Identifying related concepts of URLs and domain names

• US 8,417,695 B2
• Filed: 10/30/2009
• Issued: 04/09/2013
• Est. Priority Date: 10/30/2008
• Status: Active Grant

• Current Assignee: NetSeer, Inc. (Inuvo, Inc.)
• Original Assignee: NetSeer, Inc. (Inuvo, Inc.)
• Inventors: Zhong, Qian, Boscolo, Riccardo, Rezaei, Behnam A., Talaie, Sam, Roychowdhury, Vwani P.
• Primary Examiner(s): GIRMA, ANTENEH B
• Application Number:
  US12/610,202
• Publication Number:
  US 20100114879A1
• Time in Patent Office: 1,257 Days
• Field of Search: 707/723
• US Class Current: 707/723
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

A solution for identifying related concepts of URLs and domain names includes using structural parsing to extract information from user input comprising a URL or domain name. The information includes one or more of a protocol, a location, and a subdirectory. Semantic parsing of the information is used to identify a first one or more concepts represented by one or more tokens within the extracted information. A content association map is queried to retrieve a second one or more concepts related to the first one or more concepts. Each of the concepts represents a unit of thought, expressed by a term, letter, or symbol. The concept association map includes a representation of concepts, concept metadata, and relationships between the concepts. The first one or more concepts and the second one or more concepts are ranked, and the ranked concepts are stored for displaying to one or more users of the computer platform.

Claims

1. A computer implemented method comprising:

using structural parsing to extract information from user input comprising a URL or domain name, the information comprising one or more of a protocol, a location, and a subdirectory;

using semantic parsing of the information to identify a first one or more concepts represented by one or more tokens within the extracted information;

determining whether the domain name can be mapped to one or more concepts in the concept association map by switching term positions or changing numbers;

when the domain name can be mapped and if the mapped concepts have high score, identifying the concepts as seed concepts for further querying the concept association map;

when the mapped concepts do not have a high enough score, or the domain name cannot be mapped, then determining whether the input domain name can be mapped to a concept in the concept association map by typographical error correction, the correction comprising one or more of insertion, deletion, and switching or replacement of 1 or 2 characters; and

when the input domain name cannot be mapped by typographical error correction, or if concepts mapped as a result of typographical error correction do not have a high score, determining how to break the domain name into URL tokens by inserting separators at correction positions and correcting the tokens;

querying a concept association map to retrieve a second one or more concepts related to the first one or more concepts, each of the concepts representing a unit of thought, expressed by a term, letter, or symbol, the concept association map comprising a representation of concepts, concept metadata, and relationships between the concepts;

ranking the first one or more concepts and the second one or more concepts to create ranked concepts; and

storing the ranked concepts for displaying to one or more users of the computer platform.

2. The method of claim 1, further comprising:

responsive to the displaying, using feedback from the one or more users to update the concept association map.

3. The method of claim 1, further comprising utilizing a top level of the URL to associate a geographical location with the URL.

4. The method of claim 1, further comprising associating a penalty with concept term position switching, number mismatch, and character correction.

5. The method of claim 1, wherein breaking the URL into URL tokens further comprises:

associating a penalty with breaking a domain name into several tokens; and

associating a score used to rank, prune and expand URL splitting paths.

6. The method of claim 5, further comprising expanding a set of URL splitting paths according to at least one of:

online generated typographical error candidates; and

offline generated typographical error candidates.

7. The method of claim 1 wherein the ranking comprises assigning a score to each concept according to at least one of:

a number of occurrences of the concept in query logs;

a number of occurrences of the concept in web pages;

cost per click; and

click through rate.

8. The method of claim 1 wherein the querying comprises assigning a score to each expanded concept according to at least one of:

a concept frequency, an edge weight between a concept and a seed concept;

a cost per click; and

a click through rate.

9. The method of claim 2 wherein the updating comprises updating a concept edge weight according to at least one of:

a cost per click; and

a click through rate.

10. The method of claim 1, further comprising

querying an ads bank to obtain one or more ads based on the ranked concepts;

and

storing the one or more ads for displaying to the one or more users.

11. An apparatus comprising:

a memory; and

one or more processors configured to;

use structural parsing to extract information from user input comprising a URL or domain name, the information comprising one or more of a protocol, a location, and a subdirectory, wherein structural parsing comprises;

determining whether the domain name can be mapped to one or more concepts in the concept association map by switching term positions or changing numbers;

when the domain name can be mapped and if the mapped concepts have high score, identifying the concepts as seed concepts for further querying the concept association map;

when the mapped concepts do not have a high enough score, or the domain name cannot be mapped, then determining whether the input domain name can be mapped to a concept in the concept association map by typographical error correction, the correction comprising one or more of insertion, deletion, and switching or replacement of 1 or 2 characters; and

when the input domain name cannot be mapped by typographical error correction, or if concepts mapped as a result of typographical error correction do not have a high score, determining how to break the domain name into URL tokens by inserting separators at correction positions and correcting the tokens;

use semantic parsing of the information to identify a first one or more concepts represented by one or more tokens within the extracted information;

query a concept association map to retrieve a second one or more concepts related to the first one or more concepts, each of the concepts representing a unit of thought, expressed by a term, letter, or symbol, the concept association map comprising a representation of concepts, concept metadata, and relationships between the concepts;

rank the first one or more concepts and the second one or more concepts to create ranked concepts; and

store the ranked concepts for displaying to one or more users of the computer platform.

12. The apparatus of claim 11 wherein the one or more processors are further configured to:

responsive to the displaying, use feedback from the one or more users to update the concept association map.

13. The apparatus of claim 11 wherein the one or more processors are further configured to utilize a top level of the URL to associate a geographical location with the URL.

14. The apparatus of claim 11 wherein the one or more processors are further configured to associate a penalty with concept term position switching, number mismatch, and character correction.

15. The apparatus of claim 11 wherein the one or more processors are further configured to break the URL into URL tokens by:

associating a penalty with breaking a domain name into several tokens; and

associating a score used to rank, prune and expand URL splitting paths.

16. The apparatus of claim 15 wherein the one or more processors are further configured to expand a set of URL splitting paths according to at least one of:

online generated typographical error candidates; and

offline generated typographical error candidates.

17. The apparatus of claim 11 wherein the ranking comprises assigning a score to each concept according to at least one of:

a number of occurrences of the concept in query logs;

a number of occurrences of the concept in web pages;

cost per click; and

click through rate.

18. The apparatus of claim 11 wherein the querying comprises assigning a score to each expanded concept according to at least one of:

a concept frequency, an edge weight between a concept and a seed concept;

a cost per click; and

a click through rate.

19. The apparatus of claim 12 wherein the updating comprises updating a concept edge weight according to at least one of:

a cost per click; and

a click through rate.

20. The apparatus of claim 11 wherein the one or more processors are further configured to:

query an ads bank to obtain one or more ads based on the ranked concepts; and

store the one or more ads for displaying to the one or more users.

21. A program storage device readable by a machine, embodying a program of instructions executable by the machine to perform a method, the method comprising:

using structural parsing to extract information from user input comprising a URL or domain name, the information comprising one or more of a protocol, a location, and a subdirectory, wherein structural parsing comprises;

determining whether the domain name can be mapped to one or more concepts in the concept association map by switching term positions or changing numbers;

when the domain name can be mapped and if the mapped concepts have high score, identifying the concepts as seed concepts for further querying the concept association map;

when the mapped concepts do not have a high enough score, or the domain name cannot be mapped, then determining whether the input domain name can be mapped to a concept in the concept association map by typographical error correction, the correction comprising one or more of insertion, deletion, and switching or replacement of 1 or 2 characters; and

when the input domain name cannot be mapped by typographical error correction, or if concepts mapped as a result of typographical error correction do not have a high score, determining how to break the domain name into URL tokens by inserting separators at correction positions and correcting the tokens;

using semantic parsing of the information to identify a first one or more concepts represented by one or more tokens within the extracted information;

querying a concept association map to retrieve a second one or more concepts related to the first one or more concepts, each of the concepts representing a unit of thought, expressed by a term, letter, or symbol, the concept association map comprising a representation of concepts, concept metadata, and relationships between the concepts;

ranking the first one or more concepts and the second one or more concepts to create ranked concepts; and

storing the ranked concepts for displaying to one or more users of the computer platform.