.

18. The method of claim 17, wherein the received data are received by one of a search bot, a web feed, and a news feed.

19. The method of claim 9, wherein generating, using one or more computing devices, a set of one or more potential alternative NL queries comprises algorithmically varying, using one or more computing devices, a received query to generate a set of one or more potential NL queries that are related to the received query.

20. A natural language (NL) system comprising:

a first database to store data relevant to entities;

a second database to store web page data, the web page data including web pages configured to display answers to NL queries;

a caching manager, communicatively connected to each of the first database and second database, the caching manager configured to;

receive data representing a received NL query,

determine whether the second database contains a first web page configured to display an answer to the received NL query, and

fetch the first web page from the second database when the first web page is stored in the second database, wherein the first web page is made available to the network; and

a NL query answering engine, communicatively connected to the caching manager, the NL query answering engine configured to determine an answer to the received NL query when it is determined that the first web page is not stored on the second database, wherein the answer to the received NL query is determined using data from the first database;

wherein the caching manager is further configured to, after the NL query answering engine determines the answer to the received NL query;

determine, using one or more computing devices, whether to store a new web page to the second database, the new web page including the answer to the received NL query determined by the NL query answering engine, and store in the second database the new web page when it is determined that the new web page should be stored.

21. The NL system of claim 20, wherein the first database comprises:

a plurality of entities categorized by class and sub-class,

a plurality of attributes associated with the entities.

22. The NL system of claim 20, wherein the first database comprises:

a plurality of entities categorized into one of a plurality of classes and sub-classes; and

a plurality of attributes, each attribute associated with at least one of the plurality of classes and sub-classes.

23. The NL system of claim 20, wherein the second database comprises web page data representing a plurality of pre-generated web pages configured to display answers to NL queries, each of the plurality of pre-generated web pages corresponding to one or more of a plurality of queries.

24. The NL system of claim 23, wherein each of the plurality of pre-generated web pages comprises one or more of:

static information, programming code indicating a source from which to retrieve dynamic information.

25. The NL system of claim 23, wherein one or more of the plurality of pre-generated web pages has an associated date on which the generated web page will be deleted or updated.

26. The NL system of claim 20, wherein the NL query answering engine comprises:

a parser;

one or more scanners; and

one or more post-processors.

27. A non-transitory tangible storage medium having stored thereon machine executable instructions, the machine executable instructions, when executed by one or more machines, causing the one or more machines to:

generate a set of one or more potential alternative natural language (NL) queries that are related to a first NL query;

provide the set of one or more potential alternative NL queries to an NL query answering system implemented by one or more computing devices;

receive, from the NL query answering system, one or more respective answers to the one or more potential alternative NL queries in the set, wherein the one or more respective answers are based on data stored in a first database;

generate one or more web pages corresponding to the one or more respective answers, wherein each web page is configured to display the respective answer; and

store in a second database the one or more web pages.

28. The non-transitory tangible storage medium of claim 27, wherein the instructions cause the one or more machines to:

receive the first NL query;

interpret the first NL query to generate a disambiguated second query; and

algorithmically vary the disambiguated second query to generate one or more third queries.

29. The non-transitory tangible storage medium of claim 28, wherein the instructions cause the one or more machines to:

disambiguate the first NL query to determine a first entity associated with the first NL query; and

determine a category of the first entity; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting for the first entity one or more second entities of the same category as the first entity.

30. The non-transitory tangible storage medium of claim 28, wherein interpreting the first NL query to generate a disambiguated second query comprises disambiguating the first NL query using one or more first parameters not included in the first NL query, the one or more first parameters including one or more of:

browser type, language, location, internet service provider, time of the query, and date of the query; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting for the one or more first parameters one or more second parameters.

31. The non-transitory tangible storage medium of claim 27, wherein the instructions cause the one or more machines to:

disambiguate the first NL query to determine a first entity associated with the first NL query; and

determine a queried attribute of the first entity; and

wherein algorithmically varying the disambiguated second query to generate one or more third queries comprises substituting for the queried attribute one or more second attributes of the first entity.

32. The non-transitory tangible storage medium of claim 27, wherein determining a set of one or more potential NL queries that are related to a first NL query comprises determining for at least one of the potential NL queries a likelihood of receiving the at least one potential NL query.

33. The non-transitory tangible storage medium of claim 32, wherein the determined likelihood is based on a plurality of previously received NL queries.

34. The non-transitory tangible storage medium of claim 33, wherein the plurality of previously received NL queries is received within a predetermined span of time.

35. The non-transitory tangible storage medium of claim 27, wherein determining a set of one or more potential NL queries comprises receiving one or more of the following information:

data from a news source;

data from a blogging service;

data from a social networking service;

data from a community-based news article website;

data from a news aggregator;

trend data from a web-based reference source, the trend data describing recent search statistics related to the web-based reference source; and

trend data from a search engine, the trend data describing recent search statistics related to the search engine.

36. The non-transitory tangible storage medium of claim 35, wherein the received data are received by one of a search bot, a web feed, and a news feed.

37. The non-transitory tangible storage medium of claim 27, wherein determining a set of one or more potential NL queries comprises algorithmically varying a received query to generate a set of one or more potential NL queries that are related to the received query.