Repetitive fusion search method for search system

• US 20100036802A1
• Filed: 08/21/2008
• Published: 02/11/2010
• Est. Priority Date: 08/05/2008
• Status: Active Grant

• Current Assignee: Tokyo Denki University
• Original Assignee: Tokyo Denki University
• Inventors: Tsuruta, Setsuo, Sakurai, Yoshitaka
• Granted Patent: US 8,972,370 B2
• Time in Patent Office: Days
• US Class Current: N/A
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06Q 30/0603 Catalogue ordering

Abstract

Provided in the present invention is a repetitive fusion search method for a search system, in which: a search interface 20 displays parameter numerical value setting fields 25 for setting a numerical value for each of multiple instinct based feature parameters in order to input instinct based rating information, and a text input field 24 and a keyword selection field 26 for inputting keywords; and a search engine 13 searches, in response to search inputs input by a searcher using the search interface, a search target database 11 based on the keywords and extracts relevant targets, determines priority for the relevant targets based on the instinct based rating information, and transmits the relevant targets to the search interface 20.

Claims

1. A repetitive fusion search method for a search system that includes:

a search target database that stores information about targets to be searched for;

a searcher information database that stores information about a searcher;

a search interface that has the searcher input a keyword and instinct based rating information to be used for a search and displays input information and a search result;

a search engine that searches the search target database and the searcher information database based on the information input via the search interface, extracts a relevant target and displays the relevant target in the search interface; and

a database management unit that manages entering and updating data in the search target database and searcher information database, the method comprising;

the search interface displaying a parameter numerical value setting field for setting a numerical value for each of multiple instinct based feature parameters for inputting instinct based rating information and a keyword input field and a keyword selection field for inputting a keyword; and

the search engine searching, using search inputs input by the searcher using the search interface, the search target database based on the input keyword, extracting relevant targets, determining priority for the relevant targets based on the input instinct based rating information, and transmitting the relevant targets as a search result to the search interface.

2. The repetitive fusion search method for the search system according to claim 1, wherein the search engine performs the steps of:

(1) performing a search by setting a highest usage frequency value for a search parameter with a highest search parameter significance level in search parameters that are not selected in the search inputs set in the search interface, extracting a relevant target, and transmitting the relevant target as a first alternative for the search result to the search interface;

(2) performing a search by setting a second highest usage frequency value for a search parameter with a highest search parameter significance level in search parameters that are not selected in the search inputs set in the search interface, extracting a relevant target, and transmitting the relevant target as a second alternative for the search result;

(3) defining a search input for an alternative, in the first alternative and the second alternative for the search result, selected by the searcher using the search interface as a new search input; and

(4) repeating, using the new search input, searching for the first alternative to the search result in (1) and searching for the second alternative to the search result in (2) and transmitting a relevant target to the search interface, and thereby narrowing down the search targets.

3. The repetitive fusion search method for the search system according to claim 1, wherein:

the search target database stores histories of past search input and target selection for each searcher; and

the search engine estimates, with respect to the search input input by the searcher using the search interface, searcher search tendency information for an instinct based-feature-parameter setting criterion deviation and a search parameter significance level based on the histories of the past search input and target selection stored in the search-target database for the relevant searcher, searches the search target database by adjusting the search input using the estimation result or adjusting the matching degree calculation between the search input and the search target, extracts a relevant search target and transmits the relevant search target to the search interface.

4. The repetitive fusion search method for the search system according to claim 3, wherein every time a certain search target is selected, the search engine estimates, based on the deviation of each search parameter set for the selected search target from each search parameter in the search input, an instinct based-feature-parameter setting criterion deviation for each search parameter for the selected search target.

5. The repetitive fusion search method for the search system according to claim 3, wherein every time a certain search target is selected, the search engine judges a match rate between each search parameter in the selected search target and each parameter in the search input and estimates a search parameter significance level for each search parameter by increasing the significance level of the search parameter that matches the counterpart parameter.

6. The repetitive fusion search method for the search system according to claim 3, wherein the search engine obtains a search input distribution for each searcher based on the search histories of each searcher stored in the searcher information database, sets a lower significance level for the search parameter with a larger search input distribution and sets a higher significance level for the search parameter with a small search input distribution, and thereby estimates the search parameter significance level of each search parameter for each searcher.

7. The repetitive fusion search method for the search system according to claim 3, wherein the search engine obtains an input percentage for each search input for each searcher based on the search histories of each searcher stored in the searcher information database and estimates that the search input with the highest input percentage is a preferred search input for the relevant searcher.

8. The repetitive fusion search method for the search system according to claim 3, wherein the search engine obtains a statistic for search parameters set for search targets selected in the past based on the search target and selection histories for each searcher stored in the searcher information database and estimates that the search parameter with the average value is estimated as a preference search input.

9. The repetitive fusion search method for the search system according to claim 3, wherein the search engine stores information within one session from when a searcher starts a search until the end of the search in a short-term memory and stores information for a session in a long-term memory when the session ends.

10. The repetitive fusion search method for the search system according to claim 3, wherein the search engine stores information within one session from when a searcher starts a search until the end of the search in a short-term memory, stores information for a session in a long-term memory when the session ends and estimates latest searcher tendency information in the long-term memory from past chronological session-unit searcher tendency information pieces.

11. The repetitive fusion search method for the search system according to claim 3, wherein:

the search interface displays parameter numerical value setting fields for setting a numerical value for each of multiple instinct based features;

the search engine searches, using the set numerical value for each instinct based feature parameter set in the parameter numerical value setting fields, the search target database, extracts a search target to which a criterion value close to the numerical value set for each of the input instinct based feature parameters and transmits the extracted search target to the search interface;

the search engine obtains, when the search target is selected in the search interface, a setting criterion deviation of a setting criterion value for each instinct based feature parameter for the search target from the set value for each of the input instinct based feature parameters, and stores the deviation for each of the instinct based feature parameters in the searcher information database; and

the search engine corrects, with respect to a newly-input numerical value setting for each instinct based feature parameter, by retrieving the setting criterion deviation for each instinct based feature parameter for the relevant searcher in the searcher information database and adding the called setting criterion deviation to the newly-input numerical value setting for each instinct based feature parameter, searches the search target database with the corrected numerical value setting for each instinct based parameter as a search condition, and extracts a relevant target and transmits the relevant target to the search interface.

12. The repetitive fusion search method for the search system according to claim 3, wherein:

the search interface displays a setting field for setting each of multiple search parameters;

the search engine calculates a matching degree for each search target by weighting, by the search parameter significance level, the deviation of the search input from each of the search parameters set for the search target, obtained for each of the search parameters, then summing the weighted deviations and displays in the search interface the search targets in descending order of the matching degree.

13. The repetitive fusion search method for the search system according to claim 3, wherein, if no search target matches a search input, which is an input search value, the search engine repeats searches by relaxing search conditions, starting from an item with the lowest search parameter significance level and extracts a relevant search target.

14. The repetitive fusion search method for the search system according to claim 3, wherein the search engine obtains past search tendency for the relevant searcher based on the searcher'"'"'s search histories in the searcher information database, extracts a search target that is highly likely selected based on the obtained past search tendency, transmits the extracted search target to the search interface, receives a searcher'"'"'s response transmitted from the search interface, and updates the search tendency of the relevant searcher in the searcher information database based on the content of the response.

15. The repetitive fusion search method for the search system according to claim 3, wherein the search engine obtains an input percentage for each search input based on the search histories of each searcher, estimates the search input with the highest input percentage as a preferred search input for the relevant searcher, searches the search target database using the preferred search input, transmits the search result as a recommended target to the search interface and displays the recommended target in the search interface.

16. The repetitive fusion search method for the search system according to claim 3, wherein the search engine learns how a searcher configures search settings during one session from when the searcher starts a search until the end of the search, enters the learned content in the searcher information database, calculates, in response to a search request from the searcher, search condition setting with the highest usage frequency in the relevant search time with reference to the searcher information database, searches the search target database based on the search condition setting, transmits the search result as a recommended target to the search interface and displays the recommended target in the search interface.

17. The repetitive fusion search method for the search system according to claim 3, wherein the search engine holds in advance multiple situations during a search process as definition information, learns the searcher'"'"'s tendency for setting search conditions in each of the multiple situations, enters the learned content in the searcher information database, judges which search situation in the definition information is relevant to the search request from the relevant searcher, extracts a preference set by the searcher with reference to the searcher information database based on the relevant search situation, searches the search target database based on the preference, transmits the search result as a recommended target to the search interface and displays the recommended target in the search interface.

18. The repetitive fusion search method for the search system according to claim 3, wherein:

the search engine divides the feature of a search target into multiple feature items, displays in the search interface a parameter numerical value setting field for each feature parameter for setting a numerical value for each feature item; and

reads a numerical value, for each feature parameter, set in the parameter numerical value setting field for each feature item in the search interface and transmitted by the searcher;

the search engine estimates the deviation of the numerical value for each feature parameter set by the searcher from a setting criterion value for each feature parameter in the system based on the search histories of the relevant searcher in the searcher information database;

the search engine obtains a corrected setting numerical value for each feature parameter by correcting the deviation of the numerical value for each feature parameter set by the searcher from the setting criterion value for each feature parameter in the system; and

the search engine searches the search target database using the corrected search numerical value for each feature parameter, extracts a search target, transmits information for the search target as a recommended target to the search interface and displays the recommended target in the search interface.