JP5068605B2 - Database generation method, database generation device, and computer program - Google Patents

Description

  The present invention relates to a technique for evaluating the quality of information represented by document data.

  Due to the widespread use of the Internet, general consumers evaluate various stores that provide products and services, and the evaluation information is released on the Internet. However, the information published on the Internet includes inaccurate information that is not useful to third parties. For this reason, it is not easy to extract highly accurate information from information published on the Internet.

  In a certain prior art, an address, a place name, a building name, a company name, an organization name or a phrase for deriving them are extracted from information mainly described from the home page 103, and a plurality of map databases 108 and information databases 109 are extracted. Search for that information. Then, the best candidate is obtained and a pair of information on the home page 103 and the map is created. Further, the latitude / longitude or the coordinate value of the rectangular coordinate system peculiar to the map is obtained from the top selected map, a combination of the home page 103, the information on the map and the coordinate value is created, and the coordinated home page information database 110 is obtained. Forward to.

  In this technique, the transferred homepage 103 is evaluated based on the publicity and name recognition of information described and described, the information update frequency, and the identity and reliability of the homepage creator or sender. For the evaluation of the information described, the general part-of-speech dictionary and proper noun dictionary 112 and various information databases 109 are referred to, and the frequency of appearance of information and the size of items indicating how many information sources the information is introduced. It is obtained by judging the public nature and the publicity by etc. and continuously monitoring the creation time of the homepage file. Evaluation of the identity and reliability of the creator or sender of the homepage is based on the description method of information in the homepage 103, the presence / absence of presence in the map database 108 and various information databases 109, and a search service providing device for description information Each search result from is scored, and the score in the design evaluation is added and evaluated.

JP 2000-339330 A

  However, in the above technique, the evaluation of the home page is performed based on the presence / absence of information on the home page in other existing databases, the presence frequency, the size of items, and the like. For this reason, for example, the latest information collected by general consumers at stores, such as those described in blogs (Weblogs), and articles written in original terms and expressions are meaningful information. Even if it contains, high evaluation cannot be obtained. Such problems can occur widely when evaluating the quality of information represented by a document.

  The present invention addresses at least a part of the above-described problems, and an object thereof is to provide highly reliable information by selecting a highly reliable document from a plurality of documents. .

In order to achieve the above object, the present invention performs the following processing when generating a database based on a plurality of document data.
(A) A plurality of document data including document data including an address character string that is a character string representing an address and an article character string that is a description relating to a store is prepared.
(B) Analyzing the data of the address character string to generate detail level data representing the details of the address represented by the address character string.
(C) Based on the detail level data, at least one document data including an address character string is selected from a plurality of document data.
(D) A database relating to the store is generated based on the selected document data.

  It can be presumed that the author who describes the address in detail in the document accurately and objectively makes other descriptions in the document. Therefore, if it is set as the above aspects, a highly reliable document can be selected from a plurality of documents, and a database having highly reliable information can be generated based on the highly reliable document. .

  Steps (a) to (d) can be executed by a computer. For example, the step (a) is executed when the first computer acquires character string data from original document data stored in a second computer (server) connected to the first computer via a network. be able to. Then, the step (b) can be executed by the first computer generating detailed level data and storing it in a predetermined first memory. Furthermore, the step (d) can be executed by the first computer generating a database and storing it in a predetermined second memory. The first and second memories may be provided in the first computer, or may be provided in another device connected to the first computer via a network.

  When selecting at least one document data including an address character string from a plurality of document data, selecting the document data including the most detailed address character string as one of the at least one document data. Is preferred.

  In the above (a), when preparing a plurality of document data, it is preferable to perform the following processing. That is, for a user who has made a predetermined contract, access the server of a service provider that provides a service for publishing original document data including a character string created by the user via the network via the network, The document data is generated based on the character string included in the original document data.

  If it is such an aspect, the database which can utilize the description of the store by the user who does not have a stake in a store can be created.

In the above (d), it is preferable to perform the following processing when generating a database relating to a store.
(D1) The latitude and longitude of the point represented by the address character string are specified.
(D2) A database including the latitude and longitude and the information on the acquisition destination of the address character string and the article character string of the selected document data is created.

  If it is such an aspect, the database which can obtain easily both the description regarding a store and the location of a store can be created.

Furthermore, when providing a service using a database, it is preferable to perform the following processing.
(E1) Map data representing a map of a predetermined area is prepared.
(E2) A map based on the map data is displayed.
(E3) A predetermined mark is displayed at a position corresponding to the latitude and longitude of the point represented by the address character string on the map.
(E4) Receives designation of a mark from the user, and provides data representing an acquisition destination of the address character string and the article character string of the document data.

In the above (b), when generating the detail level data representing the details of the address represented by the address character string, it is preferable to perform the following processing.
(B1) The character string included in the document data is decomposed into a plurality of words each composed of one or more characters.
(B2) The words included in the document data are examined in order, and an address character string is extracted from the document data.

And in said (b2), when extracting an address character string, it is preferable to perform the following processes.
(B3) When the first word included in the document data satisfies a predetermined first condition including that it is a word representing an area, the first word is stored at the head of the address buffer.
(B4) If the second word following the word stored in the address buffer satisfies a predetermined second condition, the second word is additionally stored in the address buffer.
(B5) When the second word satisfies a predetermined third condition different from the second condition, the character string stored in the address buffer is extracted as an address character string.

  With such an embodiment, an address character string can be extracted from document data by analyzing individual words. In addition, it is preferable that the address buffer when the character string in the address buffer is extracted as the address character string is “an address buffer in which a word is stored immediately before”.

In (b5) above, when the character string stored in the address buffer is extracted as an address character string, it is preferable to perform the following processing.
If the third condition is met,
The second word is
(Sub-condition 1) A word representing a region,
(Sub-condition 2) include any of the characters “east”, “west”, “south”, “north”, “large”,
(Sub-condition 3) A word representing a number,
(Sub-condition 4) Including the character “character”,
If none of the above applies, the character string in the address buffer is extracted as an address character string.

  With such an embodiment, an address character string that does not include information below the “chome” level can be extracted as an address character string.

In (b5) above, when the character string stored in the address buffer is extracted as an address character string, the following processing is also preferably performed.
Assuming that the third condition is satisfied, if the second word is a word representing a region higher than the word examined in (b2) immediately before the second word, Extract a string as an address string.

  According to such an aspect, when two sets of address character strings are continuously described in the document data, they can be processed as two sets of address character strings instead of a single address character string. it can.

Furthermore, in the above (b5), when the character string stored in the address buffer is extracted as the address character string, it is also preferable to perform the following processing.
If the third condition is met,
The second word is
(Sub-condition 5) A word representing a number,
(Sub-condition 6) A predetermined word that is displayed after the number and represents the unit of the number,
(Sub-condition 7) It is a predetermined word that is displayed after the noun and represents the attribute of the noun,
(Sub-condition 8) Symbol
(Sub-condition 9) being a noun,
(Sub-condition 10) A predetermined word representing the number of connections displayed after the number,
(Sub-condition 11) It is a predetermined word that is displayed at least one before and after the noun and modifies the noun,
If none of the above applies, the character string in the address buffer is extracted as an address character string.

  According to such an aspect, an address character string including information on the “chome”, “address”, or “issue” level can be extracted as an address character string.

In the above (b4), when the second word is additionally stored in the address buffer, it is preferable to perform the following processing.
If the second condition is met:
The second word is
(Sub-condition 12) Representing a region,
(Sub-condition 13) It must contain any of the characters “East”, “West”, “South”, “North”, “Large”,
If either of the above is satisfied, the second word is additionally stored in the address buffer.

  With such an aspect, it is possible to extract a character string representing an address higher than the “character” (bruise) level or the “chome” level as a part of the address character string.

In (b4) above, when the second word is additionally stored in the address buffer, it is also preferable to perform the following processing.
If the second condition is met:
The second word is
(Sub-condition 14) A word representing a number,
(Sub-condition 15) Including the character “character”,
If either of the above is satisfied, the second word is additionally stored in the address buffer.

  With such an aspect, it is possible to extract the first character representing the character of “character” (bruise) or the address of the level below the “chome” level as a part of the address character string.

Further, in the above (b4), when the second word is additionally stored in the address buffer, it is also preferable to perform the following processing.
If the second condition is met:
A word that was considered before the second word
(Sub-condition 14) A word representing a number,
(Sub-condition 15) Including the character “character”,
And either
The second word is
(Sub-condition 5) A word representing a number,
(Sub-condition 6) A predetermined word that is displayed after the number and represents the unit of the number,
(Sub-condition 7) It is a predetermined word that is displayed after the noun and represents the attribute of the noun,
(Sub-condition 8) Symbol
(Sub-condition 9) being a noun,
(Sub-condition 10) A predetermined word representing the number of connections displayed after the number,
(Sub-condition 11) It is a predetermined word that is displayed at least one before and after the noun and modifies the noun,
If either of the above is satisfied, the second word is additionally stored in the address buffer.

  With such an aspect, a character string representing an address below the “chome” level can be extracted as a part of the address character string.

When generating the detail level data indicating the details of the address represented by the address character string in (b) above, it is preferable to further perform the following processing.
(B7) With regard to the address character string, it is examined whether or not the address character string includes a plurality of predetermined words that are a part of the address and represent the area and belong to a plurality of areas having a plurality of different levels.

In (c) above, when selecting at least one document data including an address character string from a plurality of document data, the following processing is preferably performed.
(C1) Select document data based on the lowest hierarchy of a predetermined word included in the address character string.

  With such an aspect, it is possible to objectively evaluate the details of the address character string. Then, based on an objective evaluation of the details of the address character string, it is possible to select document data whose contents are highly likely to be accurate and objective. In selecting document data, it is also preferable to consider the number of predetermined words included in the address character string.

In addition, a some hierarchy may include the hierarchy of a town area and the hierarchy of a block. In such an aspect, in the above (b7), when examining whether or not the address character string includes a predetermined word, it is preferable to perform the following processing.
(B8) With regard to the address character string, it is examined whether or not a predetermined word representing the above-mentioned area is included in the order from the word representing the upper-level area to the word representing the lower-level area.

In (b8) above, when examining whether or not a predetermined word representing a zone is included in the above order, the following processing is preferably performed.
For the number of full-width or half-width minus or hyphen in the address string,
The number of full-width or half-width numbers connected by minus or hyphen, whether or not there is a word included in the town level hierarchy detected so far, and whether or not there is a word included in the block hierarchy detected so far Based on this, an evaluation equivalent to the evaluation of whether or not a predetermined word is included is performed.

  In this manner, for example, in the document data, an address below the “chome” level is “○ chome × address △ number”, “○ chome × −Δ”, or “◯ −x−Δ”. In any case, it is possible to accurately evaluate the details of the address.

In the above (c), when selecting document data, it is preferable to further perform the following processing.
(C2) The address character string includes the character string “Kyoto Prefecture” or “Kyoto City”, and “rises up”, “descends”, “enters east”, “enters west”, “enters south” , "North entry", "Upper", "Lower", "East entry", "West entry", "South entry", "North entry"
Evaluation is performed in the same manner as when the address character string does not include the character strings of “Kyoto Prefecture” and “Kyoto City” and includes the character string of “Chome”.

  With such an aspect, even when an address unique to Kyoto is written, it is possible to accurately evaluate the details of the address for such a character string. More specifically, “evaluate equally” can take the form of adding the same evaluation score or setting the same flag in the evaluation of detail.

It is also preferable to perform the following processing.
(E) A telephone number is acquired from the document data.
(F) Deciding whether or not an address character string associated with the acquired telephone number is included in the document data while referring to a database in which the telephone number and the address character string are stored in association with each other. To do.

  When selecting at least one document data from a plurality of document data, it is preferable to perform the following processing. That is, the document data including the address character string associated with the acquired telephone number is preferentially compared with the document data not including the address character string associated with the acquired telephone number. Select as document data.

  It can be presumed that the author who accurately describes the telephone number is accurately and objectively making other descriptions in the document. Therefore, if it is set as the above aspects, a highly reliable document can be selected from a plurality of documents, and a database having highly reliable information can be generated based on the highly reliable document. .

In addition, when the original document data includes link data for associating reference document data including a character string and the original document data, it is preferable to perform the following processing.
(E) Linked document data is generated based on the character string included in the reference document data while referring to the reference document data based on the link data.
(F) It is determined whether or not the link destination document data includes the address character string included in the document data.

  When selecting at least one document data from a plurality of document data, it is preferable to perform the following processing. That is, the link destination document data includes the document data that includes the address character string of the document data, and the link destination document data has priority over the document data that does not include the address character string of the document data. Select as one document data.

  It can be presumed that the author who accurately copies the address character string included in the data referred to when creating the document also accurately and objectively performs other descriptions in the document. Therefore, if it is set as the above aspects, a highly reliable document can be selected from a plurality of documents, and a database having highly reliable information can be generated based on the highly reliable document. .

  Note that “A is preferentially selected over B” means that A is selected instead of B when the other evaluation parameters considered in selection are the same in A and B. means. As a method for preferentially selecting a certain document data, for example, it is possible to change the detail level data of the document data to a value representing higher detail and execute the step (c). For example, a predetermined value may be added to the value representing the detail. Further, the value representing the detail may be multiplied by a constant. Further, when executing the step (c), one or both of “include address string associated with the acquired telephone number” and “link destination document data include address string of document data” is selected. It is also possible to adopt a mode that is a necessary condition at the time.

It should be noted that the present invention can also be implemented as a method for estimating or evaluating the accuracy of the contents of document data, for example, as follows. In this method, the following processing is performed.
(A) A plurality of document data including an address character string that is a character string representing an address and an article character string including characters other than the character string representing an address is prepared.
(B) Analyzing the address character string data and examining the details of the address represented by the address character string.
(C) An evaluation representing the accuracy of the contents of the document data is determined based on the examination result of the details.

  The “determining the evaluation” includes, for example, the following modes. (I) determining a value representing the evaluation; (Ii) For each of a plurality of sets of document data, a ranking representing the highest evaluation order is determined.

In addition, the present invention can be realized as an aspect of, for example, the following document data selection method. In this method, the following processing is performed.
(A) A plurality of document data including an address character string that is a character string representing an address and an article character string including characters other than the character string representing an address is prepared.
(B) Analyzing the address character string data and examining the details of the address represented by the address character string.
(C) At least one document data including an address character string is selected from a plurality of document data based on the examination result of details.

  The present invention can be realized in various forms, for example, a document evaluation method and a document evaluation device, a document selection method and a document selection device, a document data processing method and a document data processing device, a database providing method, and The present invention can be realized in the form of a database providing apparatus, a computer program for realizing the functions of those methods or apparatuses, a recording medium on which the computer program is recorded, and the like.

A. First embodiment:
A1. Overview of device configuration and functions:
FIG. 1 is a diagram showing an outline of a map information providing system of the present invention. Servers AS1 and AS2 of application service providers are connected to the Internet INT. An application service provider provides a service for publishing data including character strings and images created by a user via the Internet to a user who has a predetermined contract. The servers AS1 and AS2 store browsing data PD1 and PD2 that are data including characters and images created by the plurality of users. These browsing data PD1 and PD2 can be browsed from other devices connected to the Internet.

  Specifically, the browsing data PD1 and PD2 are web page and blog data. On these web pages and blogs, evaluation articles about stores such as restaurants used by users are often posted together with the addresses of the stores. The data of at least some pages of the browsing data PD1 and PD2 includes characters, images (still images, images) including an address character string that is a character string representing an address and an article character string including characters other than the address character string. Including video), as well as audio data.

  A map application server 100 is connected to the Internet INT. The map application server 100 includes a CPU 110, a memory 120, and the like. The map application server 100 associates the url (Uniform Resource Locator) of the page data including the address character string among the browsing data PD1 and PD2 in the servers AS1 and AS2 with the database DB (hereinafter “ Article link map database MDB ") is generated, and can be viewed for a fee or free of charge via the Internet.

  Furthermore, a client 200 is connected to the Internet INT. The client 200 is connected to a liquid crystal display 210 as an output device, and a keyboard 220 and a mouse 230 as input devices. The client 200 can use the article link map database MDB of the map application server 100 via the Internet INT.

  In FIG. 1, as an example, two application service provider servers and one client are shown. However, a large number of application service provider servers and clients are connected to the Internet INT.

  FIG. 2 is a diagram illustrating an example of a display on the display 210 when the article link map database MDB is used by the client 200. A map of a location designated by the user via the client 200 is displayed in the area A10 in the center of the screen. In the upper area A20, a search input window A21 and a search button A22 are displayed.

  When the user uses the article link map database MDB, the following processing is performed. That is, the user inputs a predetermined character string in the input window A21 using the mouse 230 and the keyboard 220, and clicks the search button A22 using the mouse 230. Then, the map application server 100 retrieves data including the corresponding character string from the document data collected from the browsing data PD1 and PD2 of the servers AS1 and AS2. If the document data collected from the browsing data PD1 and PD2 includes data including those character strings, the map application server 100 uses the data including these character strings (hereinafter referred to as corresponding document data). ) Is displayed on the display 210 of the client 200.

  Specifically, the map application server 100 displays a part of the corresponding document data in the area A30 on the right side of the screen on the display 210. In the area A10, a point represented by a character string of an address existing in the document data is displayed with a predetermined mark.

  In the example of FIG. 2, a character string “Italian cuisine” is input to the input window A21. A part of relevant document data generated from articles on a blog or homepage including a character string “Italian cuisine” is listed in an area A30 on the right side of the screen. And in the area A10 in the center of the screen, the point indicated by the address in the articles of those blogs is represented by a balloon BL. Note that articles that appear to be commercial advertisements are accompanied by a balloon with the letters “PR”. The user can jump to those blogs and homepages by clicking on a list of articles in the area A30 or a balloon BL in the area A10.

  According to such an aspect, the user inputs predetermined words such as “Italian food”, “delicious”, “recommended”, etc. from the input window A21, and uses those words on a blog or a homepage. A location on the map can be specified for the mentioned store with a simple procedure.

  Further, the user can easily find a store mentioned in a blog or a homepage from a predetermined area displayed on the map based on the balloon. And you can read the evaluation of the store by jumping to those blogs and homepages.

  In the following, the map application server 100 extracts document data including an address character string from the browsing data PD1 and PD2, selects document data from the document data, and links the article data with the latitude and longitude information. The process for generating the map database MDB will be described in detail.

A2. Map application server behavior:
FIG. 3 is a flowchart showing the processing of the map application server. In step S10, document data is acquired from predetermined servers AS1, AS2, etc., among servers connected to the Internet INT. In step S10, for example, html (HyperText Markup Language) data is collected from a page with the browsing data PD1 of the server AS1, and only text data is extracted from the data to generate document data. One document data consists of character data for one page displayed on the display 210 at the same time when the browsing data PD1, PD2, etc. are browsed by the client 200 or the like. In step S 10, the url of the page that is the basis of the document data (hereinafter also referred to as “article page”) is acquired and stored in the memory 120. A functional unit of the CPU 110 of the map application server 100 that performs the process of step S10 is shown in FIG.

  The document data prepared in step S10 includes document data that does not include the address character string and document data that does not include the store evaluation. However, by repeatedly executing Steps S10 to S60 in FIG. 3 to obtain character data from the browsing data PD1 and PD2 and generate a plurality of document data, at least some of the document data represents an address. The document data includes an address character string that is a character string and an article character string that is a description of the store such as introduction and evaluation of the store.

  In step S20, the sentence of the document data is decomposed into words. The “word” here is not a “word” in Japanese grammar, but a character group composed of one or more characters defined for the purpose of analyzing a document. In this specification, this character group is called a "word" for convenience. For example, the character string “Higashi Kanagawa” is one proper noun in Japanese grammar, but is divided into the words “East” and “Kanagawa” in the process of step S20.

  In step S20, the part of speech of each word is determined. Note that the “part of speech” here is not a “part of speech” in Japanese grammar, but is a classification representing the attribute of a word that is defined for the purpose of analyzing a document.

  FIG. 4 is a table showing an example of “part of speech” in the present specification. The leftmost column shows the part number classification part. The middle column shows the part of speech classification. The rightmost column shows examples of parts of speech. The “part of speech” in this specification has a classification number and one or more classifications. For example, the classification number 11 is a classification number given to “part of speech” of “noun-proprietary noun-region-general”. The word “Tokyo” is classified into “part of speech” of “noun—proper noun—region—general”. Further, the term “Tokyo” is classified into classification number 11. The classification of “part of speech” shown in FIG. 4 is an example. The process of step S20 in FIG. 3 can be executed by various engines, and the classification of “parts of speech” may differ depending on the engines.

  5 and 6 are diagrams showing examples of processing results of step S20 for address character strings in document data. 5 and 6, only the part-of-speech classification number is shown for each word. 5 and 6 show examples of address character strings in the document data. However, the processing in step S20 is not limited to such character strings, and is performed for all words in all sentences of the document data. Is called. The process performed in step S20 in FIG. 3 is referred to as “morphological analysis”.

  In step S30 in FIG. 3, a character string representing an address (hereinafter referred to as “address character string”) is extracted from the document data that is decomposed into words in step S20 and associated with parts of speech, and address list data is created. To do.

  In step S40 of FIG. 3, the degree of detail of each address character string in the address list data is determined. More specifically, detail level data representing the degree of detail is generated. A functional unit of the CPU 110 of the map application server 100 that performs the processes of steps S20 to S40 is shown as an address character string review unit 112 in FIG.

  In step S45, it is determined whether or not an address character string having a degree of detail equal to or greater than a predetermined threshold exists in the address character string of the address list data. If there is an address character string having a degree of detail equal to or greater than a predetermined threshold, the process proceeds to step S50. If there is no address character string having a degree of detail equal to or greater than a predetermined threshold, the process proceeds to step S65.

  In step S45, document data determined to include an address character string having a degree of detail equal to or greater than a predetermined threshold is selected. Document data including an address character string having a degree of detail greater than or equal to a predetermined threshold is then incorporated into the database in step S60. For document data determined not to include an address character string having a degree of detail equal to or greater than a predetermined threshold, and document data determined not to include an address character string, the processes in steps S50 and S60 are not performed. Therefore, such document data is not incorporated into the database. A functional unit of the CPU 110 of the map application server 100 that performs the process of step S45 is shown as a document data selection unit 113 in FIG.

  In step S50 of FIG. 3, an address character string with a high degree of detail is selected from the address list data. In this embodiment, the address character string having the highest level of detail is selected from the address list data. When there are a plurality of address character strings having the highest level of detail, the address character string located at the beginning of the address list data is selected.

  In step S60 of FIG. 3, the latitude and longitude represented by the address character string are specified based on the selected address character string. This process is called “geocoding”.

  In step S60, the url of the article page acquired in step S10 and the latitude and longitude obtained in step S60 are associated with each other and added to the article link map database MDB. Then, the updated article link map database MDB is stored in the memory 120. The article link map database MDB further stores document data collected from the browsing data PD1 and PD2, and the number of appearances of the same keyword of the same type in the document data.

  Here, “same type” means as follows. That is, each word is further classified into a noun, a verb, an adjective, and an adjective verb in step S20. And when the classification is the same, it is considered as the word of "the same kind".

  A functional unit of the CPU 110 of the map application server 100 that performs the process of step S60 is shown in FIG. The database generation unit 114 generates and updates the article link map database MDB.

  In step S65 of FIG. 3, it is determined whether or not the document data has been generated for all browsing data of all the servers AS1, AS2, and the like determined in advance. If the determination result is No, the process returns to step S10, and new document data is prepared from the browsing data of the servers AS1, AS2, and the like determined in advance. If the determination result is Yes, the process proceeds to step S70.

  In step S <b> 70 of FIG. 3, display data for display as shown in FIG. 2 is created based on the article link map database MDB and transmitted to the client 200 in response to a request from the user. A functional unit of the CPU 110 of the map application server 100 that performs the process of step S70 is shown as a service providing unit 115 in FIG.

  In the display based on the display data, the balloon BL is displayed at a position corresponding to the latitude and longitude obtained in step S50 on the map (see FIG. 2). Further, a part of the document data prepared in step S10 (that is, a part of the article page) is displayed in the area A30. A link to the url of the article page acquired in step S10 is embedded at a position where the balloon BL and a part of the article page are displayed.

  According to such an aspect, the article link map database MDB can be generated based on a blog or a homepage in which an amateur who has no interest has described an unfaithful store evaluation (particularly, see step S10 in FIG. 3). ). Therefore, the article link map database MDB useful for a third party can be generated and provided.

  Moreover, it can be estimated that the user who describes the address in detail in the article on the blog or the homepage is performing highly accurate store evaluation in the article at the same time. For this reason, by setting it as the above aspects (refer especially step S45), the document data of the blog and the homepage which can be estimated that the highly accurate store evaluation is performed are selected, and the article link map database MDB is based on them. Can be generated. For this reason, the highly accurate article link map database MDB can be generated.

A3. Generation of address list data:
7 and 8 are flowcharts showing detailed processing of step S30 in FIG. In step S310, one word is acquired from the document data. Note that the words are acquired from the document data in the order of the words in the document data. Therefore, when step S310 is executed for the first time, the first word of the document data is acquired.

  In step S315, the part of speech of the acquired word is “region” (classification numbers 10 to 12, 30, etc., see FIG. 4), the part of speech is not “country” (classification number 12), and the acquired word It is considered whether is not a katakana-only word.

  As described above, the part of speech has one or more layers (see FIG. 4). In the determination of “whether the part of speech is“ region ””, it is determined that “the part of speech is“ region ”” if any hierarchy has the classification “region”. The same applies to the determination of other part of speech classifications such as “country” and “classifier”.

  In addition, in the database that is referred to when assigning parts of speech to words, the part of speech of “region” among the actual place names is assigned up to the name at the municipal level. For example, “Osawa” that follows “Character” is not assigned the part of speech of “Region” but is assigned “Noun-General”.

  Of the document data, the part of speech is not “region” for words other than the word representing the address. For this reason, the determination result of step S315 is No for words other than the word representing the address. For example, when the word acquired in step S310 is “Japan”, the part of speech is “country”, and therefore the determination result in step S315 is No. Furthermore, when the word acquired in step S310 is “Virginia”, the determination result in step S315 is “No” because it is only a katakana word.

  If the determination result is No, the process returns to step S310. Then, the next word is acquired from the document data. The processes in steps S315 to S310 are repeated until the word acquired in step S310 becomes a word that satisfies the condition in step S315.

  On the other hand, for example, when the word is “Tokyo”, the determination result in step S315 is Yes. If the determination result is Yes, the process proceeds to step S320.

  In step S320, the word acquired in step S310 is stored in a new address buffer AB (see FIG. 1).

  The map application server 100 (address list creation unit 112) has a plurality of address buffers AB in the memory 120. When the processing of FIGS. 7 and 8 is performed, a character string estimated as an address is stored in one or more address buffers AB.

  In step S330, the next word is acquired from the document data. In step S335, the part of speech of the word acquired in step S330 is “region” (classification number 10-12, 30, etc.), or the word characters are “east”, “west”, “south”, “north”. ”Or“ Large ”.

  For example, if the words acquired in step S330 are “Miyako”, “Michi”, “Fuku”, “Prefecture”, “Ku”, “Kanagawa”, “Chiyoda”, “Marunouchi”, the part of speech is “Region”. Therefore, the determination result of step S335 is Yes. Also, if the word acquired in step S330 is “East” which is a part of “Higashi Kanagawa” described above, the determination result in step S335 is Yes. If the determination result is Yes, the process proceeds to step S340.

  In step S340, the word acquired in step S330 is added to the address buffer AB. The words stored in the address buffer AB in step S340 are stored in the address buffer AB following the words stored in the address buffer AB in step S320. Hereinafter, when “add to address buffer AB” is described, the same processing is performed.

  In step S350, the next word is acquired from the document data. Thereafter, the process returns to step S335. That is, as long as words that satisfy the condition of step S335 continue, these words are sequentially stored in the address buffer AB in accordance with the processing of steps S350 and S340.

  On the other hand, when the determination result of step S335 is No, the process proceeds to step S355. When the determination result in step S335 is No, the part of speech of the word acquired in step S330 is not “region”, and the characters of the word are “east”, “west”, “south”, “north”. , Which is not “Large”.

  In step S355, it is examined whether the part of speech of the word acquired in step S330 or S350 is “number” (classification number 19) or whether the word to be examined is “character” (bruise).

  If the determination result of step S355 is Yes, the process proceeds to step S360. In step S360, the word (number or “character” (bruise)) to be examined is added to the address buffer AB. In step S370, the next word is acquired from the document data.

  Note that the determination result in step S355 is “Yes” means that one or more words that are continuous in the document data and one or more words representing the region are all stored in the address buffer AB (step S335, S340, S350), followed by numbers or “characters” (bruises). Therefore, it can be presumed that the character string that follows this is a character string that represents an address below the town area, for example, “○ chome × address Δ number” or “(character) Osawa”.

  The “town area” is a word representing an address, and is a word representing an area at the “chome” level. On the other hand, the “block” is a word representing an address, and is a word representing an area at the “address” level. “No.” is a word representing an address, and is a word representing an area of “No.” level.

  On the other hand, if the determination result of step S355 is No, the process proceeds to step S390.

  Note that the determination result in step S355 is “No” means that one or more words in the document data that are continuous and one or more words representing the area are all stored in the address buffer AB (step S335 to S350), that is, numbers or “characters” (bruises) did not appear. In such a case, it can be estimated that the subsequent character string does not have information below the town area, such as “letters ◇◇”, “○ chome”, “× address”, “△ number”. That is, it can be estimated that the address character string has ended. Therefore, as will be described later, the accumulation of the character string in the address buffer AB started in step S320 ends (see step S390).

  In step S375 of FIG. 8, the part of speech of the word to be examined, that is, the last acquired word is “region” (classification numbers 10 to 12, 30, etc.), and “part of speech is“ country ”(classification number 12). ) And whether the acquired word is not a katakana-only word. The content of the determination in step S375 is the same as the content of the determination in step S315. If the determination result is Yes, the process proceeds to step S380.

  In addition, that the determination result of step S375 became Yes has the following meanings. That is, even though it has been estimated that a character string representing a number or “letter” (bruise) has already appeared in the character string, and then an address below the town area (Yes in step S355), This means that a word representing a wider area (for example, “Tokyo”) has appeared instead of “Chome” or the like. As such a case, for example, the character data may include an enumeration of a plurality of addresses. This is a case where the character string of the next address is arranged next to the character string of the address having no information below “chome”.

  In step S380, the character string stored in the address buffer AB is added to the address list data AL (see FIG. 1). The character string added to the address list data AL is a character string composed of a plurality of words. Thereafter, the process returns to step S320 (see [C] in FIGS. 7 and 8). That is, in step S320, the word determined as Yes in step S375, that is, the word that can be estimated to be the first word of the next address is stored in the new address buffer AB.

  On the other hand, if the determination result of step S375 is No, the process proceeds to step S385.

  In step S385, the part of speech of the word to be examined is “number” (classification number 19), “classifier” (classification number 35), “suffix-general” (classification number 28), “symbol-general” (classification number 77). Or “noun-general” (classification number 2), or the words to be examined are “no”, “−” (full-width minus), “-” (half-width minus), “-” (Hyphen), “B”, “F”, “floor”, “east”, “west”, “south”, or “north” are considered.

  As shown in FIG. 4, “a classifier” is a word indicating a unit of a number, such as “chome”, “number”, etc., which is shown following the number. Similarly, “B”, “F”, and “floor” are words that follow the number and represent the unit of the number.

  “Suffix-general” is a predetermined word that is displayed after a noun and represents an attribute of the noun, such as “No.”. “Symbol-general” is a word representing a symbol. “Noun—general” is a noun. "No", "-" (full-width minus), "-" (half-width minus), and "-" (hyphen) are words that are displayed following the number and represent the connection of the number. “East”, “West”, “South”, and “North” are words that are displayed before and / or after a noun and modify the noun.

  If the determination result of step S385 is Yes, the process returns to step S360 (see [D] in FIGS. 7 and 8). In the loop of steps S360, S370, S375, S385, and S360, character strings representing addresses below the town area are sequentially stored in the address buffer AB.

  On the other hand, if the determination result of step S385 is No, the process proceeds to step S390.

  In addition, it can be estimated that the character string showing the address below a town area is accumulate | stored in the address buffer AB, and the character string showing an address was complete | finished that the determination result of step S385 became No.

  In step S390, the character string stored in the address buffer AB is added to the address list data AL.

  In step S395, it is examined whether or not the examinations in steps S315 to S385 have been completed for all words in the document data. If the determination result is No, the process returns to step S310 (see [E] in FIGS. 7 and 8). If the determination result is Yes, the processing in FIGS. 7 and 8 (the processing in step S30 in FIG. 3) ends.

  By performing the processing of FIGS. 7 and 8, it is possible to extract address character strings described in various details and notation methods from document data made up of characters.

A4. Evaluation of address string details:
9 to 11 are flowcharts showing detailed processing of step S40 of FIG. In step S40 of FIG. 3, the degree of detail is determined for the address character string of the address list data AL (see FIG. 1). In step S40 of FIG. 3, the processes of FIGS. 9 to 11 are executed for each address character string. In the processing of FIGS. 9 to 11, whether or not the address character string includes a predetermined word that is a part of the address and represents a zone and belongs to a plurality of different zones is determined in order. Be considered.

  In step S405 in FIG. 9, it is determined whether the address character string includes any of the characters “Miyako”, “Michi”, “Fu”, and “Prefecture”. If the determination result is Yes, the prefecture flag is turned ON in step S410. The fact that the prefecture flag is ON indicates that the address character string includes address information at the prefecture level.

  After step S410, the process proceeds to step S415. Similarly, when the determination result is No in step S405, the process proceeds to step S415.

  Each flag described in FIGS. 9 to 11 is stored in the memory 120. In the initial state, each flag is OFF.

  In step S415, it is determined whether or not the character “Kyoto” is included in the address character string. If the determination result is Yes, the Kyoto flag is turned ON in step S420. The fact that the Kyoto flag is ON indicates that the address represented by the address character string is an address of Kyoto Prefecture or Kyoto City. The address of Kyoto City may have a unique notation different from other regions, such as “Shijo Kawaramachi Nishiru”.

After step S420, the process proceeds to step S425. If the determination result in step S415 is No, the process similarly proceeds to step S425.

  In step S425, it is determined whether any one of the characters “city”, “town”, and “village” is included in the address character string. If the determination result is Yes, the municipality flag is turned ON in step S430. The fact that the municipality flag is ON indicates that the address character string includes address information at the municipality level.

  After step S430, the process proceeds to step S435. If the determination result in step S425 is No, the process similarly proceeds to step S435.

  In step S435, it is determined whether or not the characters “Kyoto City” are included in the address character string. If the determination result is Yes, the Kyoto flag is turned ON in step S440.

  After step S440, the process proceeds to step S445. If the determination result is No in step S435, the process similarly proceeds to step S445.

  In step S445, it is determined whether the address character string includes the character “chome”. If the determination result is Yes, the town area flag is turned ON in step S450. The fact that the town area flag is ON indicates that the address character string includes address information at the chome level of “○ chome × address Δ number”.

  After step S450, the process proceeds to step S455. Similarly, when the determination result is No in step S445, the process proceeds to step S445.

  In step S455, it is determined whether the Kyoto flag is ON. If the determination result is Yes, the process proceeds to step S505 in FIG. On the other hand, if the determination result is No, the process proceeds to step S465 in FIG.

  In step S465, it is determined whether the address character string includes the character “No.”. If the determination result is Yes, the block flag is turned ON in step S470. The fact that the block flag is ON indicates that the address character string includes the address level address information of “○ chome × address Δ number”.

  After step S470, the process proceeds to step S475. If the determination result is No in step S465, the process similarly proceeds to step S475.

  In step S475, it is determined whether the address character string includes the character “No.”. If the determination result is Yes, the number flag is turned ON in step S480. The fact that the number flag is ON indicates that the address character string includes the information of the address at the level of “○ chome × address Δ number”.

  After step S480, the process proceeds to step S485. If the determination result in step S475 is No, the process proceeds to step S485 in the same manner.

  In steps S405, S425, S445, S465, and S475 described above, words (for example, "city", "road", "prefecture", "prefecture") that represent lower-order areas (for example, , “No.”), in order, whether or not a predetermined word representing an area is included is examined, and each determination result is stored by ON / OFF of a flag.

  In step S485, it is determined whether or not there are two or more numbers and "-" (full-width minus), "-" (half-width minus), or "-" (hyphen) in the address character string. . For example, if there is a character string such as “3-”, “3-”, “3-” in the address character string, the determination result is Yes. If the determination result of step S485 is Yes, the process proceeds to step S610 in FIG.

  If the determination result in step S485 is Yes, in the address character string, at least a part of the information of “○ chome × address △ number” is represented by a notation in which numbers are connected with hyphens or minuses. Can be estimated. For example, it is a case where “4-chome 3-1” is described.

  On the other hand, if the determination result of step S485 is No, the process proceeds to step S490 of FIG.

  If the determination result in step S485 is No, it can be estimated that the information of “○ chome × address Δ number” in the address character string is not represented by a notation in which numbers are connected with hyphens or minuses. That is, in the process up to step S475, it can be considered that all address information is examined and the examination result is reflected in each flag.

  In step S490, the degree of detail of the address character string is determined based on ON / OFF of the prefecture flag, the municipality flag, the town area flag, the city block flag, and the number flag.

  Specifically, when the number flag is ON, the detail level of the address character string is 5. When the block flag is ON, the detail level of the address character string is 4. When the town area flag is ON, the detail level of the address character string is 3. When the municipality flag is ON, the detail level of the address character string is 2. When the prefecture flag is ON, the detail level of the address character string is 1. If the plurality of flags are ON, a higher level of detail among the corresponding levels of detail is assigned to the address character string. Detail level data representing the level of detail determined in this way is stored in the memory 120.

  That is, the level of detail of the address character string is determined based on the lowest hierarchy of words representing regions included in the address character string. By performing such processing, the level of detail for the address character string can be objectively determined.

  FIG. 10 is a flowchart showing the processing for examining the level of detail of the address character string when the Kyoto address is described in the address character string (see steps S415, S435, and S455 in FIG. 9). The process in FIG. 10 is executed when the determination result in step S455 in FIG. 9 is Yes (see [I] in FIGS. 9 and 10).

  In step S505, the address string is “up”, “down”, “enter east”, “enter west”, “enter south”, “enter north”, “upper”, “lower”, “east”, It is determined whether or not there are the characters "West entry", "South entry", and "North entry". If the determination result is Yes, the town area flag is turned ON in step S510.

  That is, in the examination of the details of the address string, “up”, “down”, “east entry”, “west entry”, “south entry”, “north entry”, “upper”, “lower”, “ The fact that there is a notation of “East Iru”, “West Ile”, “South Ile”, “Kitairi” is evaluated as much as the fact that there is “Chome” when the Kyoto flag is not ON. The Thereafter, the processing returns to step S465 in FIG. 9 (see [K] in FIGS. 9 and 10).

  By performing the processing of steps S455, S505, and S510, even when the address character string has an address unique to Kyoto, it is possible to evaluate the level of detail of the address character string.

  In addition, even when an address in Kyoto is written, it may be written in a system like the usual “○ chome × address △ number”. In the present embodiment, a normal step S465 is performed after the processing of steps S455, S505, and S510. For this reason, the details of the address character string can be evaluated even when the address in Kyoto is represented by a method such as a normal “○ chome × address Δ number”.

  FIG. 11 is a flowchart showing the processing for examining the level of detail of the address character string when the chome, address, and number are connected and represented by a hyphen or the like in the address character string. The process in FIG. 11 is executed when the determination result in step S485 is Yes (see [J] in FIGS. 9 and 10).

  In step S610, two or more character strings composed of numbers and "-", "-", "-", etc. (full-width or half-width minus or hyphen) detected in step S485 of FIG. Is divided at the minus and hyphen part.

  In step S615, it is determined whether there are three division results. If the determination result is Yes, the process proceeds to step S620. The determination result in step S615 being Yes means that “○ chome × address Δ number” is described as “◯ − × −Δ” or the like in the address character string. In the following, address notation will be described using full-width minus or hyphen among full-width or half-width minus or hyphen as an example.

  In step S620, the town area flag, the city block flag, and the number flag are turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  If the determination result of step S615 is No, the process proceeds to step S625.

  In step S625, it is determined whether there are two division results. If the determination result is Yes, the process proceeds to step S627. If the determination result in step S625 is Yes, it means that “○ chome × address” is written as “○ −x” in the address character string, or “○ chome × address △ number” is “○ chome × −”. It means that it is written as “Δ” or the like.

  In step S627, it is determined whether or not the town area flag indicating whether or not the address character string includes “chome” level information is ON. If the determination result is Yes, the process proceeds to step S630. The determination result in step S627 being Yes means that “○ chome × address Δ number” is described as “○ chome × −Δ” or the like in the address character string.

  In step S630, the block flag corresponding to the “address” level information and the number flag corresponding to the “number” level information are turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  If the determination result of step S627 is No, the process proceeds to step S640. If the determination result in step S627 is No, it means that the address character string does not include information on the “No.” level, and “○ chome × address” is written as “◯ − ×”. That is.

  In step S640, the town area flag corresponding to the “chome” level information and the city block flag corresponding to the “address” level information are turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  On the other hand, if the determination result of step S625 is No, the process proceeds to step S645. If the determination result in step S625 is No, it means that two or more character strings composed of numbers and full-width or half-width minus or hyphen detected in step S485 are one number and one. It is composed of two full-width or half-width minus or hyphen.

  In step S645, it is determined whether or not a block flag indicating whether or not the address character string includes “address” level information is ON. If the determination result is Yes, the process proceeds to step S650. If the determination result in step S645 is Yes, it means that “× address Δ number” is written as “× address−Δ” or the like in the address character string.

  In step S650, the number flag corresponding to the “number” level information is turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  On the other hand, if the determination result of step S645 is No, the process proceeds to step S655.

  In step S655, it is determined whether the town area flag indicating whether or not the address character string includes information of the “chome” level is ON. If the determination result is Yes, the process proceeds to step S660. The determination result in step S655 being Yes means that “○ chome × address” is written as “○ chome-x” or the like in the address character string.

  In step S660, the block flag corresponding to the “address” level information is turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  On the other hand, if the determination result of step S655 is No, the process proceeds to step S670. The determination result in step S655 being No means that “○ chome” is written as “− ○” or the like in the address character string.

  In step S670, the town area flag corresponding to the “chome” level information is turned ON. Thereafter, the processing proceeds to step S490 in FIG. 9 (see [L] in FIGS. 9 and 11).

  In the processing of steps S405 to S475 in FIG. 9, whether the address character string includes predetermined words representing the areas in order from the word representing the upper hierarchy area to the word representing the lower hierarchy area. It is considered whether or not. In the processing of FIG. 11, the numbers connected by minus or hyphen of full-width or half-width in the address character string have been detected so far (see steps S615 and S625 of FIG. 11). Based on ON / OFF of a town area flag indicating the presence / absence of a word included in the hierarchy of the town area, and ON / OFF of a city area flag indicating presence / absence of a word included in the area of the town block detected so far (FIG. In step S627, S645, and S655), the town area flag, the city block flag, and the number flag are turned ON or OFF (see steps S630, S640, S650, S660, and S670). For this reason, it is possible to accurately evaluate the details of an address character string in which a part or all of the information such as “○ chome × address Δ No.” is represented by numbers connected by hyphens or the like.

B. Second embodiment:
In the second embodiment, after the detail level of the address character string is determined in step S40 of FIG. 3, the detail level is further modified in consideration of other factors before the process of step S45. Further, the map application server 100 (see FIG. 1) has a telephone number database that holds a telephone number, an owner or store name of the telephone number, and an address thereof in association with each other. The other points of the second embodiment are the same as those of the first embodiment.

  FIG. 12 is a flowchart showing the process S42 in which the map application server modifies the detail level. The process S42 in FIG. 12 is performed between step S40 and step S45 in FIG. 12 is realized by the address character string review unit 112 as a function unit of the CPU 110.

  In step S710, a telephone number is acquired from the document data acquired in step S10 of FIG. A phone number is, for example, 10 consecutive full-width or half-width characters that do not include any characters other than full-width or half-width "-" (hyphen), "-" (minus), or ")" (parentheses). By retrieving the character string, it can be extracted from the document data.

  In step S720, the map application server 100 refers to the telephone number database and acquires the store name or owner corresponding to the telephone number acquired in step S710, and the address.

  In step S725, it is determined whether the store name or owner acquired in step S720 of FIG. 12 and the address exist in the document data acquired in step S10 of FIG. If the store name or owner and the address exist in the document data, the process proceeds to step S730. If neither the store name nor the owner exists in the document data, and the address does not exist in the document data, the process ends. Although not shown in FIG. 12, if the store name or owner of the telephone number corresponding to the telephone number and the address cannot be acquired in step S720, the process is also terminated.

  In step S730, the level of detail of the address character string determined in step S40 of FIG. 3 is modified. Specifically, the detail level of the address character string is changed to a higher detail level. In the second embodiment, the level of detail is increased by 2. Thereafter, the process ends.

  A user who accurately describes a telephone number in an article on a blog or a homepage can be estimated at the same time as performing highly accurate store evaluation in the article. For this reason, by setting it as an aspect like 2nd Example, the blog and the homepage document data which can be estimated that the highly accurate store evaluation is performed are selected, and based on them, the article link map database MDB is generated. be able to. For this reason, the highly accurate article link map database MDB can be generated.

C. Third embodiment:
Also in the third embodiment, after the detail level of the address character string is determined in step S40 of FIG. 3, the detail level is further modified in consideration of other factors before the process of step S45. In the third example, link destination data representing the link destination is acquired from the browsing data in step S10 of FIG. The other points of the third embodiment are the same as those of the first embodiment.

  FIG. 13 is a flowchart showing the process S43 in which the map application server modifies the level of detail. Process S43 in FIG. 13 is performed between step S40 and step S45 in FIG. Note that the process S43 of FIG. 13 is realized by the address character string review unit 112 as a functional unit of the CPU 110.

  In step S810, link destination document data is acquired from the link destination page represented by the link destination data acquired in step S10. In the third embodiment, link destination data representing a link destination is acquired in advance from the browsing data in step S10 of FIG.

  In step S810, specifically, html data is collected from a linked page that is stored in a server on the Internet. Then, only character data is extracted from them, and linked document data is generated. The content of the process in step S810 is the same as the process in step S10 of FIG. 3 except that data is acquired from the linked page.

  In step S820, the sentence of the link destination document data is decomposed into words. The process of step S820 is the same as the process of step S20 in FIG. 3 except that the target document data is the target.

  In step S830, an address character string is extracted from the linked document data that is decomposed into words in step S820 and each part of speech is associated with each other, thereby creating linked address list data. The process of step S830 is the same as the process of step S30 of FIG. 3 except that linked address list data is created for the linked document data.

  In step S835, the linked address list data generated in step S830 and the address list data generated in step S30 of FIG. 3 are collated to determine whether or not the same address character string is included in both. If the same address character string is included in the linked address list data and the address list data, the process proceeds to step S840. If the same address character string is not included, the process ends. Although not shown in FIG. 13, the process is also terminated when an address character string is not extracted in step S830.

  In step S840, the level of detail of the address character string determined in step S40 of FIG. 3 is modified. Specifically, the detail level of the address character string is changed to a higher detail level. In the third embodiment, the level of detail is increased by 2. Thereafter, the process ends.

  It can be presumed that a user who accurately copies an address described in a link destination in a blog or a homepage article is performing a highly accurate store evaluation in the article at the same time. For this reason, by setting it as an aspect like 3rd Example, the document data of the blog and the homepage which can be estimated that the highly accurate store evaluation is performed are selected, and the article link map database MDB is generated based on the data. be able to. For this reason, the highly accurate article link map database MDB can be generated.

D. Variations:
The present invention is not limited to the above-described examples and embodiments, and can be implemented in various modes without departing from the gist thereof. For example, the following modifications are possible. In addition, the modification shown below and said 1st-3rd Example can be combined suitably, and this invention can be implement | achieved. For example, a database with higher accuracy can be generated by performing both the processing of FIG. 12 of the second embodiment and the processing of FIG. 13 of the third embodiment in determining the degree of detail.

D1. Modification 1:
In the above-described embodiment, the document data is data of articles described on a blog or a homepage published on the Internet. However, the document data can also be other data, such as data available on an in-house network. That is, the document data can be any document data that can be obtained by any method. Such document data preferably includes an address character string that is a character string representing an address. Such document data preferably includes articles relating to store introductions and evaluations as article character strings.

D2. Modification 2:
In the above-described embodiment, the article pages of the browsing data PD1 and PD2 have been described by taking an evaluation article of a restaurant as an example. However, the article page of the browsing data PD1 and PD2 may be an article about a store that sells groceries and daily necessities, or may be a movie review of a movie being screened in a movie theater. In other words, the article page and the document data that can be generated based on the article page can include an article about an arbitrary store whose location can be specified by an address.

D3. Modification 3:
In the above embodiment, the level of detail of the address character string is determined, and the document data is selected based on the level of detail (see steps S40 and S45 in FIG. 3). However, the selection of the document data may be performed directly based on data such as a flag indicating the details of the address character string. In other words, the selection of the document data is not limited to the degree of detail, but can be performed based on the detail level data in an arbitrary format that represents the result of examining the details of the address character string.

D4. Modification 4:
In the above embodiment, the address character string with the highest degree of detail is selected from one document data (see step S50 in FIG. 3). However, when selecting an address character string, it is possible to select one or more address character strings whose degree of detail is equal to or greater than a predetermined threshold value. Further, when selecting an address character string, a predetermined number of two or more can be selected from the address character string having the highest degree of detail. Furthermore, when selecting an address character string, it is also possible to select all address character strings having the highest level of detail.

D5. Modification 5:
In the above embodiment, the url of the document data is acquired (step S10 in FIG. 3) and stored in the database in association with the latitude and longitude of the address character string (step S60). However, the data stored in the database in association with the latitude and longitude may be based on other description methods such as uri (Uniform Resource Identifier). That is, the data stored in the database in association with the latitude and longitude information can be in any format as long as the data represents the location of the document data.

D6. Modification 6:
Details of the address character string in the above embodiment are determined in order from the front of the address character string in the order of predetermined words (“city”, “road”, “fu”, “prefecture”, “city”, “ It is preferable to consider whether it falls under “town” or “village”). That is, if there is a character “Miya” in a certain address character string (Yes in step S405 in FIG. 9), then, in the address character string, in the part after the character “Miya”, the next It is preferable to examine whether or not there is a character of “city”, “town”, or “village” in the hierarchy (see step S425).

  In the above embodiment, the part of speech of “region” is assigned up to the name of the municipal level. Then, the part of speech of “region” is not assigned to a place name at a level lower than the municipality, such as “Osawa” following “letter”. However, it is possible to assign a part of speech of “region” to an address below the municipality level. In such an aspect, in step S385 of FIG. 8, it is preferable to have the condition that “part of speech is“ region ”” as one of the optional conditions.

D7. Modification 7:
In the above-described embodiment, document data is generated from browsing data of a predetermined server, a database is generated, and then a service using the database is provided. However, generation and update of a database and provision of a service using the database can be performed in parallel. In such an aspect, access to the browsing data of the server is performed for a predetermined number of thresholds (for example, 100 times) that can be accessed within a predetermined time (for example, 5 minutes, 10 minutes, 30 minutes, etc.). 500 times, 1000 times, etc.) is preferably performed at such a frequency that does not exceed.

D8. Modification 8:
In the above embodiment, the hardware configuration of the server 100 refers only to the CPU 100 and the memory 120. The CPU 100 can be composed of a single CPU or a plurality of CPUs. Further, the memory 120 can be a semiconductor memory, and can also be a fixed disk or another writable recording medium.
The configurations of the CPU 100 and the memory 120 may be stored in a single server, or may be provided as a plurality of devices connected via a network.

  Similarly, each functional unit of the CPU can also be realized by a CPU in a single server, and the CPUs of a plurality of devices connected via a network can each realize a predetermined functional unit. it can. Each functional unit itself may be realized by a single CPU or a plurality of CPUs.

  Further, for the application service provider server, one browsing data may be stored in one server, or may be distributed and stored in two or more servers.

D9. Modification 9:
In the above embodiment, a part of the configuration realized by hardware may be replaced with software, and conversely, a part of the configuration realized by software may be replaced by hardware.

  A computer program that realizes such a function is provided in a form recorded on a computer-readable recording medium such as a floppy disk, a CD-ROM, or a DVD. The host computer reads the computer program from the recording medium and transfers it to the internal storage device or the external storage device. Alternatively, the computer program may be supplied from the program supply device to the host computer via a communication path. When realizing the function of the computer program, the computer program stored in the internal storage device is executed by the microprocessor of the host computer. Further, the host computer may directly execute the computer program recorded on the recording medium.

  In this specification, the host computer is a concept including a hardware device and an operation system, and means a hardware device that operates under the control of the operation system. The computer program causes such a host computer to realize the functions of the above-described units. Note that some of the functions described above may be realized by an operation system instead of an application program.

  In the present invention, the “computer-readable recording medium” is not limited to a portable recording medium such as a flexible disk or a CD-ROM, but an internal storage device in a computer such as various RAMs and ROMs, An external storage device fixed to a computer such as a hard disk is also included.

The figure which shows the outline of the map information provision system of this invention. The figure which shows an example of the display on the display 210 at the time of using the article link map database MDB with the client 200. FIG. The flowchart which shows the process of a map application server. The table | surface which shows the example of "part of speech" in this specification. The figure which shows the example of the process result of step S20 about the character string of the address in document data. The figure which shows the example of the process result of step S20 about the character string of the address in document data. The flowchart which shows the detailed process of step S30 of FIG. The flowchart which shows the detailed process of step S30 of FIG. The flowchart which shows the detailed process of step S40 of FIG. The flowchart which shows the process of examination of the detail level of the address character string when the address notation of Kyoto is written in the address character string. The flowchart which shows the process of examination of the detail level of the address character string when the chome, the address, and the number are connected and represented by a hyphen etc. in the address character string. The flowchart which shows the process in which a map application server changes detail. It is a flowchart which shows the process in which a map application server changes a detail level.

Explanation of symbols

100 ... Map application server 110 ... CPU
DESCRIPTION OF SYMBOLS 111 ... Document data acquisition part 112 ... Address character string examination part 113 ... Document data selection part 114 ... Database production | generation part 115 ... Service provision part 120 ... Memory 200 ... Client 210 ... Liquid crystal display 220 ... Keyboard 230 ... Mouse A10 ... Map display Area A20 ... Area where input window and search button are displayed A21 ... Input window A22 ... Search button A30 ... Area where part of document data is displayed AB ... Address buffer AL ... Address list data AS1, AS2 ... Server BL ... Balloons displayed on the map INT ... Internet MDB ... Article link map database PD1, PD2 ... Reading data

Claims (18)

A method for generating a database based on a plurality of document data,
(A) preparing a plurality of document data including an address character string that is a character string representing an address and an article character string that is a description about a store;
(B) analyzing the data of the address character string to generate detail level data representing details of the address represented by the address character string;
(C) selecting at least one document data including the address character string from the plurality of document data based on the detail level data;
(D) generating a database relating to the store based on the selected document data. The method of claim 1, comprising:
The step (c)
A method of selecting document data including the most detailed address character string as one of the at least one document data. The method according to claim 1 or 2, comprising:
The step (b)
Analyzing the data of the address string and examining details of the address represented by the address string;
Determining an evaluation representing the accuracy of the content of the document data based on the result of examination of the details. The method of claim 3 , comprising:
The step of determining the evaluation representing the accuracy includes at least one of determining a value representing the evaluation and determining a rank representing the order of the evaluation with respect to the plurality of document data. Including a method. The method of claim 1, comprising:
The step (a)
Accessing a server of a service provider that provides a service for publishing original document data including a character string and created by the user via a network to a user who has made a predetermined contract via the network, Generating the document data based on the character string included in the original document data,
The original document data further includes link data for associating the reference document data including a character string with the original document data,
The method further comprises:
(E) generating linked document data based on the character string included in the reference document data while referring to the reference document data based on the link data;
(F) determining whether the link destination document data includes the address character string included in the document data;
In the step (c), priority is given to document data in which the linked document data includes the address character string of the document data, compared to document data in which the linked document data does not include the address character string of the document data. In particular, the method is a step of selecting as the at least one document data. A method according to any of claims 1 to 5 , comprising
The step (d)
(D1) specifying the latitude and longitude of the point represented by the address character string;
(D2) creating the database including the latitude and longitude and the address character string of the selected document data and the information on the acquisition destination of the article character string in association with each other. A method according to any of claims 1 to 6 , comprising
The step (b)
(B1) decomposing a character string included in the document data into a plurality of words each including one or more characters;
(B2) sequentially examining the words included in the document data, and extracting the address character string from the document data.
The step (b2)
(B3) storing the first word at the head of the address buffer when the first word included in the document data satisfies a predetermined first condition including that it is a word representing an area;
(B4) additionally storing the second word in the address buffer when a second word following the word stored in the address buffer satisfies a predetermined second condition;
(B5) extracting the character string stored in the address buffer as the address character string when the second word satisfies a predetermined third condition different from the second condition; Including a method. The method of claim 7, comprising:
The step (b5)
As the case where the third condition is satisfied,
The second word is
It ’s a local word,
Contain any of the characters “east”, “west”, “south”, “north”, “large”,
The word represents a number,
Contain the word “letter”,
The method includes the step of extracting a character string in the address buffer as the address character string if none of the above applies. The method according to claim 7 or 8, comprising:
The step (b5) further includes
When the third condition is satisfied, when the second word is a word representing a region higher than the word examined in the step (b2) immediately before the second word, Extracting the character string in the address buffer as the address character string. A method according to any of claims 7 to 9, comprising
The step (b5) further includes
As the case where the third condition is satisfied,
The second word is
The word represents a number,
A given word that appears after the number and represents the unit of the number,
A given word that appears after the noun and represents the attribute of the noun,
Be a symbol,
Being a noun,
A predetermined word that appears after the number and represents the connection of the number,
A predetermined word that appears before and / or after the noun and modifies the noun,
The method includes the step of extracting a character string in the address buffer as the address character string if none of the above applies. A method according to any of claims 7 to 10, comprising
The step (b4)
As a case where the second condition is satisfied,
The second word is
Representing the region,
Contain any of the characters “east”, “west”, “south”, “north”, “large”,
And if the second word is satisfied, the method further comprises storing the second word in the address buffer. A method according to any of claims 7 to 11, comprising
In the step (b4),
As a case where the second condition is satisfied,
The second word is
The word represents a number,
Contain the word “letter”,
And if the second word is satisfied, the method further comprises storing the second word in the address buffer. A method according to any of claims 7 to 12, comprising
In the step (b4),
As a case where the second condition is satisfied,
A word considered before the second word is
The word represents a number,
Contain the word “letter”,
And either
The second word is
The word represents a number,
A given word that appears after the number and represents the unit of the number,
A given word that appears after the noun and represents the attribute of the noun,
Be a symbol,
Being a noun,
A predetermined word that appears after the number and represents the connection of the number,
A predetermined word that appears before and / or after the noun and modifies the noun,
And if the second word is satisfied, the method further comprises storing the second word in the address buffer. A method according to any of claims 1 to 13, comprising
The step (b) further includes:
(B7) including a step of examining whether or not the address character string includes a plurality of predetermined words that are a part of the address and represent a zone and belong to a plurality of zones of different levels. ,
The step (c)
(C1) A method including the step of selecting the document data based on the lowest hierarchy of the predetermined word included in the address character string. 15. The method of claim 14, wherein
The plurality of levels include a town level and a block level,
The step (b7)
(B8) including a step of examining whether or not the address character string includes a predetermined word representing the area in an order from a word representing an upper hierarchy area to a word representing a lower hierarchy area;
The step (b8)
For the number of full-width or half-width minus or hyphen in the address string,
The number of the full-width or half-width minus or hyphen-connected number, and
Evaluation of whether or not the predetermined word is included based on the presence / absence of a word included in the district hierarchy detected so far and the presence / absence of a word included in the block hierarchy detected so far Including a step of performing an evaluation equivalent to. The method of claim 1, further comprising:
(E) obtaining a telephone number from the document data;
(F) Deciding whether or not an address character string associated with the acquired telephone number is included in the document data while referring to a database in which the telephone number and the address character string are stored in association with each other. Including the steps of:
In the step (c), the document data including the address character string associated with the acquired telephone number is preferentially compared with the document data not including the address character string associated with the acquired telephone number. A method of selecting as the at least one document data. A database generation device that generates a database based on a plurality of document data,
(A) a document data acquisition unit that prepares a plurality of document data including an address character string that is a character string representing an address and an article character string that is a description about a store;
(B) Analyzing the data of the address character string, and generating an address character string examination unit that generates detail level data representing the details of the address represented by the address character string;
(C) a document data selection unit that selects at least one document data including the address character string from the plurality of document data based on the detail level data;
(D) A database generation device including a database generation unit that generates a database relating to a store based on the selected document data. A computer program for generating a database based on a plurality of document data,
(A) a function of preparing a plurality of document data including an address character string that is a character string representing an address and an article character string that is a description about a store;
(B) a function of analyzing the data of the address character string and generating detail level data representing the details of the address represented by the address character string;
(C) a function of selecting at least one document data including the address character string from the plurality of document data based on the detail level data;
(D) A computer program capable of causing a computer to realize a function of generating a database relating to a store based on the selected document data.

Images