JP2002007431A - Information searching device, information searching method and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information searching device, an information searching method and a recording medium by which only necessary contents among contents following from a prescribed start page can be acquired by giving a query such as 'I wish to acquire a homepage including one or more images, using 'KYOTO' and 'SIGHTSEEING' as a keyword and described in 'Japanese' to a searching robot. SOLUTION: Only the necessary contents are acquired by checking whether or not the contents is compatible to the inputted query.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is to automatically acquire only necessary contents when a large variety of contents are stored in a large number of contents servers as represented by WWW. Can be
The present invention relates to an information search apparatus that can be applied to various services provided by using only necessary contents from a vast amount of contents existing in a plurality of content servers provided on a computer network such as the Internet.

[0002]

2. Description of the Related Art Generally, a search robot of a search engine such as WWW follows a hyperlink provided on content such as HTML and is connected from a first designated place.
Get all kinds of content.

By the way, "Basic and Boundary A-16-46 P.3, Proceedings of the 2000 IEICE General Conference"
50 related discovery browser "AssociaView"
If the content to be searched can be specialized in a predetermined domain (for example, Kyoto sightseeing, museums, art museums, etc.), all contents are acquired in advance, and In this case, a method of extracting information about what words are included from the acquired content can be considered. However, this method is disadvantageous in that both the calculation cost and the external storage device capacity are useless. There is.

[0004]

SUMMARY OF THE INVENTION The present invention provides, for example,
A query (search request) such as "I want to obtain a home page that contains one or more images and uses the keywords" Kyoto "and" sightseeing "and is described in" Japanese "" is sent to the search robot. By giving, an information search device capable of acquiring only necessary contents among contents connected from a predetermined start page,
It is an object to provide an information search method and a recording medium.

[0005] Further, the present invention provides an information search apparatus and an information search apparatus capable of efficiently obtaining content by searching for a link that is likely to be high in content connected from a predetermined start page. It is an object to provide a search method and a recording medium.

That is, as in the above-mentioned AssociateView, in a service that uses a domain that is specialized in a domain in advance and uses content that has been limited to some extent (filtered), necessary content is automatically acquired. As above
In the case of socialview, for example, Kyoto sightseeing AV
Also, content used for things such as Edo period art AV can be automatically acquired.

A further object of the present invention is to provide an information search device, an information search method, and a recording medium applicable to various services provided by using only necessary contents from a huge amount of contents on the Internet. I do.

[0008]

SUMMARY OF THE INVENTION According to the present invention, an input Q
By checking whether or not the content conforms to the query, only the necessary content is acquired.

Further, according to the present invention, when tracing a link included in a content, information of the link itself is used to preferentially search for a link from which a necessary content is considered to exist.

[0010]

FIG. 1 is a block diagram of an information search system 100 according to an embodiment of the present invention.

[0011] The information search system 100 includes a content D
B1, a server 2, a client 3, and a computer network 4.

The content DB 1 starts searching from a predetermined start page, and registers and stores the content when the page conforms to the query (search request) (when the content score exceeds a predetermined threshold). Database.

The server 2 has a function of searching for a desired content from a number of content servers existing on the computer network 4 according to a request from the client 3.

A content DB 1, a server 2, and a client 3 are connected to a computer network 4 such as the Internet.

The information search system 100 generally has an AS
P (Application Service Pro)
vider), and is not directly visible to users who use services such as the general Internet. Therefore, the content information is not always obtained by directly using a WWW browser or the like.

The above-mentioned Query is a search request, and is information that "I want such information" passed to the server 2 in order to obtain desired information.
L is also included in the above Query.

FIG. 2 is a block diagram specifically showing the server 2 in the above embodiment.

FIG. 3 is a diagram for explaining the relationship between the search robot 30 and the link score table 22 in the above embodiment.

The server 2 has a query interpretation engine 10
Search robot management engine 20, search robot 3
0.

The Query interpretation engine 10 interprets the Query input from the client 3, that is, converts the Query into a form that can be used by the content evaluation unit 32 and the link evaluation unit 35 described later. It is. For example, a process of extracting a noun by subjecting an input natural language sentence to morphological analysis is an example of the above interpretation.

Further, for example, "keyword: Kyoto (7
0), sightseeing (30) ", the importance of the keywords constituting each query is attached numerically, and the attached importance is reflected in the interpretation of the query, whereby the content and the link are linked. When evaluating, each Query may be weighted.

The Query interpreted by the Query interpretation engine 10 is converted into a form that can be used by the content evaluation unit 32 and the link evaluation unit 35, and is used when calculating a content score and a link score. You.

Items constituting the above-mentioned Query are attributes that characterize the content. Specifically, “keyword”, “file size”, “number of characters included”, “number of nouns”, “language”, “language”, "Link presence / number / number", "Table presence / number / number", "Image presence / number / number / size", "Music presence / size / size", "Update (creation) time", "URL" and the like.

The link score is a score for efficiently searching for a link to be followed before acquiring the content of the link destination without knowing the content score of the link destination.

The search robot management engine 20 has a search robot generation unit 21 and a link score table 22.

The search robot management engine 20 has a function of dynamically calculating the number N of the search robots 30 to be moved in parallel according to the network in the server 2 and the load of the calculation amount. Has a function of managing the search robot 30 by checking how many vehicles are currently moving. Further, the search robot management engine 20
When the number of search robots 30 becomes smaller than N, a function of generating the search robot 30 with reference to the link score table 22 and generating a search robot 30 for a URL that is considered to be most suitable for Query at that time. Have.

The above “URL that seems to be the most suitable for the query at that time” refers to the link score table 22 at that time and determines the value having the highest average link score in the link score list shown in FIG. U to have
RL. Here, there are a plurality of link source URLs,
When the link scores are different from each other, the link scores are compared using the “average link score” described above.

When the role of the search robot 30 corresponding to the predetermined URL has been completed, the flag (investigated flag) indicating "already investigated?" In the link score table is set to "Yes". This investigated flag is the same U
This is a flag for eliminating waste caused by checking the RL a plurality of times. When the search robot 30 is generated next time, the UR for which the investigated flag is “Yes” is set.
For L, the search robot 30 is not generated.

The search robot 30 includes the content acquisition unit 3
1, a content evaluation unit 32, a query matching determination unit 33, a content registration control unit 34, and a link evaluation unit 3.
And 5.

The search robot 30 is controlled by the search robot generation unit 21 in the search robot management engine 20.
It is generated using one URL as an argument, and it is checked whether or not it is necessary to register the content indicated by the URL in the content DB1, and if it is necessary, the content is registered in the content DB1. When the link score is calculated for all the links in the content and registered in the link score table 22, the search robot 30 disappears.

Next, the search robot 30 will be described in detail.

In the search robot 30, the given U
Based on the RL, the content acquisition unit 31 acquires the content, and the content evaluation unit 32 calculates the content score, thereby evaluating the content.
The matching determination unit 33 determines whether or not the query is compatible.

Next, the content score calculation performed by the content evaluation section 32 will be described in detail.

The content evaluation unit 32 calculates a content score based on the query interpreted by the query interpretation engine 10. The content score is a score indicating the degree to which the predetermined content conforms to the query.

When calculating the content score,
If the degree of matching with the query is high, the calculation is performed using a calculation formula that increases the content score. For example, when "Kyoto" and "sightseeing" are input as keywords, and if the predetermined content includes "Kyoto" and "sightseeing", the content score of the predetermined content increases, and the keyword Is not included, a calculation formula that lowers the predetermined content score is used.

For attributes other than keywords (for example, an attribute indicating that the content has an image, an attribute indicating that the content is provided by a company), as described above, if the degree of matching with Query is high,
It is calculated using a calculation formula that increases the content score.

Using a thesaurus or the like, if the same keyword as the input keyword is not included in the predetermined content, but a keyword having a nuance close to the input keyword is included in the predetermined content, May use a calculation formula that increases the content score of the predetermined content to some extent.

Further, for example, the importance of the keyword “Kyoto” is set to “70”, and the importance of the keyword “sightseeing” is set to “30”.
You may make it attach to each Query as a numerical value. In this case, the importance may be reflected in the content score by, for example, taking a weighted average with a value weighted by importance for the score of each query.

Here, if the predetermined content conforms to the Query (if the content score exceeds a predetermined threshold), the content registration control unit 34 registers the predetermined content in the DB1.

When a link is included in the content, the link evaluation unit 35 calculates a link score for each of the included links. The link score obtained by this link score calculation is sent to the link score table 22 of the search robot management engine 20. Multiple link source URLs for one link destination
May exist, and a plurality of link scores are stored according to the list structure. Note that the above list structure is
This is a structure for inserting a plurality of values in a situation where the number is not known in advance, and is composed of data and a pointer to the next data.

As described above, the link score in the above embodiment is determined by the content and the information attached to the content. The “information attached to the content” is not the content itself, such as a URL, but is information that characterizes the content.

Next, the link score calculation performed by the link evaluation unit 35 will be described in detail.

The link evaluation unit 35 obtains the score of the link itself in the same manner as the content score, and calculates the link score using the content score calculated by the content evaluation unit 32.

When the content of the link source content is Query
Since the degree of matching of the contents of the link destination is likely to be affected according to the degree of matching, the link score is calculated using the content score calculated by the content evaluation unit 32. For example, when searching for a query whose keyword is “sightseeing or Kyoto”, the link provided in the “homepage on tourism” is better than the link “Kyoto” provided in the “local government introduction homepage”. "Kyoto" is Que
ry is likely to match. Therefore, the link evaluation unit 35 calculates the product of the link itself score and the content score calculated by the content evaluation unit 32 as a link score.

Next, in the above embodiment, a predetermined Que
The operation of searching for a content that matches ry will be described.

(Step 1) First, when the user enters
y is input to the client 3. This input Que
ry is accessed by the server 2 connected via the Internet 4. Prior to the search for the content, it is usually necessary to input a user name, a password, and the like, but the description of the input operation is omitted here.

The content evaluation unit 32 and the link evaluation unit 35 can use the format (interpreted Quer).
y), the query input to the client 3 is replaced by Q
The query interpretation engine 10 performs the conversion.

(Step 2) The interpreted Query is
It is sent to the content evaluation unit 32 and the link evaluation unit 35. The Query item is an attribute that characterizes the content, and specifically includes “keyword”, “file size”, “number of included characters”, “number of nouns”, “language”,
“Presence / presence / number of links”, “Presence / presence / number of tables”, “Presence / presence / number / size of images”, “Presence / presence / size of music”, “Update (creation) time”, “URL”, and the like.

(Step 3) The search start page information sent to the search robot management engine 20 is sent to the search robot 30 via the search robot generation unit 21. The content acquisition unit 31 acquires the content of the search start page, and the acquired content is sent to the content evaluation unit 32. The content evaluation unit 32
A calculation formula is used such that the content takes a value high enough to conform to the Query interpreted in step 2.

For example, when the query “Keyword: Kyoto, sightseeing” is input, 0.9 is given as a content score to the content including both the keywords “Kyoto” and “sightseeing”, and one of the keywords is given. Content that contains only
A formula that gives 0.5 and gives 0.1 as a content score is used for the content that does not include any content as the above-mentioned calculation formula.

Also, for example, the query “Keyword: Kyoto, sightseeing, image: yes, URL: ^* .co.jp”
, "Keyword: Kyoto", "keyword: sightseeing", "image: available", and "URL: ^* .co.jp (company page)" satisfy four of the four conditions. A content is given a content score of 0.9, a content that satisfies the three conditions is given a content score of 0.7, and a content that satisfies the two conditions is given a content score of 0.5. An expression that gives a content score of 0.3 to content that satisfies one condition and gives a content score of 0.1 to content that does not satisfy one condition is used as the above calculation formula.

(Step 4) Here, when the predetermined content matches the input Query (when the score exceeds a certain threshold), the content registration control unit 3
4 registers the content in DB1.

(Step 5) If links exist in the content to be searched, the link evaluation unit 35 calculates a link score for each link. The link evaluation unit 35 calculates the score of the link itself in the same manner as in the case of calculating the content score, but also calculates the link score using the content score calculated by the content evaluation unit 32.

For example, a score obtained by multiplying the score of the link itself by the content score is set as a link score. In this case, the “link itself score” for the link “Kyoto is still going to be Kyoto” is calculated in the same way as the content score calculation, so it becomes 0.5 (Keyword: Kyoto, one of sightseeing) For content including only keywords, a content score of 0.
5 is used), but if the content score of the content including the link is 0.9, the link score is 0.5 * 0.9 = 0.45.

Similarly, in the case of the link “Nara was once a city,” the link itself has a score of 0.1 (keywords: Kyoto, contents that do not include any of sightseeing include contents Since the calculation formula that gives 0.1 is used as the score), the link score is 0.1. If the content score of the content including the link is 0.9, the link score is 0.1 * 0. .9
= 0.09. The obtained link score is sent to the link score table 22 of the search robot management engine 20.

The score of the link itself is determined based on the Query, in accordance with the link destination URL and the description of the link.
It is calculated as follows: For example, if you enter the query "Keyword: Kyoto, sightseeing"
Since y does not include the URL, the description of the link is targeted. For example, 0.9 if both keywords "Kyoto" and "sightseeing" are included, and 0 if only one is included.
A calculation formula that returns 0.1 when none is included is used.

Also, "keyword: Kyoto, sightseeing, image:
Yes, URL: ^* . If you enter the query "co.jp", the same applies to "keywords",
As for “presence / absence of image”, it cannot be determined only by the link destination URL or the description of the link (it cannot be determined until the link destination content is acquired).
Ignore y.

For “URL: ^* .co.jp”,
It is determined by looking at the link destination URL. So, for example,
Links that include both the keywords “Kyoto” and “sightseeing” and that are corporate pages (co.jp) include 0.9,
"Keyword: Kyoto", "Keyword: sightseeing", "UR
L: ^* . Of the four conditions of "co.jp", use an equation such that 0.6 is satisfied if it satisfies two, 0.3 if it satisfies one, and 0.1 if it does not satisfy any one. .

(Step 6) The search robot management engine 20 dynamically calculates the number N of the search robots 30 to be moved in parallel according to the network in the server 2 and the load of the calculation amount. When the number of robots 30 is smaller than N, the search robot 30 is generated for the URL that is most likely to match the query at that time by referring to the link score table 22 at that time.

(Step 7) The search robot 30 generated in step 6 executes steps 3 to 5, whereby the content information is gradually accumulated in the content DB1.

In other words, the above embodiment does not randomly collect the linked content, but determines at the time of search whether or not the content is necessary, and collects only the necessary content. Here is an example. In other words, the above-described embodiment is an effective search robot when it is desired to narrow down the search target content to only necessary content according to a predetermined rule, for example, to specialize a domain. That is, in the above-described embodiment, only necessary contents, which cannot be realized by the conventional search robot, can be efficiently acquired.

That is, according to the above embodiment, the search is started from a predetermined start page, and the page is queried.
(If the score exceeds a certain threshold)
Then, the content is registered in the DB. After that, the score that evaluated "link" is left in the table,
Content is searched for preferentially in descending order of score. Therefore, only “width-first” and “depth-first” contents can be collected first.

In the server 2, the link evaluation section 35 and the link score table 22 may be deleted. Even in this case, only necessary contents can be collected. On the other hand, the link evaluation unit 35 and the link score table 2
2, the server 2 can end the operation of collecting only the necessary contents in a short time.

The above embodiment can be understood as the invention of a recording medium. That is, in the above embodiment, the Qu interprets the Query input from the predetermined client.
ery interpretation procedure, a search robot creation procedure for creating a search robot, a content acquisition procedure for acquiring content from the computer network, and the Quer
a content evaluation procedure for calculating a content score, which is a degree of conformity of the content to the Query, based on the y and the content; and a compatibility determination for determining whether the content conforms to the input Query. It is an example of a computer-readable recording medium on which a program for causing a computer to execute a procedure and a content registration procedure for registering content conforming to the Query in a content DB is recorded.

In this case, a link evaluation procedure for calculating a link score, which is a degree of conformity of the content indicated by the link included in the content to the query, and a link score of the link included in the predetermined content are accumulated. A program that causes a computer to execute the link score accumulation procedure may be provided.

In this case, the attributes characterizing the content are at least one of a keyword, a file size, the number of included characters, the number of nouns, a language, presence / absence / size of a link, an update (creation) time, and a URL. Using the attribute selected from the attributes, the content score and the score of the link itself are calculated.

Further, in the above embodiment, a score calculation procedure for calculating the content score and the score of the link itself based on the attribute characterizing the content, and the score calculation procedure based on the calculated content score and the score of the link itself. A computer-readable recording medium storing a program for causing a computer to execute a link score calculation procedure for calculating a link score, and a search priority determination procedure for determining a search priority based on the calculated link score. This is an example.

The recording medium is FD, CD, DV
D, HD, semiconductor memory and the like can be considered.

[0069]

According to the present invention, it is determined whether or not the content is necessary at the time of searching for the content, so that only the necessary content can be collected.

[Brief description of the drawings]

FIG. 1 is an information search system 10 according to an embodiment of the present invention.
FIG.

FIG. 2 is a block diagram specifically showing a server 2 in the embodiment.

FIG. 3 is a diagram illustrating a relationship between a search robot 30 and a link score table 22 in the embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Content DB, 2 ... Server, 3 ... Client, 4 ... Computer network, such as the internet, 10 ... Query interpretation engine, 20 ... Search robot management engine, 21 ... Search robot generation part, 22 ... Link score table, 30 ... Search Robot 31: Content acquisition unit 32: Content evaluation unit 33: Query matching determination unit 34: Content registration control unit 35: Link evaluation unit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Yoshinobu Tomura 2-3-1 Otemachi, Chiyoda-ku, Tokyo F-term in Nippon Telegraph and Telephone Corporation (reference) 5B075 ND36 NK46 PP12 PP23 PR06 PR08 PRM QM08

Claims (10)

[Claims] 1. A client connected via a predetermined computer network, a server, and a content D
B, a query interpretation engine for interpreting the query input from the client; content acquisition means for acquiring content from the computer network;
Based on the Query and the content, the Q
content evaluation means for calculating a content score that is the degree of conformity of the content to the query; suitability determination means for determining whether the content conforms to the input Query; and content conforming to the Query. An information search device, comprising: a search robot having content registration control means for registering the search robot in a content DB; and a search robot management engine having search robot generation means for generating the search robot. 2. The search robot according to claim 1, further comprising: link evaluation means for calculating a link score, which is a degree of conformity of the content indicated by the link included in the content to the query. An information search device, characterized in that the management engine has a link score table for accumulating link scores, which are scores of links included in predetermined content. 3. A client, a server, and a content D connected via a predetermined computer network.
B. an information search method of the information search system in the information search system comprising: B; a Query interpretation step of interpreting the Query input from the client; a search robot generation step of generating a search robot; Acquiring the content; and the above Qu
query based on the query and the content.
A content evaluation step of calculating a content score that is a degree of conformity of the content with respect to; a compatibility determination step of determining whether the content conforms to the input Query; and a content conforming to the Query. And a content registration step of registering the content in a content DB. 4. A link evaluation step according to claim 3, wherein a link score, which is a degree of relevance of the content pointed to by the link included in the content to the query, is calculated; and a link of the link included in the predetermined content. A link score accumulating step of accumulating scores. 5. The attribute according to claim 4, wherein the attribute characterizing the content is at least one of a keyword, a file size, the number of included characters, the number of nouns, a language, presence / absence / size of a link, an update (creation) time, and a URL. An information search method, wherein the content score and the score of the link itself are calculated using an attribute selected from the attributes. 6. A score calculating step of calculating a content score and a score of the link itself based on an attribute characterizing the content; and calculating a link score based on the calculated content score and the score of the link itself. A link score calculating step of performing a search; and a search priority determining step of determining a search priority based on the calculated link score. 7. Qu input from a predetermined client
a query interpretation procedure for interpreting the query; a search robot creation procedure for creating a search robot; a content acquisition procedure for acquiring content from the computer network; and adaptation of the content to the query based on the query and the content. A content evaluation procedure for calculating a content score that is a degree; a compatibility determination procedure for determining whether or not the content conforms to the input Query;
And a computer-readable recording medium storing a program for causing a computer to execute a content registration procedure for registering content conforming to the above in a content DB. 8. The link evaluation procedure according to claim 7, wherein a link score, which is a degree of relevance of the content pointed to by the link included in the content to the query, is calculated; and a link of the link included in the predetermined content. And a computer-readable recording medium recording a program for causing a computer to execute a link score accumulation procedure for accumulating scores. 9. The method according to claim 8, wherein the attribute characterizing the content is at least one of a keyword, a file size, the number of included characters, the number of nouns, a language, presence / absence / size of a link, an update (creation) time, and a URL. Wherein the content score and the score of the link itself are calculated using an attribute selected from the attributes. 10. A score calculation procedure for calculating a content score and a score of a link itself based on an attribute characterizing the content; and calculating a link score based on the calculated content score and a score of the link itself. A computer-readable recording medium storing a program for causing a computer to execute a link score calculation procedure to perform; and a search priority determination procedure to determine a search priority based on the calculated link score.