JP2003108595A - Information retrieving device, information retrieving method and information retrieving program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a problem of requiring a long time for work when the number of retrieval results increases for requiring a user to successively check validity of the respective retrieval results since order of these retrieval results does not always coincide with validity in a retrieval purpose when obtaining a plurality of retrieval results. SOLUTION: A character string having the same attribute as a retrieving keyword and a character string having an attribute different from the retrieving keyword are extracted from a Web page gathered by a WWW information gathering part 24, and a score of the Web page is set according to a description degree of the character string to the Web page.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information retrieval device, an information retrieval method and an information retrieval program for retrieving a document (web page) matching a retrieval condition.

[0002]

2. Description of the Related Art Today, various kinds of information have been released on the Internet based on the rapidly popular WWW (World Wide Web) service. is happening. The address of information on the WWW is URL (Unif
Orm Resource Locator), but since the naming method of the URL has no regularity and is often changed, it is rare that the WWW user directly specifies the URL.

Therefore, HTML (Hyper Te
Target URL by following a hyperlink in another web page described in xt Markup Language) or using a search engine that returns the URL of the web page containing the specified keyword.
It has been reported that nearly 90% of the cases access the Internet (GVU Center 10th, Georgia Institute of Technology
Annual WWW User Survey 1998 http: //
www. cc. gatech. edu / gvu / use
r_surveys / survey-1998-10
/).

FIG. 25 is a block diagram showing a conventional information retrieving apparatus. In the figure, 1 is a WWW server, 2 is the Internet, 3 is a user terminal, and 4 is a list of URLs that match a keyword of a retrieval condition. It is an internet search engine provided to the worker terminal 3. Reference numeral 5 is a WWW information collection unit that collects a large number of web pages from the WWW server 1, 6
Is an index generation unit that generates an index for keyword search from a copy of the web pages collected by the WWW information collection unit 5, 7 is an index storage device that stores the index generated by the index generation unit 6, and 8 is a user terminal 3 is a search query processing unit that receives a search condition from No. 3 and refers to an index stored in the index storage device 7 to generate a list of URLs matching the search condition.

Next, the operation will be described. First, WWW
The information collecting unit 5 collects a large number of web pages from the WWW server 1. That is, when the URL of the web page is initialized, the web page is downloaded. Next, by extracting the HTML hyperlink described in the web page, the web page linked to the web page is downloaded. By repeatedly executing this process, web pages are downloaded one after another while tracing the web pages.

[0006] When the WWW information collecting unit 5 collects a large number of web pages, the index generating unit 6 generates an index for keyword search from a copy of those web pages. That is, the index generation unit 6 performs the HT on the web pages collected by the WWW information collection unit 5.
The process of separating the ML tag and normal text is executed. Then, a sentence which constitutes a normal text is divided into word units, and index information of the lexical element (word) is generated. Here, FIG. 26 is an explanatory diagram showing an example of index information, and in the example of FIG. 26, URLs of web pages in which each word appears are listed as index information.

Upon receiving the search condition from the user terminal 3, the search query processing unit 8 acquires the index entry stored in the index storage device 7 and generates a list of URLs that match the search condition. . If the search condition specifies that all keywords are included,
A list of URLs is generated by finding the common part of the URL set among a plurality of corresponding entries. Further, when the search condition specifies that one or more of the plurality of keywords are included, a URL list is generated by requesting the unification of the URL sets among the corresponding plurality of entries.

For example, when the index information of FIG. 26 is generated and a search is carried out under the condition that both "Mitsubishi" and "Tokyo" are included, the search result is {URL1,
URL3} and {URL1, URL2, URL4, URL
It becomes {URL1} which is a common part of 5}. Similarly, the search result for the condition that contains either "Mitsubishi" or "Tokyo" is {URL1, URL2, URL3, URL4.
URL5}.

On the other hand, when actually conducting a search for the purpose of identifying a company's home page, it is necessary to repeatedly perform the search using some conditions in order to prevent omission of the search and narrow down to the most appropriate result. is there. Specifically, as shown in FIG. 27, the first condition is the company's telephone number (phone numbers separated by hyphens), and the second condition is the company's telephone number (the AND condition of the part excluding the hyphen). ), The third condition is an AND condition of the whole company name and a part of the location, and the fourth condition is an AND condition of a part of the company name and a part of the location.

If the user conducts a search under the first condition and one or more results are obtained, the user selects the most appropriate result from the results or judges that there is no appropriate result, and executes the search. finish. When no result is obtained under the first condition, the search is performed under the second condition and the same judgment is performed. Hereinafter, the same applies to the third condition and the fourth condition, but if the result of the search under the fourth condition is 0, it is determined that there is no valid result, and the search is terminated.

[0011]

Since the conventional information retrieval apparatus is configured as described above, when a plurality of retrieval results are obtained, the order of the retrieval results and the validity of the retrieval purpose are not always required. It does not match. Therefore, it is necessary for the user to sequentially check the validity of each search result, and there is a problem that the work takes a long time when the number of search results increases. Further, in order to obtain as few search results as possible, it is necessary to repeat the search using a plurality of search conditions, and there is a problem that the work takes a long time.

The present invention has been made to solve the above problems, and obtains an information search device, an information search method, and an information search program which can omit checking of validity of search results by a user. With the goal.

[0013]

An information retrieval apparatus according to the present invention is a ranking determination means for determining the ranking of a web page in which a character string retrieved by the retrieval means is described by referring to the score of the web page. Is provided.

The information retrieving apparatus according to the present invention is provided with a ranking determining means for determining the ranking of the web pages collected by the collecting means with reference to the score of the web page.

In the information retrieval apparatus according to the present invention, the rank determining means outputs the web pages in the order from the highest rank.

In the information retrieval apparatus according to the present invention, the rank determining means outputs the partial contents of the web page having a high rank in order.

In the information retrieving apparatus according to the present invention, the order determining means determines the order of the WWW server to which the web page belongs based on the score of the web page.

In the information retrieving apparatus according to the present invention, the retrieving means normalizes the character string and the retrieval keyword extracted by the extracting means.

In the information retrieval apparatus according to the present invention, the retrieval means uses the telephone number as the retrieval keyword.

The information search device according to the present invention determines the ranking of the web pages collected by the collecting means with reference to the score of the web page, and changes the address of the web page to the address of the higher web page. A means for determining the order is provided.

The information retrieving apparatus according to the present invention is provided with retrieving means for retrieving a character string that does not match the retrieval keyword among the character strings extracted by the extracting means.

In the information search method according to the present invention, a character string that matches the search keyword is searched, and the ranking of the web page in which the character string is described is determined by referring to the score of the web page. It is a thing.

In the information retrieval method according to the present invention, the ranking of the collected web pages is determined by referring to the scores of the web pages.

The information retrieval method according to the present invention is such that the web page addresses are output in order from the highest.

In the information retrieval method according to the present invention, the partial contents of a web page having a high rank are sequentially output.

In the information search method according to the present invention, the WW to which the web page belongs based on the score of the web page.
The order of W servers is determined.

The information retrieval method according to the present invention is adapted to normalize the extracted character string and retrieval keyword.

The information retrieval method according to the present invention uses a telephone number as a retrieval keyword.

In the information retrieval method according to the present invention, the ranking of the collected web pages is determined by referring to the score of the web page, and the address of the web page is changed to the address of the higher web page. It is a thing.

The information search method according to the present invention searches for a character string that does not match the search keyword among the extracted character strings.

The information search program according to the present invention is provided with a rank determination processing procedure for determining the rank of the web page in which the character string searched by the search processing procedure is described by referring to the score of the web page. Is.

The information search program according to the present invention is provided with an order determination processing procedure for determining the order of the web pages collected by the collection processing procedure with reference to the score of the web page.

The information search program according to the present invention determines the ranking of web pages collected by the collection processing procedure with reference to the score of the web page, and sets the address of the web page to the address of the higher web page. The order determination processing procedure for changing is provided.

The information retrieval program according to the present invention has a retrieval processing procedure for retrieving a character string that does not match the retrieval keyword among the character strings extracted by the extraction processing procedure.

[0035]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below. Embodiment 1. 1 is a block diagram showing an information retrieval apparatus according to Embodiment 1 of the present invention, in which 21 is a W
A WWW server, 22 is the Internet, 23 is an Internet information search system, 24 is a WWW information collection unit (collection means) that accesses the WWW server 21 and collects web pages from the WWW server 21, and 25 is a WWW information collection unit 24. A character string having the same attribute as the search keyword (for example, a character string indicating a telephone number, name, location, etc.) and a character string having an attribute different from the search keyword are extracted from the collected web pages, and the character string for the web page is extracted. The information extracting unit (extracting means) that sets the score of the web page according to the degree of the description, 26 is the extraction information in which the character string extracted by the information extracting unit 25 and the score are associated with the URL of the web page. It is an extracted information storage device for storing a list.

Reference numeral 27 denotes a search target information storage device for storing search target information including a search keyword designating a search target,
28 is a character string extracted by the information extraction unit 25,
A join operation processing unit (search means) that searches for a character string that matches the search keyword stored in the search target information storage device 27, and a web page 29 in which the character string retrieved by the join operation processing unit 28 is described. Is determined by referring to the score of the web page, and a ranking processing unit (rank determination means) that sequentially outputs the URLs of the web pages having higher rankings, and 30 is the URL of the web page output from the ranking processing unit 29. Is a result information storage device for storing.

FIG. 2 is a block diagram showing the inside of the WWW information collecting unit 24. In the figure, 24a is an acquisition request URL queue holding URLs of web pages to be collected, and 24.
b is an acquisition request URL queue 24a from the WWW server 21.
Download section for downloading a web page related to the URL held in
In order to prevent the same web page from being downloaded by b, the already acquired URL storage device that stores the URL list of the web pages downloaded by the download unit 24b, and 24d is the URL content that stores the web page downloaded by the download unit 24b. The storage device 24e takes out an HTML hyperlink from the web page stored in the URL content storage device 24d, and acquires a URL indicating the link destination UR.
It is a link extraction unit to be inserted into the L queue 24a.

FIG. 3 is a block diagram showing the inside of the information extraction unit 25. In the figure, 25a is the HTM for the web page stored in the URL content storage device 24d.
H that executes the processing to separate the L tag and normal text
The TML analysis unit 25b divides a sentence constituting the text separated by the HTML analysis unit 25a into word units, and executes a morphological analysis to add part-of-speech information to each word. 25c is a syntax rule. 25d, the parser for searching a character string having the same attribute as the search keyword and a character string having an attribute different from the search keyword, 25d sets the score of the web page according to the degree of description of the character string on the web page. In addition, the extraction information management unit generates an extraction information list in which the character string and the score are associated with the URL of the web page.

Next, the operation will be described. Here, as shown in FIG. 7, the search target information storage device 27 stores the name,
It is assumed that company information including a location and a telephone number is stored as search target information. That is, in order to enable designation of a plurality of search conditions, at least one or more search conditions composed of one or more search target information are stored in a list format. Note that FIG. 7 describes only one search condition.

First, as shown in FIG. 4, when the URL of the known WWW server 21 is initialized in the acquisition request URL queue 24a (step ST11), the download section 24b of the WWW information collecting section 24 acquires the acquisition request URL. The URL of the known WWW server 21 is acquired from the queue 24a (step ST12). Then, referring to the URL list stored in the already acquired URL storage device 24c, the U
It is confirmed whether or not the RL web page has already been downloaded (step ST13). At this stage,
Since the web page has not been downloaded yet, it is determined that the web page associated with the URL has not been downloaded. In order to download many web pages, it is desirable to specify a WWW server including links to many WWW servers 21 as the initial URL.

When the download section 24b has not downloaded the web page associated with the URL, the download section 24b outputs the U
By acquiring the IP address of the WWW server 21 corresponding to the RL (step ST14) and issuing an HTTP GET request to the WWW server 21 (step ST15), the web page is downloaded from the WWW server 21. It is stored in the URL content storage device 24d (step ST16).

The link extraction unit 24e is used by the download unit 2
4b stores the web page in the URL content storage device 24d
When it is stored in, the URL of the web page is added to the URL list stored in the acquired URL storage device 24c (step ST17), and the link information is extracted from the content of the web page to acquire the URL of the link destination. Insert into the request URL queue 24a (step ST1
8).

The download section 24b uses the acquisition request URL.
It is determined whether or not a new URL is inserted in the queue 24a, and if the acquisition request URL queue 24a is empty, the process ends. On the other hand, if a new URL is inserted,
Returning to step ST12, steps ST12 to ST18
The above process is repeatedly executed (step ST19).

HTML analysis unit 25a of the information extraction unit 25
When the download unit 24b stores the web page in the URL content storage device 24d, the download unit 24b executes a process of separating the HTML tag and the normal text for each web page. When the HTML analysis unit 25a separates the texts, the lexical element analysis unit 25b divides the sentences forming the texts into word units, and also executes morphological analysis to add the part-of-speech information to each word. The part-of-speech information includes not only major classifications such as "nouns", "particles", and "verbs" but also details such as "noun-proper noun-place name". Used in.

The syntax analysis unit 25c is a lexical element analysis unit 25.
From the sequence of words and parts of speech decomposed by b, a partial sequence that matches the recognition pattern specified in the syntax rule (see FIG. 9) is extracted, and a character string is output together with the name of the matching recognition pattern. In FIG. 9, as the recognition pattern, <
Since "phone number", "name", and "location" are described, "<phone number> character string value" and "<name>character" are entered for the character strings considered to be the phone number, name, or location, respectively. Output in the format of "column value" and "<location> character string value", and output is not generated for other character strings. The recognition pattern corresponds to an "attribute" that represents the meaning of a character string, and the output of the syntax analysis unit 25c is generally a string of character string attributes and character string values.

Here, the syntax rule is, for example, RFC2.
234 (http://rfc.net/rfc223
4. ABNF (Augment)
edBackus-Naur Form, Extended Back
us-Naur notation).
The recognition according to the syntax rule can be realized using LR parsing which is a well-known compiler technique. Also, except for the pattern described in the syntax rule, an error will occur, but here the error is ignored and nothing is output.

The extraction information management unit 25d is a syntax analysis unit 25.
Based on the output of c, the score of the web page is set according to the type and appearance frequency of the character string attribute and the character string value contained in one web page. Specifically, the score of the web page is set by referring to the correspondence table (which shows the score according to the character string attribute and the character string value) as shown in FIG. The type of the character string value corresponding to the specified attribute <telephone number> is “1”, the type of the character string value corresponding to the attribute <name> is “4”, the attribute <
If the type of the string value corresponding to location> is “2”,
The score of the web page is set to "90".

In the example of FIG. 11, the score of the web page in which the character string value corresponding to each character string attribute does not exist is lowered, but the description is not complete, and the regular contact information is transmitted. This is to exclude web pages that are not intended. On the other hand, a web page with too many kinds of character string values corresponding to a certain character string attribute is also reduced in score. This is for excluding web pages published by a third party such as a list, and lowers the score of web pages in which four or more types of telephone numbers are described, for example.

When the score of the web page is set as described above, the extraction information management unit 25d sets the URL of the web page and the score, and the string of character string attributes and character string values (see FIG. 10A). (Corresponding to one line) and an extracted information index using a character string value of a certain character string attribute as a key and stored in the extracted information storage device 26 (see FIG. 10B).

Here, a telephone number is used as the character string attribute, and its value is a normalized telephone number represented only by numbers, excluding hyphens, parentheses, and blanks. The reason for using a normalized phone number is that it is expected that there will be less fluctuation (ambiguity) in the notation compared to names and locations. Generally, a plurality of extracted information indexes are generated from 0 corresponding to one web page, but for each key (normalized telephone number), there is an entry corresponding to the extracted information storage device 26. If it does not exist, the combination of the key, the URL and the score is inserted as a new entry, and if the corresponding entry in the extraction information storage device 26 already exists, the URL and the score pair is added to the entry. Then, it is written back to the extracted information storage device 26.

The combining operation processing section 28 starts its operation after the information extracting section 25 processes all the web pages. The join operation processing unit 28 searches the character strings extracted by the information extraction unit 25 for a character string that matches the search keyword stored in the search target information storage device 27. Specifically, as shown in FIG. 5, the join operation processing unit 28 first acquires one piece of search target information (search keyword) from the search target information storage device 27 (step ST21). Here, for convenience of explanation, it is assumed that the telephone number is acquired as the search keyword from the search target information storage device 27.
When obtaining the telephone number as the search keyword, the combination calculation processing unit 28 normalizes the telephone number (step ST22). Here, normalization is a process performed to absorb fluctuations in notation, such as hyphens, parentheses and spaces.
This is a process of removing the delimiters other than numbers and expressing the telephone number as a character string containing only numbers.

When the combination arithmetic processing unit 28 normalizes the telephone number which is the search keyword, it searches the character strings extracted by the information extracting unit 25 for a character string that matches the normalized telephone number. (Step ST23).

When there is a character string that matches the normalized telephone number, the combination calculation processing unit 28 determines the character string (= normalized telephone number) and the web in which the character string is described. The appearance URL list (see FIG. 10B), which is a correspondence table between the URL of the page and the score of the web page, is output to the ranking processing unit 29 (step ST).
25). It should be noted that the join operation processing unit 28 determines whether or not all the search target information has been processed. If all the search target information have not been processed, the process returns to step ST21 and steps ST21 to ST21. The processing of ST25 is repeatedly executed (step ST26).

When the combining operation processing unit 28 completes the search processing of the character string, the ranking processing unit 29 determines the rank of the Web page in which the searched character string is described by referring to the score of the Web page. , Output in order from the URL of the web page with the highest rank. In the example of FIG. 8, the top 2
It shows that the URLs of the two web pages are output.

Although the ranking processing unit 29 determines the ranking of the web page by referring to the score of the web page here, the ranking of the WWW server 21 to which the web page belongs based on the score of the web page. May be determined. Specifically, it is as shown in FIG.

The ranking processing section 29 first retrieves one search result of the join operation processing section 28 (step ST3).
1) A process of collecting a plurality of URLs in the appearing URL list for each server is executed (step ST32). That is, in HTTP, the part following http: /// is the server name, and therefore a process of collecting a plurality of URLs for each server is executed by referring to the server name.

Next, based on the score of each web page,
The maximum score of the web pages belonging to each WWW server 21 is set as the score of the WWW server 21 (step ST33). Then, the ranking of the plurality of WWW servers 21 is ranked according to the score of the WWW server 21 and the number of web pages belonging to the WWW server 21 (step ST34).

The ranking processing unit 29 determines the WWW servers 21 ranked at the top, for example, Ws within the top three.
Only the URL of the WW server 21 is sequentially output (step ST35). The ranking processing unit 29 determines whether or not all the search results have been processed, and if there is an unprocessed search result, the process returns to step ST31 and step S31.
The processes of T31 to ST35 are repeatedly executed (step ST36).

As is clear from the above, the first embodiment
According to this, a character string having the same attribute as the search keyword and a character string having an attribute different from the search keyword are extracted from the web page collected by the WWW information collecting unit 24, and the degree of description of the character string on the web page is calculated. According to the configuration, the score of the web page is set accordingly. Therefore, it is possible to eliminate the need for the user to check the validity of the search result.

In the first embodiment, the components of the Internet information search system 23 may be configured by using dedicated hardware such as IC, or a computer that executes software (information search program) may be used. It may be configured by using. FIG. 12 is a hardware configuration diagram when it is configured by using a computer. In the figure,
Reference numeral 41 denotes a CPU that executes an information retrieval program, and W
It has the functions of the WW information collection unit 24, the information extraction unit 25, the combination calculation processing unit 28, and the ranking processing unit 29. Four
2 is a memory for storing an information retrieval program and data necessary for executing the program, 43 is a console input / output device 4
4, a console input / output device, 45 a hard disk device 46
Is a disk interface, 46 is a hard disk device, and 47 is a network interface connected to the Internet 22.

Embodiment 2. In the above-described first embodiment, the output is performed in order from the URL of the web page having the higher rank, but the partial contents of the web page having the higher rank may be sequentially output. For example, by using the product name as the search target information, the price described with the product on the web page is output. As a result, the user can know the partial content of the web page, that is, the target information without actually accessing the web page.

Embodiment 3. FIG. 13 is a block diagram showing an information retrieval apparatus according to Embodiment 3 of the present invention. In the figure, the same reference numerals as those in FIG. Reference numeral 51 is a search engine that searches for a URL of a web page that includes a character string that matches the search keyword, and 52 is a search engine that gives the search keyword to the search engine 51 and acquires the URL of the web page that is the search result from the search engine 51. An inquiry unit 53 is a download unit for downloading a web page having the URL acquired by the search engine inquiry unit 52 from the WWW server 21. The search engine inquiry unit 5
2 and the download unit 53 constitute a collecting means.

Reference numeral 54 is an attribute extraction unit for extracting a character string having the same attribute as the search keyword and a character string having an attribute different from the search keyword from the web page downloaded by the download unit 53, and 55 is the web page downloaded by the download unit 53. The score setting unit sets the score of the web page according to the degree of description of the character string. Note that the attribute extraction unit 54 and the score setting unit 55 constitute extraction means. Reference numeral 56 denotes a ranking processing unit (rank determining unit) that determines the rank of the web page downloaded by the download unit 53 with reference to the score of the web page.

Next, the operation will be described. The search engine inquiring unit 52 obtains URLs of one or more web pages as a search result of the search engine 51 by passing a search keyword (for example, a company name) to the search engine 51 to request a search. The interface of the search engine 51 is based on standard specifications such as the HTTP protocol and the HTML format, and it is easy to specify the search keyword and extract the URL from the result.

When the search engine inquiry section 52 acquires the URLs of one or more web pages, the download section 53 receives web pages (HTML) corresponding to those URLs from the WWW server 21 via the Internet 22.
File) and outputs the web page to the attribute extraction unit 54.

When the attribute extraction unit 54 receives one or more web pages from the download unit 53, first, similarly to the HTML analysis unit 25a of the information extraction unit 25, the HTML tags and normal texts for each web page. Execute the process of separating. Next, like the lexical element analysis unit 25b of the information extraction unit 25, the attribute extraction unit 54 decomposes the sentences constituting the separated text into word units, executes morphological analysis, and adds the part-of-speech information to each word. Add. Note that the part-of-speech information includes not only major classifications such as “noun”, “particle”, and “verb” but also details such as “noun-proper noun-place name”, and the syntax referred to by the attribute extraction unit 54 below. Used in rules.

Next, the attribute extraction unit 54 is connected to the information extraction unit 25.
Similarly to the syntactic analysis unit 25c, the sub-string that matches the recognition pattern specified in the syntax rule is extracted from the decomposed word and part-of-speech strings, and the character string is output together with the name of the matching recognition pattern. In the syntax rule example in Figure 9,
The recognition patterns of <telephone number>, <name>, and <location> are described in accordance with the search purpose of the third embodiment. Therefore, for each character string considered to be one of these, “<telephone number>No.> character string value ”,“ <name> character string value ”,“ <location> character string value ”are output, and no output is generated for other character strings. The recognition pattern corresponds to the "attribute" that represents the meaning of the character string, and the output of the attribute extracting unit 54 is generally a string of character string attributes and character string values (see FIG. 14).

Here, the syntax rule is, for example, RFC2.
234 (http://rfc.net/rfc223
4. ABNF (Augment)
edBackus-Naur Form, Extended Back
us-Naur notation).
The recognition according to the syntax rule can be realized using LR parsing which is a well-known compiler technique. Also, except for the pattern described in the syntax rule, an error occurs, but here the error is ignored and nothing is output.

The score setting unit 55, like the extracted information management unit 25d of the information extraction unit 25, based on the output of the attribute extraction unit 54, the type of the character string attribute and the character string value included in one web page. And the score of the web page is set according to the appearance frequency. Specifically, the score of the web page is set by referring to the correspondence table (which shows the score according to the character string attribute and the character string value) as shown in FIG.
For example, the type of the character string value corresponding to the attribute <phone number> retrieved from a certain web page is “1”, and the attribute <name>
If the type of the character string value corresponding to "4" and the type of the character string value corresponding to the attribute <location> are "2", the score of the web page is set to "90".

In the example of FIG. 11, the score of the web page in which the character string value corresponding to each character string attribute does not exist is lowered, but this is not fully described, and the regular contact information is transmitted. This is to exclude web pages that are not intended. On the other hand, we also try to lower the score for web pages that have too many types of string values corresponding to a certain string attribute, but this is to exclude web pages published by third parties such as rosters. Yes, for example, the score of a web page in which four or more types of telephone numbers are written is lowered. When the score of the web page is set as described above, the score setting unit 55 sets the URL of the web page and the score as a pair, and the ranking processing unit 5
6 (see FIG. 15).

The ranking processing unit 56 uses the search engine 5
When a set of URLs and scores corresponding to all the search results of No. 1 is received from the score setting unit 55, those plural URs are received.
A process of collecting L for each WWW server 21 is executed. That is, in HTTP, the part following http: /// is W
Since it is the WW server name, a process of collecting a plurality of URLs for each WWW server 21 is executed by referring to the WWW server name. Next, based on the score of each URL, each WWW
The maximum value of the URL scores belonging to the server 21 is set as the score of the WWW server 21.

Then, the ranking of the plurality of WWW servers 21 is ranked according to the score of the WWW server 21 and the number of URLs belonging to the WWW server 21. The ranking processing unit 56 uses the WWW ranked at the top.
Server 21, for example WWW server 21 within the top three
For, the URLs having the maximum score in the WWW server 21 are sequentially output (see FIG. 16).

As is clear from the above, the third embodiment
According to the above, for a plurality of web pages as a result of the search by the search engine 51, a character string attribute and its value are extracted for character strings other than the search keyword, and the score of the web page is set according to the degree of the description. Since it is configured to determine the ranking of the web page based on
Inappropriate results that cannot be excluded only by specifying the search keyword can be excluded, and the user can omit the work of selecting an appropriate result (see FIG. 17).

In the third embodiment as well, similar to the first embodiment, the components of the Internet information search system 23 may be configured by using dedicated hardware such as IC, or software. A computer that executes the (information search program) may be used. In addition, in the third embodiment, the URL is returned, but the extracted information (for example, a telephone number to a company name and a product name to a price) may be returned.

Fourth Embodiment 18 is a block diagram showing an information retrieval apparatus according to Embodiment 4 of the present invention. In the figure, the same reference numerals as those in FIGS. 61 is a score setting unit 55
An extraction information management unit 62 that generates an extraction information index with a character string value of a certain character string attribute as a key from the URL and score of the web page received from It is a URL correction unit (order determining means) that changes the URL of a web page to the URL of a higher-level web page. FIG. 20 is a flowchart showing the processing contents of the URL correction unit 62.

Next, the operation will be described. When the score setting unit 55 sets the score of the web page as in the third embodiment, the extraction information management unit 61 receives the URL and score of the web page received from the score setting unit 55, the character string attribute and the character string. A column of value pairs (see FIG. 10 (a))
(Corresponding to one line) and the extracted information index using the character string value of a certain character string attribute as a key and stored in the extracted information storage device 26 (see FIG. 19).

Here, a telephone number is used as the character string attribute, and its value is a normalized telephone number represented only by numbers, excluding hyphens, parentheses, and blanks. The reason for using a normalized phone number is that it is expected that there will be less fluctuation (ambiguity) in the notation compared to names and locations. Generally, a plurality of extracted information indexes are generated from 0 corresponding to one web page, but for each key (normalized telephone number), there is an entry corresponding to the extracted information storage device 26. If it does not exist, the combination of the key, the URL and the score is inserted as a new entry, and if the corresponding entry in the extraction information storage device 26 already exists, the URL and the score pair is added to the entry. Then, it is written back to the extracted information storage device 26.

The ranking processing section 56 starts its operation when the WWW information collecting section 24 completes the information collection and the extracted information corresponding to all the collected web pages is stored in the extracted information storage device 26. Ranking processing unit 56
Retrieves the entries one by one from the extracted information storage device 26, executes the same processing as in the third embodiment, and executes the same process as that of the key (normalized telephone number) of each entry, and the web page with the highest score in the URLs. And those related to the WWW server 21 are output.

The URL correction unit 62 is a ranking processing unit 5.
When the key (normalized telephone number) and URL (plurality) are received from 6, the already-acquired URL storage device 2 is received as shown in FIG.
4c and the information stored in the URL content storage device 24d, the URL character string is corrected and the corrected UR
The L character string is stored in the result information storage device 30. That is, U
The HTML file corresponding to the RL is searched from the URL content storage device 24d, and if the content includes a hyperlink pointing to a higher URL, the original URL is replaced with the URL.

Here, with respect to the upper URL, if the last element (corresponding to the file name) is delimited by delimiting the URL character string with “/”, the number of elements is smaller than the original URL, and all the elements are original. Is included in the URL. For example,
The original URL is http: // www. a. co. jp /
products / pc. If it is html, ht
tp: // www. a. co. jp / index. ht
ml is the upper URL, but http: // www.
a. co. jp / products / or http: ///
www. a. co. jp / products / inde
x. html is not considered as a high-level URL.

Furthermore, when the URL character string represents a file name (when it does not end with "/"), the file name portion at the end of the URL character string is removed to generate the directory URL, and the already acquired URL is obtained. When the storage device 24c is searched and an access record to the directory URL exists, it is determined that the directory URL is valid and the original URL is replaced with the directory URL. For example, the original URL is http: // www. a. co. j
p / gaiyou. If the URL is http: // www. a. c
o. jp / is generated, the acquired URL storage device 24c is searched, and if the existence of the URL can be confirmed, the URL is ht
tp: // www. a. co. Rewrite to jp /.

If the access record to the directory URL does not exist in the acquired URL storage device 24c, it is determined that the directory URL is invalid. In this case, of the file URLs in the directory, the one having the highest validity as the URL of the home page is selected. For example, the original URL http: // w
ww. a. co. jp / gaiyou. The file URL in the same directory obtained by searching the acquired URL storage device 24c for http is http: // w
ww. a. co. jp / gaiyou. html (including the original URL itself for convenience), http: // www.
a. co. jp / products. html, htt
p: // www. a. co. jp / index. htm
If it is 1, the home page URL is generally index. Since the relevance of http is high, the URL is http: // www. a. co. jp / index.
Rewrite to html. The URL is not rewritten unless the validity is evaluated. Note that FIG. 21 is an explanatory diagram showing an example of setting the validity of the file name in the URL correction unit 62.

When retrieving the company home page from the result information storage device 30, the URL is obtained by retrieving the telephone number of the target company with a normalized character string (removing non-numeric characters such as hyphens, parentheses and spaces). To be

As is clear from the above, the fourth embodiment
According to the above, the validity of the web page as the search result is determined based on the information collected from the WWW server 21, and the URL of the more valid web page is used as the result, so that the search keyword is included. It is possible to return a reasonable result which is not always the case, and the user can omit the work of correcting the result (see FIG. 22).

Also in the fourth embodiment, as in the first embodiment, the components of the Internet information search system 23 may be configured by using dedicated hardware such as an IC, or software. A computer that executes the (information search program) may be used.

Embodiment 5. 23 is a block diagram showing an information retrieval apparatus according to Embodiment 5 of the present invention. In the figure, the same reference numerals as those in FIG. Reference numeral 71 is a known information storage device that stores a known normalized telephone number (search keyword), and 72 is a search keyword stored in the known information storage device 71 among the character strings stored in the extracted information storage device 26. An anti-join operation unit (search means) that searches for a character string that does not match with, and 73 is an unknown information storage device that stores the character string searched by the anti-join operation unit 72.

Next, the operation will be described. The anti-join calculation unit 72 starts its operation when the WWW information collection unit 24 completes the information collection and the extracted information corresponding to all the collected web pages is stored in the extracted information storage device 26. The anti-join operation unit 72 stores, in the unknown information storage device 73, an entry whose key (normalized telephone number) does not exist in the known information storage device 71 among the entries stored in the extracted information storage device 26.

Here, it is assumed that the known information storage device 71 stores a key (telephone number) in a pre-normalized format, similarly to the extraction information storage device 26 (see FIG. 24). Specifically, the anti-join operation unit 72 acquires one key of the entry from the extracted information storage device 26, searches the known information storage device 71, and outputs the entry only when the key does not exist. , Is executed for all the entries in the extracted information storage device 26.

As described above, according to the fifth embodiment, since the character string having the attribute (telephone number) of the search target is extracted from the information collected from the WWW server 21, the search target is specifically described. This has the effect of performing a search in which it is not possible to specify the value of.

Also in the fifth embodiment, as in the first embodiment, the components of the Internet information search system 23 may be configured by using dedicated hardware such as an IC, or software. A computer that executes the (information search program) may be used.

[0091]

As described above, according to the present invention, the ranking determining means for determining the ranking of the web page in which the character string retrieved by the retrieval means is described by referring to the score of the web page is provided. Since it is configured as described above, there is an effect that the check of the validity of the search result by the user can be omitted.

According to the present invention, the ranking determining means for determining the ranking of the web pages collected by the collecting means with reference to the score of the web page is provided, so that the ranking cannot be excluded only by designating the search keyword. It becomes possible to exclude inappropriate results, and as a result, there is an effect that the user can omit the work of selecting an appropriate result.

According to the present invention, since the rank determining means is configured to output the web pages in the order of their ranks in order, it is possible to access from the web pages of high validity.

According to the present invention, the rank determining means is configured to output the partial contents of the web page having a high rank in order, so that the user can know the target information without actually accessing the web page. There is an effect that can be.

According to this invention, the ranking determining means determines that the W to which the web page belongs based on the score of the web page.
Since it is configured to determine the rank of the WW server, there is an effect that the WWW server having a highly valid web page can be recognized.

According to the present invention, since the search means is configured to normalize the character string and the search keyword extracted by the extraction means, it is possible to absorb the fluctuation of the notation and improve the search accuracy. There is.

According to the present invention, since the search means is configured to use the telephone number as the search keyword, there is an effect that information about a company or a group can be accurately searched.

According to the present invention, the ranking of the web pages collected by the collecting means is determined with reference to the score of the web page, and the ranking of changing the address of the web page to the address of the upper web page. Since the means is provided, it is possible to return a valid result that does not necessarily include the search keyword, and as a result, the user can omit the work of correcting the result. .

According to the present invention, since the search means for searching the character string which does not match the search keyword among the character strings extracted by the extracting means is provided, the value to be searched is specifically specified. The effect is that it is possible to perform an impossible search.

According to the present invention, the character string that matches the search keyword is searched, and the ranking of the web page in which the character string is described is determined by referring to the score of the web page. This has the effect of omitting the validity check of the search results by the user.

According to the present invention, since the ranking of the collected web pages is determined by referring to the score of the web page, it is possible to exclude inappropriate results that cannot be excluded only by specifying the search keyword. As a result, the user can omit the work of selecting a proper result.

According to the present invention, since the web pages are output in order from the highest-ranked web page address,
There is an effect that it is possible to access from a highly valid web page.

According to the present invention, since the partial contents of the web page with the highest rank are output in order,
Without the user actually accessing the web page,
It has the effect of being able to know the desired information.

According to the present invention, since the ranking of the WWW server to which the web page belongs is determined based on the score of the web page, it is possible to recognize the WWW server having the highly valid web page. There is an effect that can be done.

According to the present invention, since the extracted character string and the search keyword are configured to be normalized,
This has the effect of absorbing fluctuations in the notation and improving the search accuracy.

According to the present invention, since the telephone number is used as the search keyword, there is an effect that information about a company or a group can be accurately searched.

According to the present invention, the ranking of the collected web pages is determined with reference to the score of the web page, and the address of the web page is changed to the address of the higher web page. As a result, a valid result not necessarily including the search keyword can be returned, and as a result, the user can omit the work of correcting the result.

According to the present invention, of the extracted character strings, a character string that does not match the search keyword is searched, so that a search in which it is impossible to specifically specify a value to be searched is executed. There is an effect that can be.

According to the present invention, the ranking determination processing procedure for determining the ranking of the web page in which the character string retrieved by the retrieval processing procedure is described by referring to the score of the web page is provided. There is an effect that the check of the validity of the search result by the user can be omitted.

According to the present invention, the ranking determination processing procedure for determining the ranking of the web pages collected by the collection processing procedure with reference to the score of the web page is provided. It is now possible to exclude inappropriate results that cannot be excluded,
As a result, there is an effect that the user can omit the work of selecting an appropriate result.

According to the present invention, the rank of the web pages collected by the collection processing procedure is determined with reference to the score of the web page, and the rank of the web page address is changed to the higher rank web page address. Since the configuration is such that the determination processing procedure is provided, as a result, it is possible to return a valid result that does not necessarily include the search keyword, and as a result, the user can omit the work of correcting the result. There is.

According to the present invention, since the retrieval processing procedure for retrieving the character string that does not match the retrieval keyword among the character strings extracted by the extraction processing procedure is provided, the value to be searched is specifically specified. The effect is that a search that cannot be specified is performed.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an information search device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram showing the inside of a WWW information collecting unit.

FIG. 3 is a configuration diagram showing the inside of an information extraction unit.

FIG. 4 is a flowchart showing the processing content of a WWW information collection unit.

FIG. 5 is a flowchart showing the processing contents of a join operation processing unit.

FIG. 6 is a flowchart showing the processing contents of a ranking processing unit.

FIG. 7 is an explanatory diagram showing search target information.

FIG. 8 is an explanatory diagram showing result information.

FIG. 9 is an explanatory diagram showing syntax rules.

FIG. 10 is an explanatory diagram showing an extraction information list and an appearance URL list.

FIG. 11 is an explanatory diagram for explaining a score setting method.

FIG. 12 is a hardware configuration diagram when configured using a computer.

FIG. 13 is a configuration diagram showing an information search device according to a third embodiment of the present invention.

FIG. 14 is an explanatory diagram showing output information of an attribute extraction unit.

FIG. 15 is an explanatory diagram showing output information of a score setting unit.

FIG. 16 is an explanatory diagram showing output information of a ranking processing unit.

FIG. 17 is an explanatory diagram showing effects of the information search device according to the third embodiment of the present invention.

FIG. 18 is a configuration diagram showing an information search device according to a fourth embodiment of the present invention.

FIG. 19 is an explanatory diagram showing an extraction information index.

FIG. 20 is a flowchart showing the processing contents of a URL correction unit.

FIG. 21 is an explanatory diagram showing a setting example of validity of a file name in the URL correction unit.

FIG. 22 is an explanatory diagram showing effects of the information search device according to the fourth embodiment of the present invention.

FIG. 23 is a configuration diagram showing an information search device according to a fifth embodiment of the present invention.

FIG. 24 is an explanatory diagram showing stored contents of an extracted information storage device and an unknown information storage device.

FIG. 25 is a block diagram showing a conventional information retrieval device.

FIG. 26 is an explanatory diagram showing an example of index information.

FIG. 27 is an explanatory diagram showing search target information.

[Explanation of symbols]

21 WWW server, 22 Internet, 23 Internet information search system, 24 WWW information collecting unit (collecting means), 24a Acquisition request URL queue, 24
b download unit, 24c already acquired URL storage device,
24d URL content storage device, 24e link extraction unit, 25 information extraction unit (extraction means), 25a HTM
L analysis unit, 25b lexical element analysis unit, 25c syntax analysis unit, 25d extraction information management unit, 26 extraction information storage device, 27 search target information storage device, 28 join operation processing unit (search means), 29 ranking processing unit (rank Determination means), 30 result information storage device, 41 CPU, 42
Memory, 43 console interface, 44 console input / output device, 45 disk interface, 4
6 hard disk device, 47 network interface, 51 search engine, 52 search engine inquiry section (collecting means), 53 download section (collecting means), 54 attribute extracting section (extracting means), 55 score setting section (extracting means), 56 ranking Processing unit (rank determination unit), 61 Extracted information management unit, 62 URL correction unit (rank determination unit), 71 Known information storage device, 72 Anti-join calculation unit (search unit), 73 Unknown information storage device.

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Takayuki Tamura             2-3 2-3 Marunouchi, Chiyoda-ku, Tokyo             Inside Ryo Electric Co., Ltd. F term (reference) 5B075 KK07 ND16 PR08 UU40

Claims (22)

[Claims] 1. A collecting means for collecting a web page from a WWW server, a character string having the same attribute as a search keyword is extracted from the web page collected by the collecting means, and a description of the character string is written for the web page. Extraction means for setting the score of the web page according to the degree, search means for searching for a character string that matches the search keyword among the character strings extracted by the extraction means, and the search means for searching An information retrieval apparatus, comprising: a ranking determining unit that determines the ranking of a web page in which a character string is described with reference to the score of the web page. 2. Collection means for collecting a web page containing a character string that matches a search keyword from a WWW server,
Extracting means for extracting a character string having the same attribute as the search keyword from the web page collected by the collecting means, and setting the score of the web page according to the degree of description of the character string on the web page; An information retrieval device comprising: a ranking determining means for determining the ranking of web pages collected by the collecting means with reference to the score of the web page. 3. The information retrieving apparatus according to claim 1, wherein the rank determining means outputs the web pages in the order of high rank in order. 4. The information retrieving apparatus according to claim 1, wherein the rank determining means outputs the partial contents of the web page having a higher rank in order. 5. The ranking determination means determines the ranking of the WWW server to which the web page belongs based on the score of the web page.
Information retrieval device described. 6. The information search device according to claim 1, wherein the search means normalizes the character string and the search keyword extracted by the extraction means. 7. The information search device according to claim 1, wherein the search means uses a telephone number as a search keyword. 8. A collecting means for collecting a web page from a WWW server, a character string having the same attribute as a search keyword is extracted from the web page collected by the collecting means, and a description of the character string is written for the web page. Extracting means for setting the score of the web page according to the degree, and determining the ranking of the web pages collected by the collecting means by referring to the score of the web page, and the address of the web page is set to the upper web. An information retrieval device comprising a ranking determining means for changing to a page address. 9. Collection means for collecting web pages from a WWW server, extraction means for extracting a character string having the same attribute as a search keyword from the web pages collected by the collection means, and characters extracted by the extraction means. An information retrieval apparatus comprising: a retrieval unit that retrieves a character string that does not match the search keyword from the columns. 10. When a web page is collected from a WWW server, a character string having the same attribute as the search keyword is extracted from the web page, and the score of the web page is scored according to the degree of description of the character string on the web page. On the other hand, while setting
An information search method for searching a character string that matches the search keyword and determining the rank of a web page in which the character string is described by referring to the score of the web page. 11. When a web page containing a character string that matches a search keyword is collected from a WWW server, a character string having the same attribute as the search keyword is extracted from the web page, and a description of the character string is written to the web page. An information retrieval method in which the score of the web page is set according to the degree, and the ranking of the collected web pages is determined by referring to the score of the web page. 12. The information search method according to claim 10, wherein the information is output in order from the address of the web page having the highest rank. 13. The information retrieving method according to claim 10, wherein the partial contents of a web page having a high rank are output in order. 14. The information retrieval method according to claim 10, wherein the ranking of the WWW server to which the web page belongs is determined based on the score of the web page. 15. The information search method according to claim 10, wherein the extracted character string and the search keyword are normalized. 16. The information search method according to claim 10, wherein a telephone number is used as a search keyword. 17. When a web page is collected from a WWW server, a character string having the same attribute as the search keyword is extracted from the web page, and the score of the web page is scored according to the degree of description of the character string on the web page. An information retrieval method in which the ranking of the collected web pages is determined with reference to the score of the web page and the address of the web page is changed to the address of the higher web page while setting the. 18. When a web page is collected from a WWW server, a character string having the same attribute as the search keyword is extracted from the web page, and a character string that does not match the search keyword is searched from the extracted character strings. Information retrieval method. 19. A collection processing procedure for collecting a web page from a WWW server, a character string having the same attribute as a search keyword is extracted from the web page collected by the collection processing procedure, and a character string of the character string for the web page is extracted. An extraction processing procedure for setting the score of the web page according to the degree of description, a search processing procedure for searching for a character string that matches the search keyword among the character strings extracted by the extraction processing procedure, and the search described above. An information retrieval program for causing a computer to execute a ranking determination processing procedure for determining the ranking of a web page in which a character string retrieved by the processing procedure is described by referring to the score of the web page. 20. A collection processing procedure for collecting a web page containing a character string that matches a search keyword from a WWW server, and extracting a character string having the same attribute as the search keyword from the web page collected by the collection processing procedure. , Refer to the score of the web page for the extraction processing procedure for setting the score of the web page according to the degree of description of the character string on the web page and the ranking of the web pages collected by the collection processing procedure. An information retrieval program for causing a computer to execute the order determination processing procedure for determining. 21. A collection processing procedure for collecting a web page from a WWW server, a character string having the same attribute as a search keyword is extracted from the web page collected by the collection processing procedure, and The extraction processing procedure for setting the score of the web page according to the degree of description and the ranking of the web pages collected by the collection processing procedure are determined by referring to the score of the web page and the address of the web page. An information retrieval program for causing a computer to execute the order determination processing procedure for changing the address of a higher web page. 22. A collection processing procedure for collecting web pages from a WWW server, an extraction processing procedure for extracting a character string having the same attribute as a search keyword from the web pages collected by the collection processing procedure, and the extraction processing procedure. An information search program for causing a computer to execute a search processing procedure for searching a character string that does not match the search keyword from the extracted character strings.