JP6960253B2 - Product evaluation device, product evaluation method, and program - Google Patents

Description

The present invention relates to a product evaluation device, a product evaluation method, and a program.

Conventionally, a web page storage unit that can store a web page, an appearance amount condition storage unit that can store an appearance amount condition that is a condition of the appearance amount of a keyword, a keyword reception unit that accepts a keyword, and a keyword in a web page. The changed part that changes the keyword or the character string other than the keyword in the web page and the web page that the changed part changed so that the appearance amount satisfies the appearance amount condition and the meaning and content of the web page are not changed. The invention of an automatic web page changing device including an output unit for outputting is disclosed (see Patent Document 1).

Japanese Unexamined Patent Publication No. 2011-18258

A query may be entered to search for a product or service (hereinafter referred to as a product, etc.) on a shopping site or the like. Since the number of search hits may affect the ranking of products etc. on the shopping site, in order to increase the number of search hits, list multiple brand names unrelated to the product in the title and product introduction column. Is being done. Such an act is called excessive SEO (Search Engine Optimization) or black hat SEO. Since excessive SEO hinders the formation of an appropriate ranking of products and the like, there is a need to discover products and the like whose sales screen is configured for the purpose of excessive SEO.

The above-mentioned Patent Document 1 aims to increase the exposure of a web page without changing the meaning of the web page. On the other hand, sufficient research has not progressed on the technology for detecting excessive SEO.

The present invention has been made in consideration of such circumstances, and an object of the present invention is to provide a product evaluation device, a product evaluation method, and a program capable of efficiently finding a product or the like to be noted. Make one.

One aspect of the present invention refers to an acquisition unit that acquires a target word / phrase included in an introduction screen of a product or service (hereinafter, product, etc.) sold via a network, and a brand dictionary that describes a brand name. A product including a brand name extraction unit that extracts the brand name included in the target phrase and a monitoring unit that identifies the product to be noted based on the number of brand names extracted by the brand name extraction unit. Equal evaluation device.

According to one aspect of the present invention, it is possible to efficiently find a product or the like to be noted.

It is a figure which shows an example of the configuration and use environment of the sales intermediary server 200 using the product evaluation apparatus, and the brand dictionary creation apparatus 300. It is a figure which shows an example of the sale screen IM1 of a product or the like. It is a figure which shows an example of the search result display screen IM2. It is a figure which shows an example of the product display screen IM3. This is an example of the product display screen IM4 in which excessive SEO is performed. It is a figure which shows an example of the content of the product data 220. It is a figure which shows an example of the content of ranking data 222. It is a figure which shows an example of the contents of a brand dictionary 240. It is a flowchart which shows an example of the flow of the process executed by the target word acquisition unit 230, the brand name extraction unit 232, and the attention target product monitoring unit 234 of the sales intermediary server 200. It is a figure for demonstrating the content of processing by narrowing-down part 312. It is a flowchart which shows an example of the flow of processing executed by the brand dictionary creation apparatus 300.

Hereinafter, with reference to the drawings, the product evaluation device of the present invention, the product evaluation method, and the embodiment of the program will be described.

[overall structure]
FIG. 1 is a diagram showing an example of a configuration and a usage environment of a sales brokerage server 200 using a product evaluation device and a brand dictionary creating device 300. The sales brokerage server 200 is a device that is accessed by the terminal device 100 to mediate the sale of a product or service (hereinafter, product or the like), and is a device that provides a so-called shopping site or auction site. Here, the "site" includes a website played by a browser and an application screen displayed by an application program. The operator of the sales brokerage server 200 is provided with product data by the seller of the product. In the terminal device 100, a UA (User Agent) 110 such as a browser or an application program is activated. The UA 110 displays the sales screen provided by the sales intermediary server 200, and transmits a request corresponding to the input operation made by the user of the terminal device 100 to the sales intermediary server 200. The sales brokerage server 200 decides to sell the product or the like in response to the request from the UA 110.

Further, the sales brokerage server 200 is provided with a brand dictionary created by the brand dictionary creation device 300. A brand dictionary is an electronic dictionary that collects brand names that should be posted on the introduction screen of a certain product in principle. In the figure, the sales brokerage server 200 and the brand dictionary creation device 300 are shown as separate bodies, but these may be integrated and realized by one hardware.

The terminal device 100, the sales brokerage server 200, and the brand dictionary creation device 300 communicate with each other via the network NW. These components include a network card for connecting to the network NW, a wireless communication module, and the like. The network NW includes, for example, the Internet, a WAN (Wide Area Network), a LAN (Local Area Network), a wireless base station, a provider terminal, a dedicated line, and the like.

[Sales brokerage server]
The sales brokerage server 200 includes, for example, a sales screen providing unit 210, a ranking processing unit 212, a log collecting unit 214, a target word acquisition unit 230, a brand name extraction unit 232, and a caution target product monitoring unit 234. Be prepared. These components are realized by, for example, a processor such as a CPU (Central Processing Unit) executing a program (software). In addition, some or all of these components are realized by hardware such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit). It may be done, or it may be realized by the collaboration of software and hardware.

Further, the sales mediation server 200 is stored in a storage device such as a RAM (Random Access Memory), an HDD (Hard Disk Drive), a flash memory, and a ROM (Read Only Memory), and includes product data 220, ranking data 222, log information 224, and the like. It stores data such as the brand dictionary 240. These data may be stored in NAS (Network Attached Storage) or the like that can be accessed by the sales brokerage server 200 via the network NW.

The sales screen providing unit 210 refers to the product data 220 and the ranking data 222, and provides the product sales screen to the terminal device 100. FIG. 2 is a diagram showing an example of the sales screen IM1 of a product or the like. The sales screen IM is provided with a query input field A1 for inputting a query for searching for a product or the like, and a search button B1 for executing a search with the input query. When a query is input to the query input field A1 and the search button B1 is operated, the screen transitions to the search result display screen IM2. Some queries are only one word, and some queries include multiple words in which multiple child queries are combined with spaces.

FIG. 3 is a diagram showing an example of the search result display screen IM2. The search result display screen IM2 includes a search result display field A2. In the search result display field A2, images and explanations of products and the like are displayed side by side in the ranking order for each query determined by the ranking processing unit 212. When an image or description of one product or the like is operated on the search result display screen IM, the screen transitions to the product or the like display screen IM3.

FIG. 4 is a diagram showing an example of a product or the like display screen IM3. The product or the like display screen IM3 includes a product or the like image display column A3-1, a title column A3-2, a product or the like explanation column A3-3, and the like. In the title column A3-2 and the product description column A3-3, the manufacturer, the material of the product, the use, and other contents are posted. Such a content is a normal product or the like display screen IM3, but the product or the like display screen in which excessive SEO is performed shows a different mode.

FIG. 5 is an example of the product display screen IM4 in which excessive SEO is performed. On the product display screen IM4, although the manufacturer is "○○ company", the product explanation column A4-3 shows unrelated companies A, B, and C (all of the same category as company XX). Brand name) is posted. As a result, as a result of performing a search by designating "Company A" as a query, this product or the like may hit the search even though it is not a product or the like of Company A. In order to avoid or suppress the listing of such products, the sales brokerage server 200 executes a specific process of the product or the like to be noted using the brand dictionary 240. This will be described later.

Returning to the story, the sales screen providing unit 210 generates various display screens illustrated in FIGS. 2 to 5 based on the product data 220. FIG. 6 is a diagram showing an example of the contents of the product or the like data 220. The product data 220 includes a product category, a product image, a title, a product description, a price (standard price or bid price in the case of an auction), shipping conditions, and the like for the product ID, which is identification information of the product. The data of is associated.

The ranking processing unit 212 performs ranking processing on each of the queries input in the past according to various criteria, and generates ranking data 222. FIG. 7 is a diagram showing an example of the contents of the ranking data 222. The ranking data 222 is, for example, associated with a query by ranking of product IDs. The ranking data 222 is, for example, based on the log information collected by the log collection unit 214, in the order of the number of clicks, the order of the number of purchases, the order of high CVR (Conversion Ratio), the order of the number of favorites registered, or The ranking is high in ascending order of price or in order of score combining these.

The log collecting unit 214 stores information including a query input in the terminal device 100, such as an action history in which pages requested by the terminal device 100 are arranged in chronological order, as log information 224.

The target word / phrase acquisition unit 230 acquires the target word / phrase for each product or the like indicated by the product or the like ID among the data included in the product or the like data 220. The target phrase may be arbitrarily defined, but for example, the phrase included in the title and the description of the product or the like is set as the target phrase.

The brand name extraction unit 232 refers to the brand dictionary 240 and extracts the brand name included in the target word / phrase acquired by the target word / phrase acquisition unit 230. More specifically, the brand name extraction unit 232 extracts a brand name that matches the category of the original product or the like from which the target phrase was acquired. FIG. 8 is a diagram showing an example of the contents of the brand dictionary 240. The brand dictionary 240 is, for example, a list of brand names for each category such as a product.

The brand name extraction unit 232 constructs an automaton that performs pattern matching from the dictionary by performing a common prefix search (Common Prefix Search) such as the AhoCorasick method in order to utilize the search results that have already been performed, and inputs text. Achieve a linear calculation time for.

The attention target product monitoring unit 234 specifies the caution target product or the like based on the number of brand names extracted by the brand name extraction unit 232. For example, the caution target product monitoring unit 234 specifies a product or the like whose target phrase includes a brand name of a predetermined number of k or more as a caution target product or the like. For example, k = 2.

Then, the attention target product monitoring unit 234 displays the product ID on the display device or pays attention to the seller in order to (1) pay attention to the seller regarding the specified product to be watched. Is generated and transmitted, (2) the ranking of the product or the like in the ranking data 222 is lowered, or (3) the product or the like is deleted from the product or the like data 220 (the listing is canceled).

The target product monitoring unit 234 may perform the processes (1) to (3) stepwise according to the number of brand names specified by the brand name extraction unit 232. For example, processing (1) for products, etc. in which two brand names are included in the target phrase, and processing (2) for products, etc. in which three brand names are included in the target phrase, four or more. The process (3) may be performed for a product or the like whose brand name is included in the target phrase.

In addition, the caution target product monitoring unit 234 counts the number of products, etc. that correspond to the caution target products, etc. for each seller, and according to the number of products, etc. that correspond to the caution target products, etc., the above (1) to The process of (3) may be performed step by step. For example, the process of (1) is performed on the seller whose number of products, etc. corresponding to the caution target product, etc. is equal to or more than the first threshold C1 and less than the second threshold C2, and the number of products, etc. corresponding to the caution target product, etc. is the second. The process of (2) is applied to the seller's attention target products, etc. of the 2nd threshold C2 or more and less than the 3rd threshold C3, and the attention target of the seller whose number of products, etc. corresponding to the caution target products, etc. is the 3rd threshold C3 or more. The process (3) may be performed on products (or all products) (C1 <C2 <C3).

FIG. 9 is a flowchart showing an example of a processing flow executed by the target word acquisition unit 230, the brand name extraction unit 232, and the attention target product monitoring unit 234 of the sales brokerage server 200. The processing of this flowchart is started at an arbitrary timing.

First, the target word / phrase selection unit 230 selects one product or the like from the product or the like data 220 (S100) and acquires the target word or phrase (S102). Next, the brand name extraction unit 232 extracts the brand name from the target words and phrases acquired in S102 (S104).

Next, the attention target product monitoring unit 234 determines whether or not the brand name extracted in S104 is a predetermined number k or more (S106). When the brand name extracted in S104 is a predetermined number of k or more, the caution target product monitoring unit 234 specifies the product or the like selected in S100 as the caution target product (S108).

Next, the target word / phrase selection unit 230 determines whether or not all the products or the like to be processed this time are selected from the product or the like data 220 (S110). If all products and the like are not selected, the process is returned to S100, and if all products and the like are selected, the process of this flowchart ends. The loop processing illustrated in FIG. 9 may be executed in parallel by parallel computing.

In this way, according to the sales brokerage server 200 (product evaluation device), the target word acquisition unit 230 for acquiring the target word included in the introduction screen of the product or the like sold via the network NW and the brand name are described. Based on the number of brand name extraction unit 232 that extracts the brand name included in the target phrase and the brand name extraction unit 232, and the number of brand names extracted by the brand name extraction unit 232, the products to be noted are selected. By providing the monitoring unit 234 for the products to be watched, the products to be watched can be efficiently found. An example of the "product evaluation device" is a combination of the target word acquisition unit 230, the brand name extraction unit 232, and the caution target product monitoring unit 234.

In the above explanation, the brand dictionary 240 according to the category of the product or the like is used to specify the product or the like to be noted, but the product or the like to be noted is specified by using all the brand dictionaries 240 without specifying the category. May be specified.

[Brand dictionary creation device]
Hereinafter, the brand dictionary creating device 300 for creating the brand dictionary 240 will be described. The brand dictionary creation device 300 includes, for example, a log acquisition unit 310, a narrowing-down unit 312, and a creation unit 314. These components are realized by, for example, a processor such as a CPU executing a program (software). In addition, some or all of these components may be realized by hardware such as LSI, ASIC, FPGA, GPU, or may be realized by collaboration between software and hardware.

Further, the brand dictionary creating device 300 stores data such as a query log 320 and a brand dictionary 322 being created in a storage device such as a RAM, an HDD, a flash memory, or a ROM. These data may be stored in NAS or the like accessible by the brand dictionary creation device 300 via the network NW.

The log acquisition unit 310 acquires information related to the query from the log information 224 accumulated by the sales intermediary server 220, and stores the information as the query log 320 in the storage device. Each query included in the query log 320 is accompanied by a product or the like displayed as a result of searching using the query and its category. The categories such as products may be assigned hierarchically, but here, the categories of the hierarchy corresponding to the categories of the brand dictionary will be described.

Based on each query included in the query log 320, the narrowing unit 312 narrows down the text included in the query according to a rule according to a category such as a product. The text in the embodiment is interpreted as a semantic unit by a predetermined logic such as morphological analysis. Further, each of the child queries separated by a space or the like may be used. Further, the text may be a combination of a plurality of child queries. Depending on the category, the narrowing-down unit 312 may add a limitation to the elements constituting the text, such as limiting the text to katakana of three or more characters.

Here, when the narrowing down unit 312 creates a brand dictionary of a certain category, the products and the like corresponding to the category (hereinafter, the target category) are most attached (that is, displayed as a result of the query search). Queries (where the majority of products, etc. correspond to that category) are targeted for narrowing down. That is, the query targeted by the narrowing unit 312 (hereinafter referred to as the target query) is a query that is presumed to be intended for the target category.

FIG. 10 is a diagram for explaining the content of processing by the narrowing-down unit 312. For each text included in the target query, the narrowing-down unit 312 appears s0 as the number of times it is input as a single query, s1 as the number of times it appears first among the queries including a plurality of texts, and appears second. The number of times is counted as s2, and the number of times of appearance after the third is counted as s3. In addition, the narrowing unit 312 also counts the number of days in which each text appears in the target period (for example, one year). In the example of FIG. 10, with respect to the text "aaa", since it appears once at the first time and appears once alone, 1 is given to s0 and s1 respectively.

In this way, when the target query of the target period is counted for each text and each appearance position, the narrowing unit 312 narrows down based on various conditions shown below.

(Condition 1)
As shown in the conditional expression (1), the narrowing-down unit 312 may leave the text whose appearance days have reached the reference days and exclude the texts that do not. In the formula, D is the number of days in the target period, and Th1 is a threshold value (0 <Th1 <1, more preferably Th1 is about 0.8 to 0.9). As a result, the narrowing-down unit 312 can exclude a large number of searched texts (buzzwords) related to buzzwords.
(Number of days of appearance) ≧ D × Th1… (1)

(Condition 2)
Further, as shown in the conditional expression (2), the narrowing-down unit 312 may leave the text in which the total number of occurrences reaches the reference value and exclude the text in which the total number of appearances does not reach the reference value. In the formula, Th2 is a threshold value (Th2 = about 10 to 100).
(S0 + s1 + s2 + s3) ≧ D × Th2… (2)

(Condition 3)
As shown in the conditional expression (3), the narrowing-down unit 312 may leave a lot of texts that appear at a specific position and exclude the texts that do not. In the formula, Th3 is a threshold (0 <Th3 <1, more preferably Th3 is 0.5 or more).
{S1 / (s0 + s1 + s2 + s3)} ≧ Th3… (3)

The condition represented by the formula (3) can be arbitrarily changed depending on the category. In particular, the molecule of formula (3) may be modified according to the category. Equation (3) focuses on the frequency with which a query containing multiple texts appears first, which is a rule suitable for a fashion category. This is because it is known in advance that when performing a query search in a fashion category, the brand name is often entered first.

The creation unit 314 collects the texts remaining as a result of the narrowing down by the narrowing down unit 312, and creates a brand dictionary. The brand dictionary is provided to, for example, the sales brokerage server 200.

Hereinafter, the flow of processing executed by the brand dictionary creation device 300 will be described with reference to a flowchart. FIG. 11 is a flowchart showing an example of the flow of processing executed by the brand dictionary creating device 300.

First, the brand dictionary creation device 300 determines the target category (S200). The target category is determined based on, for example, an instruction from the outside (including the operator). Next, the narrowing unit 312 extracts the query corresponding to the target category from the query log 320 (S202).

Next, the narrowing unit 312 selects one query extracted in S202 (S204) and divides the query into text (S206). Next, the narrowing-down unit 312 counts up the number of appearances and the number of appearance days for each appearance position for each text (S208).

Then, the narrowing-down unit 312 determines whether or not all the queries have been selected in S204 (S210). If all queries are not selected, the process is returned to S204.

When all the queries are selected, the narrowing unit 312 narrows down the text based on the various conditions described above (S212). The creating unit 314 includes the text remaining as a result of the narrowing down by the narrowing down unit 312 as a brand name in the dictionary of the target category (S214). The loop processing of S204 to S210 illustrated in FIG. 11 may be executed in parallel by parallel computing.

The method of dividing a query into text is not limited to the above method. For example, it is conceivable that "aaaa bbb" included in the query "aaaa bbb ccc" has become widespread as a brand name. On the other hand, the brand dictionary creating device 300 simply performed morphological analysis and the like from "aaaa bbb ccc" to "aaa", "bbbb", "ccc", "aaaa bbb", and "bbbb ccc". In some cases, it may be treated as one text including words that become two or more texts. In this case, for example, when both "aaa" and "aaa bbb" have passed through the narrowing down and remain, the one with the larger value on the left side of the conditional expression (3) may be adopted and the smaller one may be excluded. This is because when "aaaa bbb" has a higher public awareness than "aaaa", it is better to register "aaaa bbb" to detect excessive SEO appropriately.

According to the brand dictionary creation device 300 described above, the log acquisition unit 310 that acquires a set of queries input to the sales screen for selling products and the like using the network NW, and the text included in the target query. By providing a narrowing-down unit 312 that narrows down the items according to the rules corresponding to the categories of the products and the like, and a creating unit 314 that creates a brand dictionary for each category based on the narrowing-down result by the narrowing-down unit 312. You can create an appropriate brand dictionary for each.

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

100 Terminal device 110 UA
200 Sales brokerage server 210 Sales screen providing unit 212 Ranking processing unit 214 Log collection unit 220 Product data 222 Ranking data 224 Log information 230 Target word acquisition unit 232 Brand name extraction unit 234 Attention Target product monitoring unit 240 Brand dictionary 300 Brand dictionary Creation device 310 Log acquisition unit 312 Narrowing down unit 314 Creation unit 320 Query log 322 Brand dictionary (under construction)
NW network

Claims (6)

The acquisition department that acquires the target words and phrases included in the introduction screen of products or services (hereinafter referred to as products, etc.) sold via the network.
A brand name extraction unit that extracts the brand name included in the target phrase by referring to the brand dictionary that describes the brand name, and
It is equipped with a monitoring unit that identifies products and the like to be watched based on the number of brand names extracted by the brand name extraction unit.
The monitoring unit performs stepwise processing on the product or the like to be noted based on the magnitude relationship between the number of brand names extracted by the brand name extraction unit and a threshold value of at least 2 or more.
Product evaluation device. The acquisition department that acquires the target words and phrases included in the introduction screen of products or services (hereinafter referred to as products, etc.) sold via the network.
A brand name extraction unit that extracts the brand name included in the target phrase by referring to the brand dictionary that describes the brand name, and
It is equipped with a monitoring unit that identifies products and the like to be watched based on the number of brand names extracted by the brand name extraction unit.
The monitoring unit is stepwise for the seller who has put up the product or the like to be noted based on the magnitude relationship between the number of brand names extracted by the brand name extraction unit and a threshold value of at least 2 or more. Do the processing,
Product evaluation device. The brand dictionary is prepared for each category such as products.
The brand name extraction unit refers to a dictionary corresponding to a category given in advance to the product or the like from which the target phrase has been acquired, and extracts the brand name included in the target phrase.
The product evaluation device according to claim 1 or 2. When the number of brand names extracted by the brand name extraction unit is equal to or greater than a predetermined number, the monitoring unit identifies the acquisition source product or the like from which the target phrase has been acquired as the caution target product or the like.
The product evaluation device according to any one of claims 1 to 3. The computer
Acquire the target words and phrases included in the introduction screen of products or services (hereinafter referred to as products, etc.) sold via the network.
Refer to the brand dictionary that describes the brand name, extract the brand name included in the target phrase, and then
Based on the number of the extracted brand name, to identify the product or the like of the attention object,
Based on the magnitude relationship between the number of extracted brand names and a threshold value of at least 2 or more, the product or the like to be noted is subjected to stepwise processing.
Product evaluation method. On the computer
Get the target words and phrases included in the introduction screen of the product or service (hereinafter, product, etc.) sold via the network.
Refer to the brand dictionary that describes the brand name, extract the brand name included in the target phrase, and then
Based on the number of brand names extracted, the products to be noted are specified .
Based on the magnitude relationship between the number of extracted brand names and a threshold value of at least 2 or more, the product or the like to be noted is subjected to stepwise processing.
program.

Images