JP6193799B2 - Web element matching device, matching probability distribution generation device, method and program thereof - Google Patents

Description

  The present invention relates to a web element matching device that distributes a plurality of web elements in a web page to a plurality of devices, a matching probability distribution generation device, and a method and program thereof.

  In recent years, mobile devices with performance similar to PCs have exploded, and services linked between terminals have attracted attention. At the same time, the W3C standard web technology called HTML5 has evolved, making it possible to use multifunctional web pages from various OSs and devices.

  Against this background, the browsing experience of web pages is also diversifying. A technique for decomposing a web page in units of components and allowing them to be experienced on a plurality of devices according to the characteristics of the device is being studied.

  For example, in Non-Patent Document 1, an intermediate proxy server is interposed between a server and a client, a web page is divided into constituent elements (hereinafter simply referred to as “web elements”), and a plurality of connected to the intermediate server This makes it possible to match web elements to other devices. Non-Patent Document 2 discloses a method for optimally designing web elements such as music, moving images, and advertising content.

Yasuhiro Nakamo et al., “Proposal of Multi-Device Linkage Technology Using Intermediate Browser” Nippon Telegraph and Telephone Corporation, NTT Service Evolution Laboratories, 87th Groupware and Network Service Research Society, Information Processing Society of Japan, Volume: 2013-GN- 87, No .: 6, p1-p6, 2013-03-11. Hiroshi Hirose, “Improvement of advertising creative effect by optimization algorithm and crowdsourcing platform” IEICE technical report. Artificial intelligence and knowledge-based processing 110 (212), 13-18, 2010-09-21.

  For example, assume that there is a web page composed of three elements: video content, advertising content, and remote control function. Among these elements, there is a possibility that a new browsing style can be provided by distributing video content to “TV” and advertising content and remote control functions to “smartphones”. Thus, if a plurality of web elements can be provided to a plurality of devices in an optimum combination, there is a possibility that new value can be provided, and a technique for combining the web elements with the optimum device is required.

  However, in the combination method disclosed in Non-Patent Document 1, only one device type to be combined with each web element is specified by the developer. In addition, the web element design method disclosed in Non-Patent Document 2 does not discuss filtering a huge number of combinations, and does not consider prior knowledge. Thus, the technique which provides a web element with the optimal combination for several apparatuses is not yet disclosed.

For example, if there are three web elements and three matching devices, the number of matching combinations is 512 (2 ^{3 × 3} ). In such a case, it is difficult for the developer to determine an optimal matching combination. On the other hand, enormous amounts of operation data are required to obtain an optimal solution among 512 patterns based on user operation data.

  This invention is made in view of such a subject, and it aims at providing the web element matching apparatus and matching probability distribution generation apparatus which match a web element with the optimal apparatus, those methods, and a program. And

  The web element matching device of the present invention includes a combination information accumulation unit and an optimum combination candidate determination unit. The combination information accumulating unit stores a matching probability distribution which is a probability distribution representing a probability of a combination of a web element obtained by dividing a web page into component elements and a device set. The optimal combination candidate determination unit receives a web element and a matching probability distribution as input, and determines a matching candidate represented by a set of a web element and a device set by maximum likelihood estimation based on the matching probability distribution.

  In addition, the matching probability distribution generation device of the present invention includes a web element function estimation unit, a combination information accumulation unit, and a matching probability distribution generation unit. The web element function estimation unit estimates a function of the web element by using as input a web element obtained by dividing the web page into constituent elements. The combination information accumulation unit stores a web element matching table represented by a probability distribution of a relationship between a function of a web element across a plurality of web pages and a device set. The matching probability distribution generation unit generates a matching probability distribution by using the probability distribution of the relationship between the function of the web element and the device set with reference to the web element matching table with the function of the web element as an input.

  In addition, the web element matching method of the present invention includes an optimum combination candidate determination process. The optimal combination candidate determination process is a probability distribution representing a probability of a combination of a web element and a device set stored in the combination information accumulation unit, with a web element obtained by dividing a web page into component elements and a matching probability distribution as inputs. Matching candidates that can be represented by a set of a web element and a device set are determined by maximum likelihood estimation based on a certain matching probability distribution.

  The matching probability distribution generation method of the present invention includes a web element function estimation process and a matching probability distribution generation process. In the web element function estimation process, a web element obtained by dividing a web page into component elements is used as an input to estimate the function of the web element. In the matching probability distribution generation process, the web element matching is represented by the probability distribution of the relationship between the function of the web element across a plurality of web pages stored in the combination information accumulating unit and the device set, using the function of the web element as an input. A matching probability distribution is generated using the probability distribution of the relationship between the function of the web element and the device set with reference to the table.

  The program according to the present invention causes a computer to function the function components of each of the above devices.

  According to the web element matching apparatus, the method, and the program of the present invention, the matching candidate between the web element and the device set is determined by the maximum likelihood estimation based on the matching probability distribution, so that the web element is matched with the optimum device set. be able to. For example, it is possible to automatically match video content to a TV or the like and a remote control function to a smartphone or the like.

  Further, according to the matching probability distribution generation device, method and program of the present invention, it is possible to generate the matching probability distribution used in the web element matching device of the present invention.

The figure which shows the function structural example of the web element matching apparatus 100 of Embodiment 1 of this invention. 1 is a diagram showing a system configuration using a web element matching apparatus 100. FIG. The figure which shows an example of a matching probability distribution table. The figure which shows the function structural example of the matching probability distribution generation apparatus 200 of Embodiment 1 of this invention. The figure which shows an example of the web element matching table 221. FIG. The figure which shows the matching probability distribution table of the 101st page produced | generated by substituting the probability distribution of the web element matching table 221. The figure which shows the function structural example of the web element matching apparatus 300 of Embodiment 2 of this invention. The figure which shows the operation | movement flow of the web element matching apparatus 300. The figure which shows the function structural example of the web element matching apparatus 400 of Embodiment 3 of this invention. The figure which shows the operation | movement flow of the web element matching apparatus 400. The figure which shows an example of the matching probability distribution table which described the evaluation value together.

Embodiments of the present invention will be described below with reference to the drawings. The same reference numerals are given to the same components in a plurality of drawings, and the description will not be repeated.

Embodiment 1
FIG. 1 shows a functional configuration example of the web element matching apparatus 100 according to the first embodiment. The web element matching apparatus 100 includes a combination information accumulation unit 110 and an optimum combination candidate determination unit 120. The web element matching apparatus 100 is realized by a predetermined program being read into a computer including, for example, a ROM, a RAM, and a CPU, and the CPU executing the program. The same applies to other devices described below.

  Before describing the web element matching apparatus 100 according to the first embodiment, the configuration of the entire system in which the apparatus is used will be described with reference to FIG. The web element matching apparatus 100 functions as an intermediate server disposed between the web server 900 and the device group 800. The web element matching apparatus 100 disassembles the web page 910 of the web server 900 into the web elements 901 to 903 and then distributes the web elements 901 to 903 to the client terminals constituting the device group 800. For example, the video content 901 is distributed on the television 801, the remote control function 902 is distributed on a PDA (Personal Digital Assistants) 802, and the audio content 903 is distributed on a smartphone 803. Here, the combination of web elements distributed over a plurality of client terminals is referred to as matching. Hereinafter, the device group 800 is also referred to as a device set.

The combination information accumulating unit 110 of the web element matching apparatus 100 stores a matching probability distribution that is a probability distribution representing a probability of a combination of a web element c _{i obtained} by dividing a web page into component units and a device set D _* . The device set is denoted as D _* . For example, if the above-described client terminal types are two types: television (d ₁ ) and smartphone (d ₂ ), the device set is D ₁ = {d ₁ , d ₂ }, D ₂ = {d ₁ }, D ₄ = {d ₂ } and D ₄ = {}. D ₄ = {} represents a null where no device exists. In this way, the device set includes client terminals such as televisions and smartphones. There are 2 ^m types of device sets (D ₁ ,..., D _2m ), where m is the type of device.

Let P _i (X _i ) be the probability distribution that combines the web elements c _i into the device set. Since the device set is finite, X _i is a discrete random variable. The probability that the web element c _i and the device set D _k are combined is represented as a matching probability distribution P _i (X _i = D _k ). In the matching probability distribution, a relation of Σ _k P _i (X _i = D _k ) = 1 is established.

The optimal combination candidate determination unit 120 receives a web element and a matching probability distribution as input, and determines a matching candidate represented by a set of a web element and a device set by maximum likelihood estimation based on the matching probability distribution. Optimal combination candidate determination section 120 determines a likely candidate from all search candidates from equation (1). Since the matching probability distribution P _i (X _i = D _k ) is given in advance, it may be referred to as a prior probability distribution.

  FIG. 3 shows an example of the matching probability distribution table stored in the combination information accumulating unit 110. The matching probability distribution table stores a matching probability distribution of web elements for each web page.

  The first column of the matching probability distribution table shows web elements of a certain web page A. The second to fifth columns show the matching probability distribution of each device set. The sixth column shows the function of the web element (the function will be described in detail later). The seventh column shows the optimal set of devices that combine the web elements. This matching probability distribution table is prepared for each web page.

For example, when three web elements c ₁ , c ₂ , c ₃ and matching probability distributions p ₁ , p ₂ , p ₃ are input, the web element matching device 100 selects matching candidates represented by, for example, the formula (2). Output. After outputting the combination shown in Expression (2), the process of evaluating Expression (1) in combinations other than the output combination may be repeated a plurality of times (L times). If L = 3, in this example, the top three combinations can be determined from a total of 64 combinations. For example, the combinations shown in Expression (3) and Expression (4) are matching candidates output at the second time and the third time.

The matching candidate of the formula (2) is that the web element c ₁ (Video content: f ₁ ) is the television D ₂ = {d ₁ }, and the web element c ₂ (remote control function: f ₃ ) is the smartphone D ₃ = {d ₂ }. , Web element c ₃ (advertising content: f ₆ ) means to be combined with both the television and the smartphone D ₁ = {d ₁ , d ₂ }. This combination is determined by the optimal combination candidate determination unit 120 when the matching probability distribution table is illustrated in FIG.

  As described above, according to the web element matching apparatus 100 of the first embodiment, it is possible to determine the optimal web element for the device set by the maximum likelihood estimation based on the matching probability distribution. The input web element, the matching probability distribution, and the matching candidate (optimum device set) estimated by maximum likelihood estimation are added to the matching probability distribution table stored in the combination information accumulating unit 110. Thus, the web element and the matching probability distribution are stored as prior knowledge.

  Next, the matching probability distribution generation device 200 according to the first embodiment that generates the matching probability distribution will be described.

[Matching probability distribution generator]
FIG. 4 shows a functional configuration example of the matching probability distribution generation device 200. The matching probability distribution generation device 200 includes a web element function estimation unit 210, a combination information accumulation unit 220, and a matching probability distribution generation unit 230.

  The web element function estimation unit 210 receives a web element obtained by dividing a web page into constituent elements, and estimates the function of the web element. The technique for estimating the function of the web element uses a conventional technique. The conventional method will be briefly described.

The estimation of the function of the web element is performed by a rule-based determination using an html tag or a method using machine learning. For example, by using a rule base, it is possible to determine whether each is a moving image content, an audio content, or a menu using <video>, <audio>, and <navigation> tags. Furthermore, by supervised machine learning with features such as words in the web element, html tag information, and structure, estimation of whether the article is the main article or advertisement content is executed. The estimated web element functions include, for example, f ₁ : Video content, f ₂ : Audio content, f ₃ : Remote control function, f ₄ : Menu function, f ₅ : Main news article, f ₆ : Advertising content, etc. It is.

  The matching probability distribution generation unit 230 uses the web element function estimated by the web element function estimation unit 210 as an input, refers to the web element matching table 221, and uses the probability distribution of the relationship between the function of the web element and the device set. To generate a matching probability distribution.

  The web element matching table 221 is a table represented by a probability distribution of a relationship between a function of a web element across a plurality of web pages and a device set, and is stored in the combination information accumulation unit 220. Here, using the probability distribution refers to, for example, a web element matching table 221 indicating a probability distribution of a relationship between a function of a web element and a device set, and a combined probability value of the function of the web element and the device set. Is a matching probability distribution.

FIG. 5 shows an example of the web element matching table 221. The first column of the web element matching table 221 represents the function of the web element. The second to fifth columns represent the matching probability distribution of each device set. The first line indicates that, for example, the optimal device set of f ₁ : Video content is “TV” of device set D ₂ = {d ₁ } having a probability value of 0.6. The second row indicates that, for example, the optimal device set of f ₂ : Audio content is “TV” of device set D ₂ = {d ₁ } with a probability value of 0.5. The third line indicates that, for example, f ₃ : the optimum device set of the remote control function is a “smart phone” of device set D ₃ = {d ₂ } with a probability value of 0.7.

  For example, it is assumed that the web element matching table 221 shown in FIG. 5 shows a probability distribution of a relationship between a function of a web element and a device set across 100 web pages.

From the web element matching table 221, the optimal device set of web elements having the function of f ₁ : Video content is 15 times when D ₁ = {d ₁ , d ₂ }, and D ₂ = {d ₁ } Is 60 times, D ₃ = {d ₂ } is 25 times, and D ₄ = {} is 0 times, and the probability distribution of the optimal device candidate for the function is (0.15, 0.60). , 0.25, 0.0).

Assuming that the function of the web element c _{1 on} the 101st page is f ₁ , the matching probability distribution between the web element (c ₁ ) on the 101st page and the function (f ₁ ) is the probability distribution of the web element matching table 221. based on the high equipment set D ₂ the most probable value is the optimum equipment set. FIG. 6 shows a matching probability distribution table for the 101st page generated by substituting the probability distribution of the web element matching table 221.

  Thus, according to the matching probability distribution generation device 200 of the first embodiment, it is possible to generate a matching probability distribution that is a probability distribution representing the probability of a combination of a web element of a web page and a device set.

[Embodiment 2]
In FIG. 7, the example of a function structure of the web element matching apparatus 300 of Embodiment 2 is shown. The web element matching apparatus 300 further includes a web element function estimation unit 210, a matching probability distribution generation unit 230, a combination information accumulation unit 310, and a matching probability distribution presence / absence determination with respect to the web element matching device 100 (FIG. 1). It differs in the point provided with the part 320. FIG.

  The optimum combination candidate determination unit 120 is the same as the web element matching apparatus 100 as is apparent from the reference numerals. The combination information accumulating unit 310 also stores a web element matching table 221 represented by a probability distribution of a relationship between a web element function and a device set across a plurality of web pages. Different from 1). Since the configuration including the combination information accumulating unit 310, the web element function estimating unit 210, and the matching probability distribution generating unit 230 is substantially the same as the matching probability distribution generating device 200, its reference numeral is denoted as 200 '. Yes.

  The matching probability distribution presence / absence determination unit 320 is a functional configuration unit that determines whether or not a matching probability distribution is given in advance as a prior probability distribution to a web element. This matching probability distribution presence / absence determination unit 320 is not necessary when it is clear whether or not a matching probability distribution is set in advance for a web element. Therefore, the structure is shown by a broken line in FIG. The case where the matching probability distribution presence / absence determining unit 320 is necessary will be described later.

  First, the operation of the web element matching apparatus 300 will be described with reference to FIG. 8 on the assumption that the matching probability distribution is not set in advance for the web element. When the matching probability distribution is not set in advance for the web element (No in step S320), the web element function estimation unit 210 estimates the function of the web element by using the web element obtained by dividing the web page into the constituent elements. (Step S210). The matching probability distribution generation unit 230 receives the function of the web element estimated by the web element function estimation unit 210 as an input, and generates a matching probability distribution using the probability distribution of the relationship between the function of the web element and the device set ( Step S230).

  Then, the optimum combination candidate determination unit 120 determines a matching candidate represented by a set of a web element and a device set by maximum likelihood estimation based on the matching probability distribution generated by the matching probability distribution generation unit 230. The generated matching probability distribution is added to the matching probability distribution table stored in the combination information accumulating unit 310.

  Next, the operation when the matching probability distribution is set in advance for the web element will be described. When only the web element for which the matching probability distribution is set in advance is input (Yes in step S320), the process in the web element function estimation unit 210 in step S210 and the process in the matching probability distribution generation unit 230 in step S230 are: Not done. In this case, the optimal combination candidate determination unit 120 determines a matching candidate by maximum likelihood estimation based on the input matching probability distribution (step S120). The input web element, the matching probability distribution, and the matching candidate (optimum device set) estimated by maximum likelihood estimation are added to the matching probability distribution table stored in the combination information accumulating unit 310. That is, the web element matching apparatus 300 in this case performs the same processing as the web element matching apparatus 100 (FIG. 1).

  Next, a case where the matching probability distribution presence / absence determination unit 320 is necessary will be described. The matching probability distribution presence / absence determination unit 320 is a functional configuration unit necessary when it is unknown whether a matching probability distribution is set in advance for a web element.

  The matching probability distribution presence / absence determining unit 320 determines whether or not a matching probability distribution has been set for the web element every time a web element is input (step S320). When the matching probability distribution is not set for the web element (No in step S320), the web element matching apparatus 300 performs the process including steps S210 and S230 described on the assumption that the matching probability distribution is not set. . Therefore, even when a web element of an unknown web page for which no matching probability distribution is set for the web element is input, it is possible to match an optimal device set to the web element.

  If the matching probability distribution is set for the web element (Yes in step S320), the web element matching apparatus 300 determines the optimal web element for the device set by maximum likelihood estimation based on the input matching probability distribution ( Step S120). The web element matching apparatus 300 in this case performs the same processing as the above-described web element matching apparatus 100 (FIG. 1).

  As described above, according to the web element matching apparatus 300, it is possible to match a web element to an optimum device set regardless of whether or not a matching probability distribution is preset.

[Embodiment 3]
In FIG. 9, the example of a function structure of the web element matching apparatus 400 of Embodiment 3 is shown. The web element matching apparatus 400 can perform matching in consideration of the user's viewpoint.

  The web element matching device 400 has a functional configuration of the web element matching devices 100 (FIG. 1) and 300 (FIG. 7), a web element / device combination unit 410, a client terminal matching unit 420, and a browsing state measurement. Unit 430 and a matching evaluation unit 440.

The operation flow of the web element matching apparatus 400 shown in FIG. 10 will also be described. The web element / device combination unit 410 receives a plurality (L) of matching candidates output from the web element matching devices 100 and 300 as input, and has the highest probability value or evaluation value of the matching probability distribution of the web candidate of the matching candidate. The combination information that sets the device set as the optimum device set is output. In the initial state where the web element matching apparatus 400 starts operation, there may be a web element having no evaluation value. In that case, the web element / device combination unit 410 outputs combination information based on the probability value of the matching probability distribution (step S410).
Here, the plurality (L) of matching candidates are expressed by, for example, the above formulas (2) to (4). Matching candidates output for the third time among the matching candidates (c ₁ , D ₂ ), (c ₁ , D ₂ ), and (c ₁ , D ₃ ) of the web element c _{1 in} the expressions (2) to (4) Assuming that the probability value or evaluation value of the matching probability distribution of (c ₁ , D ₃ ) is the highest, the web element / device combination unit 410 outputs (c ₁ , D ₃ ) as combination information.

The client terminal matching unit 420 matches the web element with the client terminal included in the device group 800 based on the combination information output by the web element / terminal combination unit 410 (step S420). For example, when the combination information is (c ₁ , D ₃ ), the web element c ₁ is matched with d ₂ : a smartphone that forms the device set D ₃ = {d ₂ }.

Is one example smartphone client terminal 803 performs the rendering of the web elements c ₁ received. Note that the device group 800 (in this example, the smartphone 803) periodically transmits user operation information to the web element matching device.

View state measuring unit 430 detects the viewing stay time and the user operation of the web page of the user and outputs the viewing status in the client terminal the web element c ₁ are matched (step S430). When the browsing stay time exceeds a predetermined time, it can be estimated that the user is satisfied to some extent with the combination of the web element and the client terminal. Further, when the user leaves the web page by a user operation before the predetermined time, it can be estimated that the user is unsatisfied with the combination of the web element and the client terminal. In addition, the user satisfaction can be estimated based on the change log data of the page configuration by the user operation. Thus, it is possible to estimate the user's satisfaction based on the browsing stay time and the user operation.

  The browsing status measuring unit 430 outputs, for example, an evaluation value that can be used in the search algorithm as the browsing status. In addition, you may make it output browsing stay time, a user's operation log, etc. as it is as a browsing condition.

  The matching evaluation unit 440 updates the evaluation value written together in the matching probability distribution using a search algorithm based on the browsing situation output by the browsing situation measuring unit 430 (step S440).

FIG. 11 shows an example of a matching probability distribution table in which evaluation values are written together with the matching probability distribution of the matching probability distribution table (FIG. 3) described in the first embodiment. From the sixth column to the ninth column of the matching probability distribution table in FIG. 11, the evaluation values from the device set D _{1 to} the evaluation value of the device set D ₄ are added in order. The evaluation value is a value normalized so that the sum of all evaluation values becomes 1, for example.

  Examples of search algorithms that can be used in this embodiment will be briefly introduced. An A / B test can be used as a simple search algorithm. The A / B test presents the combination A and the combination B to the user at random, and determines success or failure from the number of clicks. This search algorithm is also called a split run, and is a technique that employs a combination with a high success rate after a certain time.

  In addition, a bandit algorithm can be adopted as an algorithm that guarantees the user's satisfaction level to some extent during the search. Examples of the banded algorithm include a softmax method and a UCB (Upper Confidence Bound) method. Both are formulated in consideration of the correct answer data in consideration of the expected value and the number of trials for each trial. Any search algorithm may be used in this embodiment.

  As the correct answer data, for example, a predetermined time ε of the browsing stay time is used. If the user's browsing stay time exceeds a predetermined time ε, the user is evaluated to be satisfied (successful) with the combination of the web element and the device, and a feedback value of the search algorithm is given. The evaluation value fed back in this case is increased from 0.5 to 0.6, for example. Conversely, if the user's browsing stay time is equal to or shorter than the predetermined time ε, the user evaluates as unsatisfactory (failure) and decreases the evaluation value, for example, from 0.5 to 0.4.

  The feedback value of the search algorithm is given by updating the evaluation value of the matching probability distribution stored in the combination information accumulation unit 450. When the evaluation value is updated, the web element / terminal combination unit 410 outputs the combination information again based on the updated evaluation value. As described above, the search algorithm of the matching evaluation unit 440 recursively repeats a predetermined number of times (a loop leading to steps S440 to S410).

  As described above, according to the web element matching apparatus 400 of the third embodiment, the web search is performed by a combination search algorithm based on log data such as the user's staying time of the web page, the departure from the web page, and the change of the page configuration. It is possible to optimize the matching between elements and equipment.

  If there is a function of freely customizing between the web element and the device, the user's change operation (correct answer after change) may be given as a feedback value to update the evaluation value. Further, as described above, the browsing stay time may be output as it is as the browsing status output by the browsing status measurement unit 430. In that case, the matching staying unit 440 converts the browsing stay time into an evaluation value. Alternatively, the user operation log may be output as it is as the browsing status, and the operation log may be converted into an evaluation value. In short, the matching evaluation unit 440 uses the search algorithm based on the browsing situation to update the evaluation value used in the algorithm, and the browsing situation is any type of information as long as it is based on a user operation. It doesn't matter.

  As described above, according to the web element matching apparatus 100 of the first embodiment, it is possible to determine the optimum web element for the device set by the maximum likelihood estimation based on the matching probability distribution. Further, according to the matching probability distribution generation device 200 of the first embodiment, a matching probability distribution used in the web element matching devices 100 and 400 of the present embodiment can be generated.

  Further, according to the web element matching apparatus 300 of the second embodiment, even when the matching probability distribution is not clear, an optimal device set can be matched with the web element. That is, even when advance information from the web page provider side is not given, it is possible to estimate the optimal combination of web elements and devices by complementing the information.

  Further, according to the web element matching apparatus 400 of the third embodiment, it is possible to determine the optimum web element for the device set that also considers the viewpoint of the user. That is, optimization can be performed in consideration of both the sense (sense) on the web page providing side and the sense (sense) of the user.

  As described above, according to the web element matching apparatus of the present embodiment, when a web page is divided and the page is browsed in cooperation between client terminals, the web element is automatically matched with an optimal device. For example, when a user browses one web page with a television and a smartphone, it is possible to provide a combination of web elements and devices that maximizes user satisfaction. In addition, the web element matching apparatus according to the present embodiment can also be used when a web service provider provides a service with a combination that maximizes the convergence rate of advertisement content.

  When the processing means in the above apparatus is realized by a computer, the processing contents of the functions that each apparatus should have are described by a program. Then, by executing this program on the computer, the processing means in each apparatus is realized on the computer.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Further, the program may be distributed by storing the program in a recording device of a server computer and transferring the program from the server computer to another computer via a network.

Each means may be configured by executing a predetermined program on a computer, or at least a part of these processing contents may be realized by hardware.
As described above, the present invention is not limited to the above-described embodiment, and various modifications can be made within the scope of the gist thereof.

100: Web element matching device 110: Combination information accumulation unit 120: Optimal combination candidate determination unit 200: Matching probability distribution generation device 210: Web element function estimation unit 220: Combination information accumulation unit 230: Matching probability distribution generation unit 300: Web element Matching device 310: Combination information accumulating unit 320: Matching probability distribution presence / absence determining unit 400: Web element matching device 410: Web element / device combination unit 420: Client terminal matching unit 430: Browsing situation measuring unit 440: Matching evaluation unit 800: Device Group 900: Web server

Claims (8)

A combination information accumulating unit storing a matching probability distribution which is a probability distribution representing a probability of a combination of a web element and a device set obtained by dividing a web page into constituent elements;
With the web element and the matching probability distribution as inputs, an optimum combination candidate determining unit that determines a matching candidate represented by a set of the web element and the device set by maximum likelihood estimation based on the matching probability distribution;
A web element matching apparatus comprising: A web element function estimation unit that estimates a function of the web element by using a web element obtained by dividing the web page into component elements,
A combination information accumulating unit for storing a web element matching table represented by a probability distribution of a relationship between a function of a web element across a plurality of web pages and a device set;
Using the web element function as an input, referring to the web element matching table, and using the probability distribution of the relationship between the function of the web element and the device set, a matching probability distribution generation unit,
A matching probability distribution generation device comprising: In the web element matching apparatus according to claim 1,
The combination information accumulating unit also stores a web element matching table represented by a probability distribution of a relationship between a web element function and a device set across a plurality of web pages.
Furthermore,
A web element function estimation unit that estimates a function of the web element by using a web element obtained by dividing the web page into component elements,
Using the web element function as an input, referring to the web element matching table, and using the probability distribution of the relationship between the function of the web element and the device set, a matching probability distribution generation unit,
A web element matching device comprising: In the web element matching apparatus according to claim 1 or 3,
Furthermore,
A web element / apparatus combination unit that outputs a combination information having the highest probability value or evaluation value of the matching probability distribution, using a plurality of matching candidates or update information as input,
A client terminal matching unit that matches the web element to a client terminal included in the device set based on the combination information;
A browsing status measuring unit that detects a browsing stay time and user operation of the user's web page in the client terminal and outputs a browsing status;
A matching evaluation unit that outputs the update information that updates the evaluation value using a search algorithm based on the browsing status;
A web element matching device comprising: The web element matching device includes an optimal combination candidate determination unit,
Based on the matching probability distribution, which is a probability distribution representing the probability of the combination of the web element and the device set stored in the combination information accumulation unit, with the input of the web element obtained by dividing the web page into constituent elements and the matching probability distribution An optimal combination candidate determination process for determining a matching candidate represented by a combination of the web element and the device set by maximum likelihood estimation;
Web element matching method. The matching probability distribution generation device includes a web element function estimation unit, a combination information accumulation unit, and a matching probability distribution generation unit,
A web element function estimation process for estimating a function of a web element by inputting a web element obtained by dividing a web page into constituent elements,
Using the function of the web element as an input, refer to the web element matching table represented by the probability distribution of the relationship between the function of the web element across a plurality of web pages stored in the combination information accumulation unit and the device set, and A matching probability distribution generation process for generating a matching probability distribution using a probability distribution of a relationship between a function of a web element and a device set;
A matching probability distribution generation method.   A program for causing a computer to function as the web element matching device according to claim 1, 3 or 4.   A program for causing a computer to function as the matching probability distribution generation device according to claim 2.