JP5195281B2 - Input candidate recommendation method and input candidate recommendation program - Google Patents

Description

  The present invention provides an input candidate recommendation method for displaying a list of input candidates for each input field of an input form displayed on a display of a user terminal on the display of the user terminal based on a past input history, and The present invention relates to an input candidate recommendation program for realizing such an input candidate recommendation method.

  For example, on a computer network such as the Internet, various documents such as application forms and application forms that have been exchanged in a conventional paper form are digitized and exchanged. For example, on the Internet, a website requesting such a digitized document is preliminarily uploaded with a digitized template that includes one or more input fields (hereinafter referred to as “input forms”). A user who wishes to issue / submit such a digitized document to a website operates the web browser program executed on his / her terminal, and thereby the template of the digitized document is transferred to the website (web server). ), And after completing the digitized document by inputting (supplementing) necessary information in each input form, the system is configured to transmit the digitized document to the website.

  By the way, in order to facilitate the input (replenishment) of information to the input form, the currently popular main web browser programs generally include the input forms in the user terminals in which the web browser program is installed. Each input history is stored, the past input history is presented for each input form, and when the user arbitrarily selects a character string included in each input history, the selected character string is input to each input form ( A function that can be supplemented) is provided.

  Also, in a certain business system, when the input history of each input form at each terminal is stored on the server side and any template is displayed on the terminal display, each input form of that template is displayed. The server apparatus is provided with a function of presenting input candidates based on the input history for and selecting a character string to be input (supplemented) from the corresponding input form.

Furthermore, for example, an OS (operating system) such as Windows (registered trademark of Microsoft Japan Co., Ltd.) presents input candidates from a dictionary or past input history, and allows the user to select an input candidate for easy character string input. (In the above OS, a function called IME) is incorporated, and according to this, since the input history is accumulated regardless of the input form, the same applies to any input form. Input candidates are presented.
JP 2000-276487 A JP-A-8-255174

  However, according to the former two functions, an input history is created for each value of the NAME attribute assigned to each input form, so that the input history is not shared between input fields to which different NAME attributes are assigned. . The input history is managed for each input form, and the input history in another input form is not referred to. Therefore, even if a desired value has been input in another input form before, the input history is diverted. I couldn't.

  In addition, according to the latter function, one huge input history that is stored independently of each input form is referenced, so the number of input candidates is too large, and unrelated input candidates. This presents a problem that the input candidate desired by the user is less likely to be presented and selection becomes difficult.

  Therefore, the problem of this proposal is to refer not only to the history of character strings input in the past for the same input form, but also to the history of character string input in input forms that tend to input similar character strings. The present invention proposes an input candidate recommendation method capable of presenting input candidates and an input candidate recommendation program for causing a computer to perform the service of the input candidate recommendation method.

  In the present plan, for each of a plurality of predetermined input forms, an input history of a character string input in the past is stored in a database, and an input target form that is an input form selected as an input target by the computer operator is described. An input history and an input history for a non-input target form that is an input form other than the input target form are respectively acquired from the database. Of the character strings included in the input history for the non-input target form, those that are not included in the input history for the input target form, respectively, the similarity to all the character strings included in the input history for the input target form The character string included in the input history for the input target form is a predetermined number of character strings having a high sum of the similarities among the character strings included in the input history for the non-input target form. At the same time, it is output as an input candidate.

  As described above, according to the present plan, the character strings included in the input history of the input target form selected by the operator are output as input candidates, and the character strings included in the respective input histories of the non-input target forms. Among these, a predetermined number of character strings are output as input candidates in descending order of the sum of the similarities with each character string included in the input history of the input target form. Therefore, the input candidates presented to the operator are not limited to the character strings included in the input history of the input target form, but the adverse effect that a large number of unrelated character strings are output as input candidates is also prevented. .

  The present proposal can be applied to a network system in which a client terminal as a computer operated by an operator and a server device are connected via a network. Also good. In the former case, processing such as acquisition of input history from the database, calculation of the sum of similarities, and selection of input candidates are also performed at the client terminal. In this case, the input form may be prepared in advance in the client terminal or may be downloaded from the server device. When downloading from the server device, it is desirable that the input history for the input form is also added to the database each time a new type of input form is downloaded. When the database exists on the server device side, it is possible to store the input history for each user as the input history for each input form, and to store the input history without distinguishing the input user. It is also possible.

  As a method of calculating the similarity, various known methods can be adopted, and as one of them, the inner product correlation method can be adopted. In this case, not only the morphological coincidence rate between character strings but also weights such as the frequency and ratio of each character string input to each input form may be reflected in the similarity. Also, if there is a tendency for similar character strings to be input between input forms, this is evaluated as similarity between forms, and the similarity of character strings input to input forms with high inter-form similarity You may rate higher.

  According to the present plan configured as described above, not only the history of character strings input in the past for the same input form, but also the history of character string input in input forms that tend to input similar character strings. Also, input candidates can be presented.

Hereinafter, an example of a network system according to an embodiment of the present plan will be described with reference to the drawings.
<System configuration>
First, the system configuration of the network system will be described based on the block diagrams of FIGS.

  As shown in FIG. 1, this network system is an electronic document that uses the above input support method (for example, membership application forms for various member services, product purchase application forms for online shopping, electronic application forms for various administrative institutions, etc. ) And a client device 2 that transmits an electronic document to the server device 1. In the following description, it is assumed that the network that relays the server device 1 and the client terminal 2 is the Internet N. In addition, a large number of such server devices 1 and client terminals 2 may exist on the Internet N, but one unit is illustrated in FIG.

The client device 2 is a personal computer having an Internet connection function, a PDA (Personal Digital Assistant), a mobile phone or the like, and is generally commercially available. That is, an OS and a browser program 21 including a communication program are installed in a memory or a disk device (not shown) of the client device 2, and a CPU (not shown) reads and executes these programs to thereby execute a network adapter or wireless communication (not shown). It is only necessary to be able to connect to the Internet N through a device (SIM card or the like). As a result, according to the browser program 21, the CPU sends various structured document data (HTML data, XML data, etc., hereinafter referred to as HTML data, which will be described below) from a web server (such as the server device 1) on the Internet. HTML data S, HTML data 0 of the input candidate presentation screen, etc.) are downloaded, and the downloaded HTML data is analyzed to display a screen corresponding to each HTML data on a display (not shown).

On the other hand, the server device 1 has a hardware configuration such as a CPU, a network adapter, a main memory, and a hard disk that are normally required as a web server on the Internet N, and an OS and WWW that are stored in the hard disk and read and executed by the CPU. (World
Wide Web) server program.

In addition, the hard disk (not shown) of the server device 1 is linked to the template of the input form read from the disk device (Web page DB 13 described later) in response to a request from the client terminal 2 in cooperation with the WWW server program. The electronic document HTML data S is generated after being generated and transmitted to the requesting client terminal 2, and each input form of the electronic document template HTML data is appropriately supplemented with a character string. Will be described later in the template of the input candidate presentation screen read out from the disk device (Web page DB 13 to be described later) according to the content (the content meaning the character string supplemented to each input form). By embedding the input candidate group notified from the similar candidate recommendation function 11 as described above, Various application programs (CGI) for causing a CPU (not shown) to execute processing for generating HTML data 0 of an input candidate presentation screen having a function of selecting and supplementing the input form and causing the request source client device 2 to respond Programs, servlets, etc.) are also installed. In FIG. 1, the functions realized by the CPU, OS, WWW server program, and the various application programs are illustrated as “Web server 10”.

Further, on the hard disk (not shown) of the server device 1, a process for determining an input candidate group for the requested input form and responding to the WWW server 10 in response to a request from the WWW server 10 is sent to the CPU (not shown). Various application programs (CGI program, servlet, etc.) to be executed are also installed. In FIG. 1, functions realized by the CPU and the various application programs are illustrated as “similar candidate recommendation function 11”. The function of the similar candidate recommendation function 11 will be described later in detail with reference to FIG.

Furthermore, the hard disk (not shown) of the server device 1 is supplemented with each input form in the electronic document HTML data S received from the client device 2 by cooperating with the WWW server 10. Various application programs (CGI program, servlet, etc.) for causing a CPU (not shown) to execute processing for storing a column as an input history in a disk device (an input history database 15a described later) are also installed. In FIG. 1, functions realized by the CPU and the various application programs are illustrated as “history update function 12”.

The web page database 13 is a database that stores in advance various web data that can be responded to by a WWW server program alone, the template of the input form described above, and the template of the leading candidate presentation screen described above. It is an element that constitutes a part. FIG. 3 is a table exemplifying a data structure of the Web page database 13 and a part storing one input form template and the input candidate presentation screen template.

  The input history database 15 is created for each input form included in the HTML data of various electronic documents stored in the Web page database 13, and is input (supplemented) to the corresponding input form in the past. This is a database that accumulates the history of input character strings (input history) (corresponding to database means for storing input history of character strings input in the past for each of a plurality of predetermined input forms). In FIG. 1, the input target form input history database 15a and the other input history database 15b are illustrated separately, but the difference between them is only relatively, and when processing for one input form is performed. In addition, the input history database 15 corresponding to the input form is merely referred to as “input target form input history database 15a”. Further, these input history databases 15 may be integrated as one database as long as the input history is accumulated and searchable for each input form. 4 and 5 are tables illustrating the data structure of the input history database 15 and its contents. As shown in FIGS. 4 and 5, the input history database 15 has a “form ID” field for recording the ID of the corresponding input form, and each character input (supplemented) in the past to the input home. An “input history” field for recording each column, an “number of input” field for recording the number of times of input for each character string, and a ratio of the number of times of input for each character string to the total number of times of input Consists of the "weight" field.

  FIG. 2 is a block diagram showing a detailed configuration of the similar candidate recommendation function 11 described above. As shown in FIG. 2, the similar candidate recommendation function 11 (that is, a group of application programs for realizing the similar candidate recommendation function 11) includes an input target form history acquisition function 111, an entire history acquisition function 112, and a similarity calculation function. 113 and the ranking function 114 (that is, each application program for realizing the function of each module).

  The input target form history acquisition function 111 reads the contents of the input history (hereinafter referred to as “input target form”) from the input history database (input target form input history database 15a) corresponding to the input form (input target form) designated from the Web server 10. (Referred to as acquisition means for acquiring the input history of the input target form from the database means).

  The entire history acquisition function 112 is a module that acquires the contents of each input history (hereinafter referred to as “non-input target form history”) from all input history databases 15b other than the input target form input history database 15a (input target). This corresponds to an acquisition unit that acquires from the database unit the input history of a non-input target form that is an input form other than a form.

  The similarity calculation function 113 is a module that calculates the similarity between each character string forming the input history of the input target form and each character string forming the input history of each non-input target form. Note that various methods can be applied as the similarity calculation method. Here, as an example, the inner product correlation method is used (input among the character strings included in the input history of the non-input target form). For items not included in the input history for the target form, this corresponds to a similarity total calculation means for calculating the sum of similarities for all character strings included in the input history for the input target form).

  The ranking function 114 selects several character strings having a high similarity calculated by the similarity calculation function 113 as recommended input candidates, and the input target form input history read by the input target form history acquisition unit 111. This is a module that responds to the Web server 10 as a candidate for input included in (a predetermined number of character strings having a high sum of similarities among character strings included in the input history for the non-input target form, This corresponds to input candidate output means for outputting as an input candidate together with a character string included in the input history of the input target form).

The inter-form similarity database 14 is a database for recording the inter-form similarity, which is a comprehensive evaluation of the similarity calculated by the similarity calculation function 113 between the input forms. Is an element that constitutes a part of FIG. 6 is a table illustrating the data structure of the inter-form similarity database 14 and its contents.
<Processing content>
Hereinafter, with reference to the flowchart, the contents of the operations of the Web server 10, the similar candidate recommendation function 11 and the history update function 12 in the server device 1 (that is, processing executed by a CPU (not shown) according to the WWW server program and each application program). Will be described with reference to the flowcharts of FIGS. 7 to 15 and the timing charts of FIGS. Note that the same step numbers are assigned to the same processes in the flowchart and the timing chart. In the following description, in order to facilitate understanding, specific examples of processing in the case where various data illustrated in FIGS. 3 to 6 are stored in the databases 13 to 15 will be described together.

  First, FIG. 7 shows processing executed by the Web server 10 and is activated for each session with each client terminal 2. In the first S001, the Web server 10 waits for an HTTP request message from the client terminal 1. When any HTTP request message is received, the Web server 10 advances the process from S001 to S002.

In S002, the Web server 10 checks whether the HTTP request message received from the client terminal 1 is an electronic document template HTML acquisition request (FIG. 18). If an electronic document template HTML acquisition request is received, the process proceeds to step S003. If any other HTTP request message is received, the process proceeds to step S007.

  In S <b> 003, the Web server 10 acquires an HTML template of an input form that should form a digitized document template requested by the received HTTP request message from the Web base database 13. For example, if the requested electronic document template is composed of only one input form stored in the Web page database 13 having the contents shown in FIG. 3, the input form extracted in this step The HTML template is as shown in FIG.

  In the next S004, the Web server 10 empties the input value on the HTML template acquired in S003. For example, when the HTML template having the content shown in FIG. 19 is acquired, the result is as shown in FIG.

  In the next S005, the Web server 10 generates the requested electronic document template HTML data by synthesizing the HTML template processed in S004. However, when the requested electronic document template is composed of only one input form, the HTML template processed in S004 is used as it is as the electronic document template HTML data.

  In next S006, the Web server 10 returns the digitized document template HTML data subjected to the processing in S005 to the requesting client terminal 2. Then, in the client terminal 2, the browser program 21 executed by the CPU (not shown) displays an electronic document template on a display (not shown) based on the received HTML data. For example, when the digitized document template HTML data is composed only of the HTML template shown in FIG. 20, the digitized document template to be displayed is as shown in FIG. As is apparent from FIGS. 20 and 21, in this electronic document template, a character string indicating an address is input by being operated by an input form 31 to be input (supplemented), an input device (not shown). A “history search” button 32 for transmitting a candidate recommendation request message to the server device 1 and a character string input (supplemented) in the input form 31 by being operated by an input device (not shown) are referred to as a “transmission request message”. "Send" button 33 for sending as "." As shown in FIG. 22, the input candidate recommendation request message includes the URL of “history search function” and the ID of the input form 31 corresponding to the operated “history search” button 32. Yes.

  When S006 is completed, the Web server 10 returns the process to S001 and waits for the next HTML message. When the HTML message is received from the client terminal 2, the Web server 10 advances the process from S001 to S002, and if the received HTML message is an HTML message other than an electronic document template HTML acquisition request, the process is performed. The process proceeds from S002 to S007, and further, it is checked whether or not the received HTML message is an input candidate recommendation request message. The Web server 10 advances the process to S008 if an input candidate recommendation request message is received, and advances the process to S011 if any other HTTP request message is received.

In S008, the Web server 10 calls the similar candidate recommendation function 11. Then, the called similar candidate recommendation function 11 executes the subroutine of FIG. 8 and calls the input target form history acquisition function 111 in the first S101. Then, the called similar candidate recommendation function 11 and the subroutine of FIG. 9 are executed, and in S201, the input history database 15 corresponding to the input form ID included in the input candidate recommendation request message received by the Web server 10, that is, From the input target form input history database 15a, all the character strings (and their “weight” field values [hereinafter simply referred to as “weights”) as input history are read. For example, when the input history database 15 for the form ID = 0 having the contents shown in FIG. 4 is the input target form input history database 15a, the input history to be read is as shown in FIG. The process of S201 corresponds to a process of acquiring the input history for the input target form from the database.

  In the next S102 in the subroutine of FIG. 8, the similar candidate recommendation function 11 calls the entire history acquisition function 112. Then, the called entire history acquisition function 112 executes the subroutine of FIG. 10, and in S301, all the input history databases 15 other than the input target form input history database 15a, that is, all the non-input target form input history databases 15b. From this, all input histories (character strings) and their weights are read out. For example, when two input history databases 15 for the form ID = 1 and the form ID = 2 having the contents shown in FIG. 5 exist separately from the input target form input history database 15a, The weights are as shown in FIG. The process of S301 corresponds to a process of acquiring an input history for a non-input target form that is an input form other than the input target form from the database.

  In the next S 103 in the subroutine of FIG. 8, the similarity candidate recommendation function 11 calls the similarity calculation function 113. Then, the called similarity calculation function 113 executes the subroutine of FIG.

  In the first step S401 after entering the subroutine of FIG. 11, the similarity calculation function 113 concatenates the input histories (character strings) read from each non-input target form input history database 15b in S301 as one data string. For example, when the input history (character string) having the contents shown in FIG. 24 is read from each non-input target form input history database 15b, the data string is as shown in FIG. As shown in FIG. 25, at this point, the data string may include the same character string in duplicate.

  In next step S402, the similarity calculation function 113 deletes a duplicate character string from the data string connected in step S401. For example, in the data string having the contents shown in FIG. 25, the character string “EE Tokyo EE Ward” is duplicated, so that the character string (appeared for the second time or later) is deleted as shown in FIG. Is done.

In the next S403, the similarity calculation function 113 deletes the character string included in the input history acquired in S201 from the data string that has undergone the process of S402. For example, assuming that the input history acquired in S201 is as shown in FIG. 23, the character string “Tokyo AA Ward” and the character string “Tokyo BB Ward” from the data string shown in FIG. Is deleted, and the data string after deletion is as shown in FIG.

In the next S404, the similarity calculation function 113 describes the “weight” stored for each character string included in the data string that has undergone the process of S403 for each non-input target form input history database 15b. A merge table is generated by adding a field (column). At this time, an invalid value N / A is set to indicate that there is no value in a record (row) corresponding to a character string for which no value is stored in each non-input target form input history database 15b. Since such an invalid value N / A is set, the value of each field (column) in the merge table is referred to as “expansion weight”. For example, if the data string has the contents shown in FIG. 27 and the input history read from each non-input target form input history database 15b is as shown in FIG. 24, the merge table created is as shown in FIG. Become. That is, the character strings “GG city in Chiba Prefecture”, “HH city in Kanagawa Prefecture”, and “FF city in Saitama Prefecture” are not stored in the input history database 15 corresponding to the form 1; In the field of “extension weight”, an invalid value N / A is set in a record having these as a heading. In addition, since the character strings “Tokyo CC Ward” and “Tokyo DD Ward” are not stored in the input history database 15 corresponding to Form 2, in the field of “Extension Weight in Form 2” in the merge table, An invalid value N / A is set in a record having these as a headline.

  In next step S405, the similarity calculation function 113 reads the inter-form similarity between the input target form and each other input form (non-input target form) from the inter-form similarity database 14. For example, the inter-form similarity for all the non-input target forms corresponding to the input target form is read from the inter-form similarity database 14 having the contents shown in FIG. The significance (calculation method) of the inter-form similarity registered in the inter-form similarity database 14 will be described in the description of S501 described later.

  In next step S406, the similarity calculation function 113 executes a “character string similarity weight value calculation process” subroutine shown in FIG. In the first step S501 after entering this subroutine, the similarity calculation function 113 extracts one unprocessed heading (character string) from the merge table generated in step S404.

  Subsequently, the similarity calculation function 113 executes a loop process from S502 to S505. In the first step S502 after entering this loop process, the similarity calculation function 113 extracts one unprocessed character string from the input history of the input target form acquired in S201.

  In the next step S503, the similarity calculation function 113 calculates the character string similarity using the character string extracted in S502 as a comparison source character string and the headline extracted in S501 as a comparison destination character string. Specifically, the ratio of the number of matching characters from the beginning of the comparison target character string and the comparison source character string to the number of characters of the comparison source character string is calculated as “character string similarity”.

  In the next S504, the similarity calculation function 113 multiplies the character string similarity calculated in S503 by the “weight” acquired in S201 corresponding to the character string extracted in S502, and calculates the product. To do. The process of S504 corresponds to a process of reflecting the weight calculated for each character string in the similarity value.

  In next step S505, the similarity calculation function 113 checks whether or not the processing has been completed for all the character strings included in the input history of the input target form acquired in step S201. If an unprocessed character string remains, the process returns to S502, and the process for the next character string is performed. On the other hand, if the unprocessed character string does not remain as a result of repeating the loop processing of S502 to S505, the similarity calculation function 113 advances the processing to S506.

In S506, the similarity calculation function 113 calculates the sum of the products calculated in S504 for all the character strings included in the input history of the input target form acquired in S201, and calculates the calculated sum. , “Weight value of character string similarity” of the headline (character string) extracted from the merge table in S501. For example, if the input history of the input target form is as shown in FIG. 23 and the contents of the merge table are as shown in FIG. 28, the “weight value of character string similarity” calculated for each heading (character string) "Is as shown in FIG. That is, for the “sum” for the heading “Tokyo CC Ward” in the merge table, the “character string similarity” with each input history “Tokyo AA Ward” and “Tokyo BB Ward” of the input target form respectively. Since it is calculated as “0.5” (
S503), the weight values “0.67” and “0.33” of the corresponding input target forms are multiplied (S504), and the sum is calculated (S506), and “0.50 × 0.67 + 0.50 × 0.33 = 0.50” is obtained. It is. Similarly, for other headings “Tokyo DD Ward”, “Tokyo EE Ward”, “Tokyo GG Ward”, “Tokyo HH Ward”, and “Tokyo FF Ward”, “weight of string similarity” Value is calculated as “0.50”, “0.50”, “0.00”, “0.00”, “0.00”, respectively. For the character string included in the input history for the input target form among the character strings included in the input history for the non-input target form, the processing in S506 is performed for all the character strings included in the input history for the input target form. This corresponds to the process of calculating the sum of the similarities with respect to.

  In next step S507, the similarity calculation function 113 checks whether or not the processing has been completed for all the headings (character strings) included in the merge table. If an unprocessed heading (character string) remains, the process returns to S501 to perform processing for the next heading (character string). On the other hand, if the unprocessed character string does not remain as a result of repeating the loop processing of S501 to S507, the similarity calculation function 113 ends this subroutine and proceeds to S407 in FIG.

  In S <b> 407, the similarity calculation function 113 executes the “inter-form similarity weight value calculation processing” subroutine shown in FIG. 13. In the first step S601 after entering this subroutine, the similarity calculation function 113 extracts one unprocessed heading (character string) from the merge table generated in step S404.

  Subsequently, the similarity calculation function 113 executes a loop process of S602 to S607. In the first step S602 after entering this loop process, the similarity calculation function 113 has not yet extracted out of the extension weights in each non-input target form corresponding to the heading (character string) extracted in S601 from the merge table. Take out one thing.

  In the next S603, the similarity calculation function 113 checks whether or not the extension weight extracted in S602 is an invalid value N / A. If the value is invalid value N / A, the similarity calculation function 113 sets “calculation result” to “0” in S604, and advances the process to S607.

  On the other hand, if it is determined in step S603 that the extension weight is not the invalid value N / A, the similarity calculation function 113, in step S605, searches the inter-form similarity database 14 for all the input target forms. Read the similarity between forms for non-input target forms.

  In the next S606, the sum of all the inter-form similarity extracted in S605 is calculated. Then, the inter-form similarity between the input target form and the non-input target form corresponding to the extended weight extracted in S602 is multiplied by the extended weight extracted in S602, and the calculated product is further calculated. Divide by the sum of degrees and finally calculate the quotient as “calculation result”. When S606 is completed, the similarity calculation function 113 advances the process to S607.

  In S <b> 607, the similarity calculation function 113 checks whether or not the processing in S <b> 603 and subsequent steps has been completed for the extended weights of all the non-input target forms corresponding to the heading (character string) extracted in S <b> 601. If the processing from S603 onward is not yet completed for the extended weights of all non-input target forms, the similarity calculation function 113 returns the processing to S602. On the other hand, when the processing from S603 onward is completed for the extended weights of all non-input target forms as a result of repeating the loop processing of S602 to S607, the similarity calculation function 113 starts the processing from S607. Proceed to S608.

In S608, the sum of the “calculation results” obtained in S604 or S606 is calculated for the extension weights of all the non-input target forms for the headline (character string) extracted in S601, and the calculated sum is calculated in S601. The “weight value of similarity between forms” for the headline (character string) taken out in (1). For example, assuming that the content of the inter-form similarity database 14 is as shown in FIG. 6 and the content of the merge table is as shown in FIG. 28, the “inter-form similarity degree calculated for each heading (character string)” is calculated. The “weight value” is as shown in FIG. That is, the sum of the similarity between forms of the input target form and each non-input target form in the inter-form similarity database 14 is “0.65 + 0.25 = 0.90”. The “extension weight in form 1” for the heading “Tokyo CC Ward” in the merge table is “0.40”, and the inter-form similarity between the input target form described in the inter-form similarity database 14 and form 1 Since the degree is “0.65”, the “calculation result” is “(0.65 × 0.40) /0.90=0.289” (S606). On the other hand, since “extension weight in form 2” for the heading “Tokyo CC Ward” in the merge table is an invalid value N / A, “calculation result” is “0” (S604).
). Therefore, the “weight value of similarity between forms” is calculated as “0.289”. Similarly, for the other headings “Tokyo DD ward”, “Tokyo EE ward”, “Tokyo GG ward”, “Tokyo HH ward”, and “Tokyo FF ward”, “weight of similarity between forms” Value ”is“ 0.144 ”,
Calculated as “0.128”, “0.097”, “0.056”, “0.028”. The processing of S604 to S608 corresponds to processing for reflecting the sum of values proportional to the inter-form similarity calculated for the non-input target form including each character string in the similarity value.

  In next step S609, the similarity calculation function 113 checks whether or not the processing in step S601 and subsequent steps has been completed for all headings (character strings) in the merge table. If the processing for all headings (character strings) has not been completed, the similarity calculation function 113 returns the processing to S601. On the other hand, if the processing from S602 onward has been completed for all headings (character strings) as a result of repeating the loop processing of S601 to S609, the similarity calculation function 113 ends this subroutine. The process proceeds to S408 in FIG.

  In S <b> 408, the similarity calculation function 113 calculates the “character string similarity weight value” calculated in S <b> 406 and the “inter-form similarity weight value” calculated in S <b> 407 for each heading (character string) in the merge table. Is calculated as “input candidate appropriate evaluation value”. For example, assuming that the “character string similarity weight value” and “form similarity weight value” calculated for each heading (character string) are as shown in FIGS. 29 and 30, respectively, each heading (character string) The “input candidate appropriate evaluation value” calculated for) is as shown in FIG. In other words, the “character string similarity weight value” calculated for the heading “Tokyo CC Ward” is “0.50”, and the “form similarity weight value” is “0.289”. Is calculated as “0.144”. Similarly, for the other headings “Tokyo DD ward”, “Tokyo EE ward”, “Tokyo GG ward”, “Tokyo HH ward”, and “Tokyo FF ward”, “input candidate appropriate evaluation value” Are calculated as “0.072”, “0.064”, “0.000”, “0.000”, and “0.000”, respectively.

  In the next S409, the “input candidate appropriate evaluation value” calculated for each heading (character string) in S408 is stored in a disk device (not shown) in the list format shown in FIG. 31, and a merge table is recommended for similar candidates. Pass to function 11. When S409 is completed, the similarity calculation function 113 ends the subroutine of FIG. 11 and returns the process to the similarity candidate recommendation function 11.

  In the next S104 in the subroutine of FIG. 8, the similar candidate recommendation function 11 calls the ranking function 114. Then, the called ranking function 114 executes the subroutine of FIG.

In the first step S701 after entering the subroutine of FIG. 14, the ranking function 114 determines the similarity in descending order of the “input candidate appropriate evaluation value” stored in the disk 301 (not shown) by the similarity calculation function 113 in S409. The records in the merge table passed from the calculation function 113 are sorted. For example, if the contents of the “input candidate appropriate evaluation value” are as shown in FIG. 31 and the contents of the merge table are as shown in FIG. 28, the merge table records after sorting are as shown in FIG. Become.

  In next step S702, the ranking function 114 extracts a predetermined upper number of records from the merge table after the sorting in step S701, and stores it in a disk device (not shown). For example, if the sorted merge table is as shown in FIG. 32 and the top three records are extracted, the extracted records are as shown in FIG.

  In next step S703, the ranking function 114 selects a recommended input candidate consisting of the input history of the input target form acquired in step S201 (that is, the input candidate) and the heading (character string) in the record extracted in step S702. The set is returned to the Web server 10. For example, if the contents of the input history of the input target form are as shown in FIG. 23 and the extracted records are as shown in FIG. 33, the input candidates and recommended input candidates that are responded to the Web server 10 are 34 (a) and 34 (b), respectively. The above-described processing of S701 to S703 is performed by using a predetermined number of character strings having a high sum of similarities among the character strings included in the input history for the non-input target form as characters included in the input history for the input target form. It corresponds to outputting as an input candidate together with a column.

  When S703 is completed, the ranking function 114 ends the subroutine of FIG. 14 and returns the process to the similar candidate recommendation function 11. Then, since the subroutine of FIG. 8 is also ended, the similar candidate recommendation function 11 returns the processing to the Web server 10.

  In the next S009 in the main routine of FIG. 7, the Web server 10 reads the HTML template of the input candidate presentation screen from the Web page database 13. For example, if the content of the Web page database 13 is as shown in FIG. 3, the content of the HTML template of the input candidate presentation screen to be read is as shown in FIG.

  In the next S010, the Web server 10 adds each input candidate and the character string of each recommended input candidate received from the similar candidate recommendation function 11 in S008 to the list in the HTML template of the input candidate presentation screen acquired in S009. By inserting, the HTML data of the input candidate presentation screen is generated. For example, when the input candidates and recommended input candidates having the contents shown in FIG. 34 are returned from the similar candidate recommendation function 11, as shown in FIG. 36, the URL of the input target form is equal to the number of input candidates and recommended input candidates. The description of the anchor tag including is copied, and each input candidate and recommended input candidate are inserted.

  In the next S006, the Web server 10 returns the HTML data O of the input candidate presentation screen that has undergone the processing of S010 to the requesting client terminal 2. Then, in the client terminal 2, the browser program 21 being executed by the CPU (not shown) displays an input candidate presentation screen on a display (not shown) based on the received HTML data O. For example, if the HTML data O of the input candidate presentation screen is as shown in FIG. 36, the displayed input candidate presentation screen is as shown in FIG. As is clear from FIGS. 36 and 37, each input candidate and recommended input candidate are displayed in a list on the input candidate presentation screen, and each input candidate and recommended input candidate is selected (for example, by a pointing device (not shown)). 38, an input candidate selection request message including the URL of the input target form and the selected input candidate or recommended input candidate as parameters is sent to the Web server 10, as shown in FIG.

When S006 is completed, the Web server 10 returns the process to S001 and waits for the next HTML message. When the HTML message is received from the client terminal 2, the Web server 10 advances the process from S001 to S002. If the received HTML message is other than the digitized template request message and the input candidate recommendation request message, The process proceeds to S007 and S011, and it is further checked whether or not the received HTML message is an input candidate selection request message. The Web server 10 advances the process to S012 if an input candidate selection request message has been received, and advances the process to S014 if any other HTTP request message has been received.

  In S012, the Web server 10 acquires the HTML template of the input form corresponding to the URL included as a parameter in the received HTTP request message from the Web base database 13.

  In the next S013, the Web server 10 inserts an input candidate or a recommended input candidate included as a parameter in the received HTTP request message into the input value of the HTML template acquired in S012, thereby generating HTML of the electronic document template. Complete the data. For example, if the HTML template of the input form is as shown in FIG. 19, as shown in FIG. 39, the selected input candidate or recommended input candidate is inserted into the anchor tag.

  In the next S006, the Web server 10 returns the digitized document template HTML data that has undergone the processing of S013 to the requesting client terminal 2. Then, in the client terminal 2, the browser program 21 executed by the CPU (not shown) displays the electronic document on the display (not shown) based on the received HTML data. For example, when the recommended input candidate “Tokyo EE Ward” is inserted in FIG. 39, the electronic document to be displayed is as shown in FIG. As is clear from FIG. 40, this electronic document template is the one shown in FIG. 21 except that the selected input candidate or recommended input candidate character string is input (supplemented) to the input form 31. Is exactly the same. Accordingly, the “history search” button 33 can be operated again, the character string input (supplemented) in the input form 31 can be corrected, and the “send” button 33 is operated. It is also possible. Therefore, when the “Send” button 33 is operated, as shown in FIG. 41, the ID of the input form and the character string (content of the input form) input (supplemented) in the text box 31 at that time are displayed. A transmission request message is sent to the Web server 10 as a parameter and addressed to the URL of the Web server 10.

  When S006 is completed, the Web server 10 returns the process to S001 and waits for the next HTML message. When the HTML message is received from the client terminal 2, the Web server 10 advances the process from S001 to S002, and the received HTML message is an electronic document template request message, input candidate recommendation request message, and input candidate selection request. If it is not a message, the process proceeds to S007, S011, and S014, and it is further checked whether or not the received HTML message is a transmission request message. Then, if the Web server 10 has received a transmission request message, the process proceeds to S015, and if another HTTP request message has been received, the process returns to S001.

  In S015, the Web server 10 performs processing for receiving a transmission request message. That is, a computerized document whose contents are specified by parameters included in the transmission request message is received and processing for coping with it is executed by itself, or another server device (not shown) for performing such coping Forward to. For example, if the electronic document is a certificate issuance application form, a certificate issuance procedure is performed.

In the next S015, the Web server 10 calls the history update function 12. Then, the called history update function 12 executes the subroutine of FIG. 15, and in the first S801, corresponds to the input target form input history database 15a (that is, the ID included as a parameter in the received transmission request message). Update input history database). That is, if the character string included as a parameter in the received transmission request message is not yet written in the “input history” field, add a new record, fill in the character string, and enter the “input count” field. Enter the initial value “1”. If the character string is already described in the “input history” field, the “number of times of input” corresponding to the character string is incremented by one. Then, the “weight” value is recalculated and overwritten in the “weight” column according to the ratio of the number of times of input of each character string. For example, if the contents of the input target form input history database 15a are as shown in FIG. 4, and the received transmission request message includes “Tokyo EE Ward”, the input target form input The history database 15a is updated as shown in FIG. The processing in S801 corresponds to the processing for storing the input history of character strings input in the past in the database, and the ratio of the number of times of input between character strings included in the input history for each input form. This corresponds to the process of calculating the weight in advance.

  Subsequently, the history update function 12 executes the first loop process of S802 to S813 for each input history database 15 other than the input target form input history database 15a (that is, the non-input target form input history database 15b).

  In the first loop process, the history update function 12 selects one of the non-input target form input history database 15b that has not been selected in the first S802.

  Subsequently, the history update function 12 executes the second loop processing of S803 to S810 for each input history (character string) in the non-input target form input history database 15b selected in S802.

  In the second loop process, the history update function 12 selects one of the input histories (character strings) in the non-input target form input history database 15 b selected in S 802 in the first S 803 that has not been extracted. Take out.

  Subsequently, the history update function 12 executes the third loop processing of S804 to S809 for each input history (character string) extracted in S803.

  In the third loop process, the history update function 12 selects one of the input histories (character strings) in the input target form input history database 15a after the update in S801 in the first S804. Take out.

  In next step S805, the history update function 12 stores the input history (character string) in the non-input target form input history database 15b extracted in S803 and the input history (character string) in the input target form input history database 15a. Check if any matches.

If the two match, the history update function 12 determines that the input history (character string) in the non-input target form input history database 15b and the input history (character string) in the input target form input history database 15a in S806. ) Is set to “1.0”, and then the process proceeds to S808. On the other hand, if the two do not match, the history update function 12 uses the input history (character string) comparison source character string in the input target form input history database 15a and the non-input target form input history in S807. Using the input history (character string) in the database 15b as a comparison target character string, the character string similarity is calculated as in S503. Specifically, the ratio of the number of matching characters from the beginning of the comparison target character string and the comparison source character string to the number of characters of the comparison source character string is set as the input history (character string) in the non-input target form input history database 15b. This is calculated as “character string similarity” with the input history (character string) in the input target form input history database 15a. When S807 is completed, the history update function 12 advances the process to S808. For example, the contents of the input target form input history database 15a after the update in S801 are as shown in FIG. 42, and the contents of “form 1” corresponding to the non-input target form input history database 15b having the contents shown in FIG. If the input target form is selected in S802, the “character string similarity” acquired in S806 or S807 for each combination of the input histories (character strings) is as shown in FIG.

  In S808, the history update function 12 records the input history (character string) extracted in S804 in the input target form input history database 15a in the “character string similarity” acquired in S806 or S807. The “weight” and the “weight” recorded in the non-input target form input history database 15b are multiplied by the input history (character string) extracted in S803. When the processing of S808 calculates the similarity between the forms, the similarity between each character string included in the input history for the input target form and each character string included in the input history for the non-input target form This corresponds to the process of reflecting the weight calculated for each character string each time.

  In next step S809, the history update function 12 checks whether or not the third loop processing has been executed for all input histories (character strings) in the input target form input history database 15a. If the third loop process has not been executed for all input histories (character strings), the history update function 12 returns the process to S804. On the other hand, when the third loop process has been executed for all input histories (character strings) as a result of repeating the third loop process, the history update function 12 advances the process from S809 to S810.

  In S810, the history update function 12 checks whether or not the second loop process has been executed for all input histories (character strings) in the non-input target form input history database 15b selected in S802. If the second loop process has not been executed for all input histories (character strings), the history update function 12 returns the process to S803. On the other hand, if the second loop processing has been executed for all input histories (character strings) as a result of repeating the second loop processing, the non-input target form input history database 15b selected in S802 Since the calculation in S808 is completed for all combinations of the input history (character string) and the input history (character string) in the input target form input history database 15a, the history update function 12 The process proceeds from S810 to S811.

  In S811, the history update function 12 selects all of the input history (character string) in the non-input target form input history database 15b selected in S802 and the input history (character string) in the input target form input history database 15a. The sum of the products calculated in S808 for the combination is calculated, and this is set as the similarity between the input target form and the non-input target form selected in S802. FIG. 44 shows the contents of the input target form input history database 15a after the update in S801 as shown in FIG. 42, and “form 1” corresponding to the non-input target form input history database 15b having the contents shown in FIG. This is an example of calculation in S811 when the non-input target form is selected in S802, and once the sum of products calculated in S808 is calculated for each input history (character string) of the input target form, Further, the sum total of the total sum calculated for each input history (character string) of the input target form is calculated. The processing in S811 is the sum of the similarities between each character string included in the input history for the input target form and each character string included in the input history for the non-input target form for each non-input target form. This corresponds to a process of calculating the similarity between forms in advance.

  In next step S812, the history update function 12 updates the similarity between the input target form in the inter-form similarity database 14 and the non-input target form selected in S802 with the value calculated in S811.

In next step S813, the history update function 12 checks whether or not the first loop processing has been executed for the all non-input target form input history database 15b. If the first loop process has not been executed for all non-input target form input history database 15b, the history update function 12 returns the process to S802. On the other hand, when the first loop process has been executed for all non-input target form input history database 15b as a result of repeating the first loop process, the history update function 12 ends the subroutine of FIG. Then, the process is returned to the Web server 10.

  In the main routine of FIG. 7, after completing S015, the Web server 10 returns the process to S001 and waits for the next message. As described above, the processes of S001 to S016 are repeated as long as the session with the same client terminal 2 is connected.

As described above, according to the present embodiment, the “history search” button corresponding to each input form 31 in the template of the electronic document (FIG. 21) provided by the Web server 10 on the server device 1. 32, when the user operates the “history search” button 32 using a pointing device (not shown) of the client terminal 2, the input history database 15 corresponding to the input form (input target form input history database 15a) ) In the past, the input history (character string) stored in the input history database 15 (non-input target form input history database 15b) for other input forms is read out. Total with each input history (character string) stored in the input target form input history database 15a Specific input history of similarity is a predetermined number in order from the highest ones (string) is selected. In this way, the input history (character string) read from the input target form input history database 15a is selected as an input candidate, or selected from the input history (character string) stored in the non-input target form input history database 15b. Some of the input histories (character strings) are displayed as recommended input candidates on the input candidate presentation screen (FIG. 37), respectively. Therefore, since the (recommended) input candidates other than the past input history (character string) of the input form are also presented to the user, the range of selection by the user is expanded. In addition, since the recommended input candidates other than the past input history (character string) of the input form are carefully selected for similarity with the past input history (character string) of the input form, On the other hand, there is no problem that the input is bothered.
(Appendix 1)
An input candidate recommendation method for recommending input candidates for an input form executed by a computer connected to a client terminal via a network,
For each of a plurality of predetermined input forms, the input history of character strings input in the past is saved in the database,
An input history for an input target form that is an input form selected as an input target by an operator of the client terminal, and an input history for a non-input target form that is an input form other than the input target form, respectively, From the database,
Of the character strings included in the input history for the non-input target form, those that are not included in the input history for the input target form, respectively, the similarity to all the character strings included in the input history for the input target form Of the sum of
Among character strings included in the input history for the non-input target form, a predetermined number of character strings having a high sum of the similarities are output as input candidates together with the character strings included in the input history for the input target form. An input candidate recommendation method, wherein the computer executes the operation.
(Appendix 2)
The input candidate recommendation method according to claim 1, wherein the computer executes that the similarity is calculated according to an inner product correlation method.
(Appendix 3)
In each input form unit, a weight corresponding to the ratio of the number of times of input between each character string included in the input history is calculated in advance,
When calculating the similarity between the character string included in the input history for the input target form and the character string included in the input history for the non-input target form, the weight calculated for each character string is The input candidate recommendation method according to appendix 1, wherein the computer executes the reflection in the similarity value.
(Appendix 4)
When there are a plurality of non-input target forms, for each non-input target form, each character string included in the input history for the input target form and each character string included in the input history for the non-input target form The inter-form similarity, which is the sum of the similarities with the above, is calculated in advance, and for each character string, the sum of values proportional to the inter-form similarity calculated for the non-input target form including the same is calculated. Reflect in the degree value.
The input candidate recommendation method according to appendix 1 or 3, wherein the computer executes the above.
(Appendix 5)
In each input form unit, a weight corresponding to the ratio of the number of times of input between each character string included in the input history is calculated in advance,
When calculating the inter-form similarity, each character is used for the similarity between each character string included in the input history for the input target form and each character string included in the input history for the non-input target form. The input candidate recommendation method according to appendix 4, wherein the computer executes that the weight calculated for the column is reflected and the sum of the similarities reflecting the weight is used as the inter-form similarity.
(Appendix 6)
In a network system including a client terminal and a server device, the computer stores the database in the server device and outputs the input device from the server device to the client terminal. The input candidate recommendation method according to appendix 1.
(Appendix 7)
A computer connected to a client terminal via a network is an input candidate recommendation program that is read and executed in order to recommend input candidates for an input form, the computer comprising:
Database means for storing an input history of a character string input in the past for each of a plurality of predetermined input forms;
An input history for an input target form that is an input form selected as an input target by the operator and an input history for a non-input target form that is an input form other than the input target form are respectively acquired from the database means. Acquisition means,
Of the character strings included in the input history for the non-input target form, those that are not included in the input history for the input target form, respectively, the similarity to all the character strings included in the input history for the input target form Similarity operation calculation means for calculating the sum of, and
Among character strings included in the input history for the non-input target form, a predetermined number of character strings having a high sum of the similarities are output as input candidates together with the character strings included in the input history for the input target form. An input candidate recommendation program that functions as an input candidate output means.

The block diagram which shows schematic structure of the network system which is embodiment of this plan Block diagram showing detailed configuration of similar candidate recommendation function Table showing example data of Web page database Table showing example data of input history database (input target form input history database) Table showing data example of input history database (non-input target form input history database) Table showing data example of inter-form similarity database A flowchart showing processing by the Web server 10 The flowchart which shows the process by the similar candidate recommendation function performed in S008 The flowchart which shows the process by the input object form log | history acquisition function performed in S101 The flowchart which shows the process by the whole log | history acquisition function performed in S102 The flowchart which shows the process by the similarity calculation function performed in S103 The flowchart which shows the weight value calculation processing subroutine of the character string similarity performed in S406 The flowchart which shows the weight value calculation processing subroutine of the similarity between forms performed in S407 The flowchart which shows the process by the ranking function performed in S104 The flowchart which shows the process by the similarity calculation function performed in S015 Timing chart showing the flow of data exchanged between each unit by the processing according to the flowcharts of FIGS. FIG. 16 is a partially enlarged timing chart Diagram showing parameters in HTML request message of electronic document template Figure showing input form HTML template Figure showing HTML template of input form with empty input value Diagram showing digitized document template The figure which shows the parameter in the input candidate recommendation request message Table showing input history read from the input target form input history database Table showing input history read from each non-input target form input history database A table showing data columns that combine the input history of each non-input target form Table showing the data column from which the input history of the input target form has been deleted Table showing data columns from which duplicate input history has been deleted Table showing merge table Table showing string similarity weight values Table showing weight values for similarity between forms Table showing input candidate suitability evaluation values Table showing merge table headings after sorting Table showing the top three extracted headings The table | surface which shows the input candidate and recommendation input candidate which are responded to a web server. Table showing HTML template of input candidate presentation screen Table showing HTML data of input candidate presentation screen Table showing input candidate presentation screen The figure which shows the parameter in the input candidate selection request message Table showing the HTML data of the digitized document template with the selected (recommended) input candidates inserted Diagram showing the digitized document template with the selected (recommended) input candidate inserted Diagram showing parameters in the send request message Table showing input target form input history database after update A table listing calculation results of character string similarity between each character string in the input history of the input target form and each character string in the input history of the non-input target form. Table showing the calculation process of similarity between forms Table showing the inter-form similarity database after update

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Server apparatus 10 Web server 11 Similarity candidate recommendation function 12 History update function 14 Inter-form similarity database 15 Input history database 111 Input object form history acquisition function 112 Whole history acquisition function 113 Similarity calculation function 114 Ranking function N network

Claims (5)

An input candidate recommendation method for recommending input candidates for an input form executed by a computer connected to a client terminal via a network,
For each of a plurality of predetermined input forms, the input history of character strings input in the past is saved in the database,
An input history for an input target form that is an input form selected as an input target by an operator of the client terminal, and an input history for a non-input target form that is an input form other than the input target form, respectively, From the database,
Of the character strings included in the input history for the non-input target form, those that are not included in the input history for the input target form, respectively, the similarity to all the character strings included in the input history for the input target form Of the sum of
Among character strings included in the input history for the non-input target form, a predetermined number of character strings having a high sum of the similarities are output as input candidates together with the character strings included in the input history for the input target form. An input candidate recommendation method, wherein the computer executes the operation. In each input form unit, a weight corresponding to the ratio of the number of times of input between each character string included in the input history is calculated in advance,
When calculating the similarity between the character string included in the input history for the input target form and the character string included in the input history for the non-input target form, the weight calculated for each character string is The input candidate recommendation method according to claim 1, wherein the computer executes the reflection to the similarity value. When there are a plurality of non-input target forms, for each non-input target form, each character string included in the input history for the input target form and each character string included in the input history for the non-input target form The inter-form similarity, which is the sum of the similarities with the above, is calculated in advance, and for each character string, the sum of values proportional to the inter-form similarity calculated for the non-input target form including the same is calculated. The input candidate recommendation method according to claim 1, wherein the computer executes the reflection to the degree value. In each input form unit, a weight corresponding to the ratio of the number of times of input between each character string included in the input history is calculated in advance,
When calculating the inter-form similarity, each character is used for the similarity between each character string included in the input history for the input target form and each character string included in the input history for the non-input target form. 4. The input candidate recommendation method according to claim 3, wherein the computer executes that the weight calculated for the column is reflected and the sum of the similarities reflecting the weight is set as the inter-form similarity. A computer connected to a client terminal via a network is an input candidate recommendation program that is read and executed in order to recommend input candidates for an input form, the computer comprising:
Database means for storing an input history of a character string input in the past for each of a plurality of predetermined input forms;
An input history for an input target form that is an input form selected as an input target by the operator and an input history for a non-input target form that is an input form other than the input target form are respectively acquired from the database means. Acquisition means,
Of the character strings included in the input history for the non-input target form, those that are not included in the input history for the input target form, respectively, the similarity to all the character strings included in the input history for the input target form A similarity sum calculating means for calculating the sum of
Among character strings included in the input history for the non-input target form, a predetermined number of character strings having a high sum of the similarities are output as input candidates together with the character strings included in the input history for the input target form. An input candidate recommendation program that functions as an input candidate output means.

Images