JP2020042333A - Monitoring device, shopping server, monitoring method, and program - Google Patents

Abstract

To provide a monitoring device, a shopping server, a monitoring method, and a program capable of returning a search result reflecting a user's intention.SOLUTION: A monitoring device narrows down to a specific category for queries inputted by a user, in an e-commerce site, and outputs information on the necessity of updating rules for introducing a product or service (hereinafter product etc.). The monitoring device comprises: an extraction unit which refers to history information of the queries inputted by the user, and extracts the query in which a change in the frequencies inputted as the queries is a reference value or more, before a prescribed timing and after the prescribed timing; and an output unit which outputs recommended information which recommends to update the rule, related to the query extracted by the extraction unit.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a monitoring device, a shopping server, a monitoring method, and a program.

  There is known a technique for narrowing down search results having higher relevance among search results searched by a query (for example, see Patent Document 1).

JP 2010-118021 A

  By the way, a query in which search results with high relevance are narrowed down may be used for a search with a new meaning of a query. In this case, even if the user inputs a query having another meaning as an intention, the search result is narrowed down to a highly relevant search result conventionally associated with the query. For this reason, search results intended by the user may not be provided in some cases.

  The present invention has been made in view of such circumstances, and provides a monitoring device, a shopping server, a monitoring method, and a program that can return a search result according to a user's intention. One of the purposes.

  One embodiment of the present invention provides information on the necessity of updating a rule for introducing a product or service (hereinafter referred to as a product or the like) by narrowing down a query input by a user to a specific category in an e-commerce site. A monitoring device that outputs a query and refers to history information of a query input by a user, and extracts a query in which a change in the number of times of input as a query is equal to or more than a reference before and after a predetermined timing. A monitoring device comprising: a unit; and an output unit that outputs recommendation information that recommends updating the rule with respect to the query extracted by the extraction unit.

  According to one embodiment of the present invention, it is possible to return a search result according to a user's intention.

It is a figure showing an example of composition of shopping server 100 concerning a 1st embodiment. FIG. 6 is a diagram illustrating an example of an interface screen IM. FIG. 4 is a diagram illustrating an example of the content of whitelist information 121. FIG. 4 is a diagram illustrating an example of the content of history information 122. FIG. 7 is a diagram illustrating an example of the content of total information 123. FIG. 7 is a diagram illustrating an example of a query to be extracted by an extraction unit 212. It is a figure which shows an example of the time series data of the number of requests of "ABC BEAT". It is a diagram illustrating an example of time-series data of "ABC BEAT" change point score Score _{(y t).} It is a flow chart which shows an example of operation of extraction processing of a query with a changed meaning according to the first embodiment. 5 is a flowchart illustrating an example of an operation of a whitelist update process according to the first embodiment. It is a figure showing an example of composition of shopping server 100a concerning a 2nd embodiment. It is a flow chart which shows an example of operation of extraction processing of a query with a changed meaning according to the second embodiment. It is a figure showing an example of composition of shopping server 100b concerning a 3rd embodiment. It is a figure showing an example of the contents of new goods information 124 concerning a 3rd embodiment. It is a flow chart which shows an example of operation of update processing of a white list concerning a 3rd embodiment.

  Hereinafter, embodiments of a monitoring device, a shopping server, a monitoring method, and a program according to the present invention will be described with reference to the drawings.

  The monitoring device is realized by one or more processors. The monitoring device is a device that monitors whether a category narrowed down based on a query input by a user is appropriate at an e-commerce site. The e-commerce site is, for example, a shopping site, an auction site, a flea market site, or the like. In the following description, it is assumed that an electronic commerce site is a shopping site, and an apparatus that provides the site is referred to as a shopping server. The monitoring device may be configured integrally with the shopping server or the like, or may be configured separately. The shopping server, for example, searches a whitelist based on the input query, narrows down to a corresponding category, and introduces a product or service (hereinafter, product, etc.) to the user. Introducing a product or the like means, for example, presenting a product or the like as a search result of a query on a screen. The white list is, for example, information in which a query is associated with a category. "If the query entered by the user is included in the whitelist, read the category specified in the whitelist and apply it to the introduction of products, etc." Is an example.

  The monitoring device refers to the history information of the query input by the user, and extracts a query whose change in the number of times the query has been input is equal to or greater than a reference between before and after a predetermined timing. Then, the monitoring device outputs recommendation information that recommends updating the whitelist for the extracted query. Accordingly, the monitoring device can determine that the meaning of the query included in the whitelist has changed, prompt the operator to update the whitelist, or automatically perform the update. As a result, the monitoring device can guide the shopping server to return the search result according to the user's intention.

<First embodiment>
FIG. 1 is a diagram illustrating an example of a configuration and a use environment of a shopping server 100 according to the first embodiment. The shopping server 100 communicates with one or more terminal devices 10 via a network NW, for example. The network NW includes, for example, a WAN (Wide Area Network), a LAN (Local Area Network), the Internet, a dedicated line, a wireless base station, a provider, and the like.

  The terminal device 10 is a terminal device including an input device, a display device, a communication device, a storage device, and an arithmetic device, such as a mobile phone such as a smartphone, a tablet terminal, and various personal computers. The communication device includes a network card such as a NIC (Network Interface Card), a wireless communication module, and the like. The terminal device 10 includes a terminal device used by a seller of a product and a terminal device used by a purchaser of a product, and these are illustrated in FIG. 2 without distinction. In the terminal device 10, a UA (User Agent) such as a browser or an application program starts. The UA displays a screen provided from the shopping server 100 and transmits a request corresponding to an input operation performed by a user of the terminal device 10 to the shopping server 100. In the following description, it is assumed that UA is a browser and shopping server 100 is a web server. In the terminal device 10, a UA such as a web browser or an application program is activated, and transmits a request corresponding to the content input by the user to the shopping server 100. In addition, the terminal device 10 on which the UA is activated causes the display device to display various images based on the information acquired from the shopping server 100.

  The shopping server 100 includes, for example, a shopping site providing unit 110, a storage unit 120, and a monitoring unit 210. The shopping site providing unit 110 includes, for example, a sales management unit 111, a history information generation unit 112, a reception unit 113, a white list generation unit 114, and a white list update unit 115. The monitoring unit 210 includes, for example, a tally information generation unit 211, an extraction unit 212, and an output unit 213. Each unit of the shopping site providing unit 110 and the monitoring unit 210 is realized by a hardware processor such as a CPU (Central Processing Unit) executing a program (software). Some or all of these constituent elements are hardware (circuits) such as an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), and a GPU (Graphics Processing Unit). (Including a circuitry), or may be realized by cooperation of software and hardware. The program may be stored in a storage device such as an HDD (Hard Disk Drive) or a flash memory in advance, or may be stored in a removable storage medium such as a DVD or a CD-ROM. It may be installed by being attached. Further, the shopping site providing unit 110 and the monitoring unit 210 may be realized by an integrated processor or may be realized by separate processors. The monitoring unit 210 provided in the shopping server 100 is an example of a “monitoring device”.

  The storage unit 120 stores, for example, whitelist information 121, history information 122, and total information 123. The storage unit 120 is realized by, for example, a RAM (Random Access Memory), an HDD, a flash memory, an SD card, or the like.

[About configuration of shopping site providing unit 110]
The sales management unit 111 performs processing of a part related to sales among services provided by the shopping site. The sales management unit 111 provides the terminal device 10 with various screens serving as an interface of a shopping site, for example. FIG. 2 is a diagram illustrating an example of the interface screen IM. The interface screen IM includes, for example, a query input column A1 for inputting a query used when searching for a product, a search button B1 for executing a search, and a product etc. introduction column A2 for introducing (presenting) a product. Etc. are provided on the screen.

  The user performs various operations on the interface screen IM provided by the sales management unit 111 to purchase a product on the shopping site. For example, when the user of the terminal device 10 performs an operation to purchase a product, the sales management unit 111 performs a settlement process related to this transaction. The settlement process is, for example, a process of transmitting settlement information relating to this sale to a settlement server (not shown) that executes the settlement process, and requesting the settlement process.

  The sales management unit 111 searches the whitelist using the query input in the query input field A1 of the shopping site as a search key, and acquires a category of a product associated with the query when the category exists. . Then, the sales management unit 111 presents the products narrowed down to the specified category from the various products indicated by the query search result in the product etc. introduction column A2, and introduces (presents) them to the user.

  FIG. 3 is a diagram illustrating an example of the content of the whitelist information 121. The whitelist information 121 is, for example, information indicating a whitelist. The whitelist is, for example, information in which a query and a specific category corresponding to the query are associated with each other. The specific category associated with the query is a category that is known in advance to be highly relevant to the query. The white list is generated by, for example, a white list generation unit 114 described later, and is stored in the storage unit 120 as white list information 121. In the whitelist information 121 shown in FIG. 3, the query “PC monitor” is associated with “liquid crystal display” which is a category highly relevant to “PC monitor”.

  Returning to FIG. 2, for example, when the query input to the query input field A1 of the shopping site is “PC monitor”, the sales management unit 111 uses the “PC monitor” as a search key to search the whitelist of the whitelist information 121. A search is performed to obtain a category (“liquid crystal display” in the example shown in FIG. 2) associated with “PC monitor”. In the product introduction column A2, the sales management unit 111 introduces (presents) a product of the specified category (that is, “liquid crystal display”) from various products indicated by the search result of “PC monitor”.

  Referring back to FIG. 1, the history information generation unit 112 generates the history information 122. The history information 122 is, for example, information indicating a user's behavior history on a shopping site for each user and for each session. The session is, for example, a period in which a certain user starts browsing a shopping site until (A) a predetermined time elapses (that is, a timeout occurs), and (B) a user moves to another site. Or (C) a period until the window (or application) is closed.

  FIG. 4 is a diagram illustrating an example of the content of the history information 122. The history information 122 shown in FIG. 4 includes, for each session ID for identifying each session, a record in which the history of the action that occurred during the session period and the date and time when the action was performed are associated with each user. The data is summarized in For example, when the user searches for “XX! Shopping site”, which is one of the sales sites, using X as a query, browses the product A, and subsequently purchases the product A, the session ID included in the history information 122 A plurality of action histories are associated with one session ID as in the record “SESH_1”.

  Returning to FIG. 1, the accepting unit 113 is an operation performed by the user on the terminal device 10 after the recommendation information for recommending updating the whitelist is output by the output unit 213 described later. Operation information indicating an operation to instruct to update or an operation to instruct not to update is received.

  The whitelist generation unit 114 generates a whitelist based on the tally information 123. FIG. 5 is a diagram illustrating an example of the content of the tally information 123. The aggregation information 123 is, for example, in a shopping site, a category of a product selected (clicked) from search results after a query search and a number of clicks on a product in the category are associated with each other. A record is data compiled for each query. The total information generation unit 211 illustrated in FIG. 4 indicates that the number of clicks on the product of the category “liquid crystal display” is 1200 times from the search result of the query “PC monitor”. Indicates that the number of clicks on the product of the “monitor arm” is 100 times.

The whitelist generation unit 114 generates a whitelist by any of the following methods, for example.
(1) The whitelist generation unit 114 narrows down, for example, to queries having a total number of clicks equal to or more than a predetermined number among the queries indicated in the aggregation information 123, and among the product categories associated with the narrowed-down queries, Identify the one with the most clicks as the category most relevant to the query.
(2) The whitelist generation unit 114 extracts, for example, a product category having the largest number of clicks from among the product categories associated with each query indicated in the aggregation information 123, and the extracted category has a predetermined number of clicks. In the case described above, the category is specified as a category highly relevant to the query.

  The whitelist generation unit 114 specifies “liquid crystal display” as a category highly relevant to the query “PC monitor” by the method (1) or (2) based on the aggregation information 123 shown in FIG. I do. The whitelist generation unit 114 generates a whitelist that associates the “PC monitor” with the specified “liquid crystal display”, and causes the storage unit 120 to store the whitelist as whitelist information 121.

  The whitelist updating unit 115 updates the whitelist of the whitelist information 121 when the user's instruction received by the receiving unit 113 is an instruction to update the whitelist. The details of the whitelist updating unit 115 will be described later.

[Configuration of monitoring unit 210]
The tally information generation unit 211 generates tally information 123 based on the history information 122. The aggregation information generation unit 211 specifies an action history (in the example illustrated in FIG. 2, “search using X as a query”) indicating that a query has been input, from the action history included in the history information 122. . Next, the tally information generation unit 211 is an operation performed after the identified action history, and is an operation of selecting a product from the search results (in the example shown in FIG. Click A)).

Next, the total information generation unit 211 specifies the category of the product clicked after the search for the query. The aggregation information generation unit 211 acquires the category of the product clicked after the search for the query, for example, by any of the following methods.
(1) The aggregation information generation unit 211 specifies a category of a product clicked after a query search based on, for example, a category tree that is correspondence information indicating a correspondence between a product category and a product and the history information 122. I do.
(2) In the history information 122 of FIG. 4, the action history is represented by the content of the action, but in reality, it may be represented by a URL (Uniform Resource Locator) browsed according to the action. In this case, for example, the product is embedded in the URL in a state where the product is associated with the category expressed stepwise. The aggregation information generation unit 211 specifies the category of the product clicked after the search for the query based on the URL indicated in the history information 122.

  The aggregation information generation unit 211 repeats the above-described processing, and obtains, for each query, the number of times the product in the specified category is clicked.

  Next, the extracting unit 212 will be described. FIG. 6 is a diagram illustrating an example of a query to be extracted by the extraction unit 212. Here, the whitelist may include a query whose meaning changes later. For example, if a store called "ABC Beat!", Which is a car parts store, exists on the shopping site and the category of the product sold at this store is "car parts", the white list includes "ABC BEAT". "Car parts" is associated as a highly relevant category. Thereafter, a smartphone named “ABC BEAT” is released. In this case, uppercase and lowercase letters are treated as the same in the whitelist, and "!" Is ignored, so the query “ABC BEAT” has two meanings, “car parts” and “home appliances” before and after the smartphone release date.

  Even after the meaning of the query “ABC BEAT” has changed, when introducing products that are narrowed down to the specified category based on the whitelist, the shopping server 100 is associated with the query of the new meaning (smartphone). It is difficult to introduce products (household appliances) in different categories. Therefore, it is preferable that the shopping server 100 performs a process of extracting a query whose meaning has changed and removing the extracted query from the whitelist.

  Here, the number of requests for the query “ABC BEAT” at the shopping site changes between before and after the release date when the smartphone named “ABC BEAT” was released (for example, Increase after a day). In other words, when the meaning of the query changes due to the release of a new product or the like, the number of query requests may increase before and after the meaning of the query changes.

  FIG. 7 is a diagram illustrating an example of time-series data of the number of requests for “ABC BEAT”. A waveform W1 shown in FIG. 7 is a waveform indicating a change over time in the number of requests for “ABC BEAT”. The extraction unit 212 is, for example, a query included in a whitelist based on the history information 122, and determines the number of requests for a query whose score is to be calculated (hereinafter, a target query) every predetermined time (for example, one day). Is calculated in time series. As shown by the waveform W1, the number of requests for “ABC BEAT” rapidly increases after the release date in February 2017. For example, the extraction unit 212 refers to the history information 122, and the change in the number of requests is equal to or greater than a reference between a certain time t before a predetermined timing (for example, three days) and a predetermined timing (for example, three days). Extract the query as a query whose meaning has changed.

  The extraction unit 212 according to the present embodiment detects, for example, a change point in the number of times of the query request using the Change Finder. Change Finder is a method of detecting a change point. The change point is a point at which the property of the data that changes with time changes (that is, the point at which the meaning of the query changes). An outlier is a value that deviates significantly from other values in statistics. The Change Finder is a change point detection method that detects a change point by regarding the continuation of an outlier as a change point. Hereinafter, a method of detecting a change point using the change finder by the extraction unit 212 will be described.

[Step 1: Probability density learning by SDAR (first stage learning)]
Extraction unit 212, based on the history information 122, a request number x _t of the target query, a predetermined time (e.g., 1 day) is calculated by a time series for each. In the following description, when a request number _{x t} calculated by serial, time-series data _{{x t: t = 1,2,} ...} is described as. The extraction unit 212 performs learning of the probability density by using an SDAR (Sequentially Discounting Auto Regressive) model using the time series data {x _t : t = 1, 2,...}.

  The extraction unit 212 uses the SDAR model to create a time-series model at each time point of time t = 1, 2,... (That is, based on time-series data up to each time point of time t = 1, 2,. Learn (construct) the probability density function. Specifically, the extraction unit 212 learns the probability density function by estimating the parameter PT1 that maximizes I on the left side of Expression (1) based on Expression (1). The parameter PT1 is calculated by, for example, the Yule Walker method.

Next, the extraction unit 212, based on the probability density function of the request number x _t learned with SDAR model to calculate the probability density. Specifically, the extraction unit 212 calculates the probability density _pt-1 ( _xt ) based on Expression (2).

If the value of the expression probability density _{p t-1 (x t)} as indicated by (2) is large, the value of the predicted actual time-series data _{{x t: t = 1,2,} ...} and that is near In the case where the value of the probability density _pt-1 ( _xt ) is small, it indicates that the predicted value deviates from the actual time-series data { _xt : t = 1, 2,...}.

[Step 2: outlier score calculation (acquisition of outlier likeness)]
Next, the extraction unit 212 substitutes the calculated probability density _pt-1 ( _xt ) into Expression (3), and outlier score Score ( _xt ) at each time point of time t = 1, 2,. Is calculated.

The outlier score Score ( _xt ) is a value indicating the likelihood of an outlier. When the outlier score Score ( _xt ) is large, it is highly possible that the number of requests xt at time _t is an outlier.

[Step 3: Calculation of Moving Average of Outlier Score (Generation of New Time Series Data)]
Next, the extraction unit 212 substitutes the calculated outlier score Score (x _t ) into Equation (4), and calculates a moving average y _t of the outlier score Score (x _t ). The moving average y _t is an average of the outlier score Score (x _t ) from before the movement time T (for example, three days) at a certain time t to after the movement time T (for example, three days) at a certain time t. Is a value that indicates

From Expression (4), the moving average y _t is obtained by smoothing the outlier score Score (x _t ).

[Step 4: Probability density learning by SDAR (second stage learning)]
Next, the extraction unit 212, the formula (4) a predetermined time the moving average y _t calculated by (e.g., one day) is calculated by a time series by. In the following description, a moving average _{y t} calculated by the time series, time series data _{{y t: t = 1,2,} ...} is described as. The extraction unit 212 learns the probability density by the SDAR model using the time series data {y _t : t = 1, 2,...}.

The extraction unit 212 estimates the parameter PT2 that maximizes I on the left side of the equation (5), as in the case of the time-series data {x _t : t = 1, 2,. Learn the density function.

Next, the extraction unit 212, based on the probability density function of the moving average y _t learned with SDAR model to calculate the probability density. Specifically, the extraction unit 212, based on the formula (6), calculates the probability density _{q t-1} a _{(y t).}

[Step 5: Change point score calculation (acquisition of change point likeness)]
Next, the extraction unit 212 substitutes the calculated probability density q _t−1 (y _t ) into Expression (7), and changes point score Score (y _t ) at each time point of time t = 1, 2,. Is calculated.

If the change point score Score (y _t ), which is a value indicating the likelihood of the change point, is large, the change point score Score (y _t ) is likely to be the time point t.

The extraction unit 212 executes the above-described processes of Steps 1 to 5, and calculates a change point score Score (y _t ). The extraction unit 212 compares the calculated change point score Score (y _t ) with a predetermined threshold, and when the change point score Score (y _t ) is equal to or greater than the predetermined threshold, the query in which the meaning of the target query has changed. Extract as

FIG. 8 is a diagram illustrating an example of the time-series data of the change point score Score (y _t ) of “ABC BEAT”. A waveform W2 shown in FIG. 8 is a waveform indicating a change with time of the change point score Score (y _t ) of “ABC BEAT”. As shown by the waveform W2, the change point score Score (y _t ) of “ABC BEAT” sharply increases on the release date in February 2017. The extraction unit 212 compares the change point score Score (y _t ) with a predetermined threshold value (30 shown in the figure). If the change point score Score (y _t ) is equal to or greater than the predetermined threshold value, the extraction unit 212 indicates the target query. Is extracted as a changed query. Therefore, the extraction unit 212 extracts “ABC BEAT” as a query whose meaning has changed.

  Note that the extraction unit 212 adopts a method other than the Change Finder as long as it can extract a query whose change in the number of times of input as a query is equal to or more than a reference before and after a predetermined timing. You may. For example, if the value obtained by subtracting the average of the predetermined period (for example, three days) before the predetermined timing from the average of the predetermined period (for example, three days) after the predetermined timing is equal to or more than the predetermined value, The target query may be extracted as a query whose meaning has changed. Further, the extraction unit 212 may extract a query in which the meaning of the target query has changed based on a moving average during a predetermined period, instead of the average during the predetermined period.

  Returning to FIG. 1, when a query is extracted by the extraction unit 212, the output unit 213 outputs recommendation information that recommends updating the whitelist for the query. The output unit 213 outputs the recommended information to, for example, the shopping site providing unit 110 or the terminal device 10 used by the administrator of the shopping server 100. In the following description, it is assumed that the output unit 213 outputs the recommended information to the terminal device 10 used by the administrator of the shopping server 100. The administrator of the shopping server 100 inputs an operation for instructing to update the whitelist or an operation for instructing not to update the whitelist to the terminal device 10 based on the output recommended information.

  The whitelist updating unit 115 updates the whitelist based on the operation information received by the receiving unit 113. When the operation information indicates an operation to instruct to update the whitelist, the whitelist update unit 115 excludes the query extracted by the extraction unit 212 from the whitelist. When the operation information indicates an operation not instructing not to update the whitelist, the whitelist update unit 115 does not exclude the query extracted by the extraction unit 212 from the whitelist.

[Query extraction processing]
FIG. 9 is a flowchart illustrating an example of an operation of a process of extracting a query whose meaning has changed according to the first embodiment. The flowchart shown in FIG. 9 is executed for each query included in the whitelist at predetermined time intervals (for example, every day). First, the extraction unit 212 calculates a change point score Score (y _t ) of the target query based on the history information 122 (S100). Next, the extraction unit 212 determines whether the calculated change point score Score (y _t ) of the target query is _{equal to} or larger than a predetermined threshold (S102). If the calculated change point score Score (y _t ) of the target query is not equal to or larger than the predetermined threshold, the extraction unit 212 ends the process. The output unit 213 outputs recommendation information to the terminal device 10 when the change point score Score (y _t ) of the target query calculated by the extraction unit 212 is equal to or greater than a predetermined threshold (S104).

[Whitelist update process]
FIG. 10 is a flowchart illustrating an example of an operation of a whitelist update process according to the first embodiment. First, the receiving unit 113 determines whether operation information has been received from the terminal device 10 (S200). The receiving unit 113 waits until receiving operation information from the terminal device 10. When the operation information is received from the terminal device 10 by the reception unit 113, the whitelist update unit 115 determines whether the operation information indicates an operation to instruct to update the whitelist (S202). When the operation information received by the receiving unit 113 indicates an operation instructing not to update the whitelist, the whitelist updating unit 115 ends the process. When the operation information received by the receiving unit 113 is an operation instructing to update the whitelist, the whitelist updating unit 115 excludes the target query extracted by the extracting unit 212 from the whitelist, The white list of the white list information 121 is updated (S204).

[Summary of First Embodiment]
As described above, the shopping server 100 of the present embodiment narrows down a query input by a user to a specific category on an e-commerce site (a shopping site in this example) to introduce a product or service. (In this example, a whitelist in this example) for outputting information (in this example, recommended information) on the necessity of updating, and refers to the history information 122 of the query input by the user, and It is recommended that the extraction unit 212 extract a query whose change in the number of requests input as a query is equal to or greater than a reference before and after a predetermined timing, and that the whitelist be updated for the query extracted by the extraction unit 212 And an output unit 213 for outputting recommended information to be performed. It is possible to return the search results along.

In the shopping server 100 of the present embodiment, the extraction unit 212 calculates an index value (in this example, a change point score Score (y _t )) indicating a change in the number of requests, and obtains a change point score Score (y _t ). A query is extracted by comparing the index value with a predetermined threshold value indicating the reference. As a result, the shopping server 100 of the present embodiment can return a search result that meets the user's intention with less processing than in the case of using an AR (Auto Regressive) model.

  In addition, when the recommendation information is output by the output unit 213, the shopping server 100 according to the present embodiment specifies whether to exclude the query extracted by the extraction unit 212 from the whitelist of the whitelist information 121. A reception unit 113 that receives information from a user, a whitelist update unit 115 that updates whitelist information 121 based on operation information received by the reception unit 113, and a sales management unit 111 that provides a service as a shopping site. , And by excluding a query whose meaning has changed from the whitelist, it is possible to appropriately introduce products.

<Second embodiment>
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. The shopping server 100 of the first embodiment outputs recommendation information when a query is extracted by the extraction unit 212, but the shopping server 100a of the second embodiment determines the part of speech of the query extracted by the extraction unit 212. And output recommended information when the query is a noun. In addition, about the structure similar to embodiment mentioned above, the same code | symbol is attached and description is abbreviate | omitted.

  FIG. 11 is a diagram illustrating an example of a configuration of a shopping server 100a according to the second embodiment. The shopping server 100a includes a monitoring unit 210a instead of (or in addition to) the monitoring unit 210 included in the shopping server 100. The monitoring unit 210a includes, for example, a tally information generation unit 211, an extraction unit 212, an output unit 213, and a query part-of-speech determination unit 214.

  The query part of speech determination unit 214 determines the part of speech of the query extracted by the extraction unit 212. Here, a query that is included in the whitelist and that changes meaning later is likely to be a noun. The query part-of-speech determining unit 214 performs a morphological analysis on the query extracted by the extracting unit 212, for example, and divides the query into words. The query part-of-speech determining unit 214 determines whether or not the divided word is a noun.

  The output unit 213 of the present embodiment outputs recommended information to the terminal device 10 when the query is determined to be a noun by the query part-of-speech determination unit 214, and outputs the recommended information to the terminal device 10 when the query is determined to be not a noun. Do not output recommended information to

[Query extraction processing]
FIG. 12 is a flowchart illustrating an example of an operation of extracting a query whose meaning has changed according to the second embodiment. Among the processing of S100 to S104 shown in FIG. 12, the processing other than S103 is the same as the processing of the same step numbers in FIG. 9, and therefore the description is omitted. When the change point score Score (y _t ) of the target query calculated by the extraction unit 212 is equal to or greater than a predetermined threshold, the query part of speech determination unit 214 determines whether the target query is a noun (S103). If the target query is not a noun, the query part-of-speech determination unit 214 ends the process. If the query part-of-speech determining unit 214 determines that the target query is a noun, the output unit 213 advances the process to S104.

[Timing of part of speech judgment]
In the above description, the case where the query part-of-speech determination unit 214 determines whether the target query extracted by the extraction unit 212 is a noun is described, but the present invention is not limited to this. For example, the extraction unit 212 may be configured to calculate a change point score Score (y _t ) for a query included in the whitelist and determined as a noun by the query part-of-speech determination unit 214. In this case, since the extraction unit 212 previously calculates the change point score Score (y _t ) for the query that is a noun, it is possible to reduce the labor involved in the process of calculating the change point score Score (y _t ). .

[Summary of Second Embodiment]
As described above, the shopping server 100a of the present embodiment further includes the query part-of-speech determining unit 214 that determines whether the query extracted by the extracting unit 212 is a noun, and the output unit 213 performs the query part-of-speech determination. If the unit 214 determines that the query is not a noun, the recommendation information is not output. Thereby, the shopping server 100a of the present embodiment can return the search result according to the user's intention with higher accuracy.

<Third embodiment>
Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. In the shopping server of the first embodiment and the shopping server of the second embodiment, a case has been described in which, based on an instruction of the administrator of the shopping server, a query that is considered to have a change in meaning is removed from the whitelist. In the shopping server 100b of the present embodiment, a case will be described in which a query whose meaning has changed is removed from the whitelist based on the new product information. In addition, about the structure similar to embodiment mentioned above, the same code | symbol is attached and description is abbreviate | omitted.

  FIG. 13 is a diagram illustrating an example of a configuration of a shopping server 100b according to the third embodiment. The shopping server 100b includes a shopping site providing unit 110b instead of (or in addition to) the shopping server 100 and the shopping site providing unit 110 included in the shopping server 100a. The shopping server 100b includes a storage unit 120b instead of (or in addition to) the storage unit 120 included in the shopping server 100 and the shopping server 100a.

  The shopping site providing unit 110b includes a sales management unit 111, a whitelist generation unit 114, a whitelist update unit 115, and a new product determination unit 116. The storage unit 120b stores whitelist information 121, history information 122, total information 123, and new product information 124.

  FIG. 14 is a diagram illustrating an example of the content of the new product information 124 according to the third embodiment. The new product information 124 is information in which the name of a recently released product or the name of a product to be released from now on is associated with the release date. The new product information 124 shown in FIG. 14 includes a record in which a product named “ABC BEAT” and a release date of the product “February 2017” are associated with each other.

Returning to FIG. 13, when the query is extracted by the extraction unit 212, the output unit 213 of the present embodiment outputs the recommended information to the shopping site providing unit 110b. When the recommendation information is output from the output unit 213, the new product determination unit 116 searches the new product information 124 using the query extracted by the extraction unit 212 as a search key, and matches the query or a name matching the query. Identify the product. Also, the new product determination unit 116 matches the release date associated with the specified product with the time t at which the change unit score Score (y _t ) is _{equal to} or greater than the predetermined threshold by the extraction unit 212. It is determined whether or not.

The whitelist update unit 115 of the present embodiment matches the release date of the product by the new product determination unit 116 with the time t at which the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold by the extraction unit 212. In such a case, it is considered that the meaning of the query has changed with the release of the new product, and the query extracted by the extraction unit 212 is excluded from the whitelist. In addition, the whitelist update unit 115 determines that the release date of the product by the new product determination unit 116 does not match the time t at which the change point score Score (y _t ) is _{equal to} or more than the predetermined threshold by the extraction unit 212. It is assumed that the meaning of the query associated with the new product has not changed, and the query extracted by the extraction unit 212 is not excluded from the whitelist.

[Whitelist update process]
FIG. 15 is a flowchart illustrating an example of an operation of a whitelist update process according to the third embodiment. First, the new product determination unit 116 determines whether recommendation information has been output from the output unit 213 (S300). The new product determination unit 116 waits until the recommendation information is input from the output unit 213. When the recommendation information is output from the output unit 213, the new product determination unit 116 matches or matches the extracted query with the product name based on the query extracted by the extraction unit 212 and the new product information 124. It is determined whether or not the release date of the product to be executed matches the time t at which the change point score Score (y _t ) is _{equal to} or greater than a predetermined threshold (S302). The whitelist updating unit 115 does not match the extracted query with the product name or the release date of the matching product and the time t at which the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold. Is completed by the new product determination unit 116, the process ends. The whitelist updating unit 115 determines that the extracted query and the product name match, or that the release date of the matching product matches the time t when the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold. If determined by the new product determination unit 116, the whitelist of the whitelist information 121 is updated by excluding the target query extracted by the extraction unit 212 from the whitelist (S204).

[Summary of Third Embodiment]
As described above, when the recommendation information is output by the output unit 213, the shopping server 100b of the present embodiment includes the new product information 124 in which the product name and the release date of the product are associated with each other, and the extraction unit The new product determination unit 116 determines whether to exclude the query extracted by the extraction unit 212 from the whitelist information 121 based on the query extracted by the new product determination unit 212, and the whitelist information 121 by the new product determination unit 116. And the new product information 124 that updates the whitelist information 121 by excluding the query determined to be excluded from the whitelist information 121 from the whitelist information 121. the extracted timing (in this example, the change point score score _{(y t)} is Tokoro If the time t) determined to be equal to or greater than the threshold value matches the query indicated in the new product information 124 or the release date of the product having the matching name, the query extracted by the extraction unit 212 is whitelisted. It is determined to be excluded from the information 121. Thereby, the shopping server 100b of the present embodiment can introduce products more appropriately.

[Other whitelist update methods]
In the above description, the case where the whitelist updating unit 115 updates the whitelist based on the release date of the product determined by the new product determination unit 116 has been described, but the present invention is not limited to this. The whitelist updating unit 115 may update the whitelist based on information other than the release date, for example.

For example, the whitelist updating unit 115 may update the whitelist based on the special sale date of the shopping site. For example, on the special sale day of the shopping site, since the number of purchasers of the product increases, the number of requests also increases. For this reason, when the time t at which the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold matches the special sale date of the shopping site, the whitelist updating unit 115 does not need to update the whitelist. Good.

Further, the whitelist updating unit 115 may update the whitelist based on weather conditions. For example, on a day when the temperature rises or a day when the temperature falls due to a rapid temperature change, the number of requests for heat-related or cold-related goods may increase. For this reason, if the time t at which the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold matches the day at which the rapid temperature change occurs, the whitelist updating unit 115 determines whether or You do not need to update the whitelist for cold goods.

Further, the whitelist updating unit 115 may update the whitelist based on the television broadcast information. For example, the number of query requests may increase even when the meaning is not changed due to the influence of television broadcasting or the like. For this reason, the whitelist updating unit 115 matches the time t at which the change point score Score (y _t ) is _{equal to} or greater than the predetermined threshold value with the date on which the query extracted by the extracting unit 212 has been broadcast on television. In this case, the whitelist need not be updated.

  As described above, the embodiments for carrying out the present invention have been described using the embodiments. However, the present invention is not limited to these embodiments at all, and various modifications and substitutions may be made without departing from the gist of the present invention. Can be added.

10 terminal device, 100, 100a, 100b shopping server, 110, 110b shopping site providing unit, 111 sales management unit, 112 history information generating unit, 113 receiving unit, 114 white list generating unit, 115 White list update unit, 116: new product determination unit, 120, 120b: storage unit, 121: white list information, 122: history information, 123: total information, 124: new product information, 210, 210a: monitoring unit, 211 ... Total information generation unit, 212 extraction unit, 213 output unit, 214 query part-of-speech determination unit, Score (y _t ) change point score, Score (x _t ) outlier score, p _t−1 (x _t ) , Q _t-1 (y _t ): probability density, xt: number of requests, yt: moving average

Claims (11)

A monitoring device that outputs information on the necessity of updating a rule for introducing a product or service (hereinafter referred to as a product or the like) by narrowing down a query input by a user to a specific category in an e-commerce site. hand,
An extraction unit that refers to the history information of the query input by the user, and extracts a query whose change in the number of times input as a query is equal to or more than a reference between before and after a predetermined timing.
For the query extracted by the extraction unit, an output unit that outputs recommendation information that recommends updating the rule,
A monitoring device comprising: The rule is defined by whitelist information that associates a category to be narrowed down with respect to the input query,
The monitoring device according to claim 1. The updating includes removing the query extracted by the extracting unit from the whitelist information,
The monitoring device according to claim 2. The extraction unit calculates an index value indicating a change in the number of times, compares a predetermined threshold value indicating a criterion of the index value with the index value, and extracts a query.
The monitoring device according to any one of claims 1 to 3. The history information is history information of a query input to the e-commerce site,
The monitoring device according to claim 1. A query part-of-speech determination unit that determines whether the query extracted by the extraction unit is a noun,
The output unit, if the query is determined by the query part of speech determination unit is not a noun, does not output recommendation information that recommends updating the rule,
The monitoring device according to any one of claims 1 to 5. A monitoring device according to any one of claims 1 to 6,
When the recommendation information is output by the output unit, the user specifies information indicating whether to exclude the query extracted by the extraction unit from whitelist information in which a category to narrow down the input query is associated. And a receiving unit to receive from
A whitelist updating unit that updates the whitelist information based on information received by the receiving unit;
A sales management unit that provides a service as the e-commerce site;
Shopping server provided with. A monitoring device according to any one of claims 1 to 6,
When the recommendation information is output by the output unit, a query input based on new product information in which a product name and a release date of the product are associated with each other and a query extracted by the extraction unit A new product determination unit that determines whether to exclude a query extracted by the extraction unit from whitelist information that associates a category to be narrowed down,
A whitelist update unit that updates the whitelist information by excluding the query determined to be excluded from the whitelist information by the new product determination unit from the whitelist information,
A sales management unit that provides a service as the e-commerce site;
Shopping server provided with. When the timing at which the change in the number of times is extracted by the extraction unit matches the release date of the query indicated in the new product information, the new product determination unit extracts the whitelist information from the whitelist information. Determine that the query was excluded
A shopping server according to claim 8. In an e-commerce site, a computer that outputs information on the necessity of updating a rule for introducing a product or service (hereinafter, product etc.) by narrowing down to a specific category in response to a query input by a user,
With reference to the history information of the query input by the user, a query whose change in the number of times input as a query is equal to or greater than a reference between before and after a predetermined timing is extracted,
Outputting recommendation information that recommends updating the rule for the extracted query;
Monitoring method. On an e-commerce site, a computer that outputs information on the necessity of updating rules for introducing products or services (hereinafter, products, etc.) by narrowing down to a specific category in response to a query input by a user,
By referring to the history information of the query input by the user, before the predetermined timing and after the predetermined timing, to extract a query whose change in the number of times entered as a query is equal to or more than a reference,
For the extracted query, output recommendation information recommending that the rule be updated,
program.

Images