KR20230026669A - Method and system for writing reviews considering synerge effect between a plurality of products and analyzing synerge effect with respect to the reviews using ai model - Google Patents

Abstract

Provided is a review management method for outputting a first analysis result and a second analysis result by using an analysis model to analyze synergy reviews written for a combination of at least two products among a plurality of products, and according to the analysis by an analysis model. The analysis model is trained to output a first analysis result including positive or negative values for a combination of the products and a second analysis result including attribute information associated with the first analysis result.

Description

Synergy effect analysis method and system for reviews based on review writing and AI model considering synergy between multiple products AI MODEL}

The following description relates to a technique for writing a review by a user and managing the written review, and relates to a review writing method considering a synergy effect between a plurality of products and a synergy effect analysis method between products for the written review.

A user who purchases a product or uses a service may leave a review on the product he/she has purchased or the service used. A review written by a user includes the actual purchaser/user's experience of the corresponding product or service, and can be an important reference material for other users to consider purchasing the product or using the service.

It is common for product reviews to be written as reviews of product purchases by users who have purchased the product, and thus reflect only the attributes of the corresponding product.

However, depending on the type and characteristics of the product, a product is often used in combination with one or more other products. In particular, in the case of products such as cosmetics, cosmetics (e.g., skin/lotion/cream) used in one step are generally used in combination with other cosmetics (e.g., base/tone) used in another step, and the combination of used cosmetics Depending on it, the effect may be different. In the case of a 'general review' written simply as a review on product purchase, it does not properly reflect the synergy effect of these products being used in combination.

Therefore, there is a need for a technique that allows a review to be written considering the synergistic effect of products being used in combination, and to analyze the review written in consideration of the synergistic effect and use it as recommendation information.

For example, Korean Utility Model Publication No. 20-2011-0003005 describes a device for evaluating a person's skin or hair and recommending an appropriate personal care product based on the evaluation.

A 'synergy review (or affinity review) that considers the synergistic effect of a combination of products for a combination of multiple products that can be used in combination with each other, rather than just a user writing a review that includes an evaluation of one product. It provides a review management method and a review management platform (system) that enables users to create 'reviews'.

Using an artificial intelligence (AI) model, by analyzing synergy reviews written for a combination of at least two products, the positive or negative degree of the combination of the products is numerically estimated, and the estimated positive or negative degree Provided is a method of determining the attribute information that has influenced and outputting it as recommendation information.

In one aspect, in a review management method performed by a computer system, using an analysis model, analyzing synergy reviews created for a combination of at least two products among a plurality of products - the analysis model is the product According to learning to output a first analysis result including a positive or negative degree of a combination of and a second analysis result including attribute information associated with the first analysis result and analysis by the analysis model, the A review management method comprising outputting the first analysis result and the second analysis result is provided.

Each synergy review of the synergy reviews is information input through a synergy review writing user interface (UI) provided to the user terminal of the user who wrote each synergy review, and the first product and the second product constituting the combination of the products. It may include information about and text for evaluating a combination of the first product and the second product.

The analysis model includes a first analysis model learned to output the first analysis result for the combination of products by analyzing the text included in the synergy reviews, and the first analysis result is for the combination of products. Data that quantifies positive or negative degrees may be included.

The first analysis model is learned by the following steps to output the first analysis result, and the steps include first tokens composed of words indicating affirmation from texts of reviews pre-written by users. and extracting second tokens composed of words indicating negation. By removing tokens of overlapping words and tokens of words corresponding to attributes of products from the first tokens and the second tokens, the first tokens are removed. The method may include generating a training data set for an analysis model, and training the first analysis model to estimate a positive or negative rating for a review written by a user using the training data set. .

The analysis model includes a second analysis model learned to output the second analysis result for the combination of the products based on the text included in the synergy reviews and the first analysis result, and the second analysis result is , as the attribute information, at least one attribute keyword indicating a attribute of a product associated with the positive or negative degree indicated by the first analysis result may be included.

The attribute keyword may include a keyword representing a product attribute with a maximum contribution to a positive or negative degree of the combination of the products indicated by the first analysis result.

The second analysis model is learned by the following steps to output the second analysis result, and the steps include extracting words representing attributes of products from reviews pre-written by users; Among the extracted words, to determine a word indicating a property of a product associated with a positive or negative degree of each review of the pre-written reviews according to an analysis result by the first analysis model on the text of the pre-written reviews 2 may include training the analysis model.

The first analysis result for the combination of products is determined based on the positive or negative degree estimated for each synergy review of the synergy reviews written for the combination of products, and the above for the combination of products The second analysis result may be determined based on an attribute keyword indicating a product attribute associated with the positive or negative degree estimated for each synergy review.

The outputting may include outputting, as a result of the first analysis, a comprehensive score obtained by digitizing the positive or negative degree of the combination of products, and the attribute keyword as a result of the second analysis, and the product indicated by the attribute keyword. It may include further outputting attribute ratings associated with attributes of .

The combination of products includes at least a first product and a second product, and the outputting of the product includes, in response to a search request for the first product received from a user terminal, the first product in response to the search request. The method may include providing information on the second product, the first analysis result, and the second analysis result to the user terminal as recommendation information on the first product.

For a combination of a plurality of products that can be used in combination with each other, a user interface (or review writing tool) may be provided that enables a user to create a 'synergy review' considering the synergistic effect of the combination of products.

By analyzing the synergy reviews written for a combination of at least two products using an artificial intelligence (AI) model, the compatibility between products used in combination with each other (i.e., whether the combination has a positive synergistic effect or a negative synergistic effect) synergistic effect) is numerically determined, and it is possible to easily grasp which attribute of the product has the greatest influence on the determined compatibility.

By recommending other products having a positive synergistic effect with the searched product to a user searching for a product and presenting the synergistic effect through numerical values and attribute keywords, more effective product recommendation to the user may be possible.

1 illustrates a method of providing results according to synergy review analysis to a user terminal, according to an embodiment.
2 is a block diagram showing a user's terminal for creating synergy reviews and outputting analysis results for synergy reviews and a server for providing analysis results for synergy reviews, according to an embodiment.
3 is a flow diagram illustrating a method for managing reviews (ie, generic reviews and synergy reviews created for a combination of products), according to one embodiment.
4 is a flowchart illustrating a method of constructing a first analysis model according to an example.
5 is a flowchart illustrating a method of constructing a second analysis model according to an example.
6 is a flowchart illustrating a method of providing recommendation information including a synergistic effect according to a combination of products according to a search request from a user according to an example.
7 illustrates a method of outputting analysis results for synergy reviews using an analysis model according to an example.
8A to 8C show a user interface (UI) for creating a synergy review according to an example.
9 illustrates a method of comparing a first product and a second product according to an example.
10 illustrates a method of displaying a synergy review on a user terminal according to an example.
11A and 11B show a method of constructing a first analysis model and numerically estimating a synergistic effect (positive or negative degree) of a combination of products through the first analysis model, according to an example.
12A and 12B illustrate a method for establishing a second analysis model and determining a factor associated with a synergistic effect (positivity or negativity) for a combination of products through the second analysis model, according to an example. indicate
13A and 13B show a method of building an AI-based analysis model according to an example.
14 illustrates a method of constructing an analysis model for synergy review analysis in a system for synergy review management according to an example.
15 illustrates a method of configuring a system for management of synergy reviews according to an example.

Hereinafter, an embodiment will be described in detail with reference to the accompanying drawings.

1 illustrates a method of providing results according to synergy review analysis to a user terminal, according to an embodiment.

In FIG. 1, a server 100 performing a review management method of an embodiment and a terminal 110 used by a user are shown.

The user may be a user who wants to purchase a product through an online shopping mall or e-commerce, or may be a user who uses a review for a purchased product. The user may search for a product through an online shopping mall or e-commerce using the user terminal 110 possessed by the user, purchase the searched product, and write a review evaluating the purchased product.

The server 100 may be a server that provides a user interface (UI) for inputting information for writing a review to the user terminal 110 so that the user can write a review, and manages the review written by the user. The server 100 may provide recommendation information related to a product for which a search is requested according to a product search request from the user terminal 110 .

In the embodiment, the server 100, in providing a UI for inputting information for the user to write a review to the user terminal 110, provides a UI that allows the user to write a review for a combination of at least two products. can do. The user may write a review evaluating a combination of at least two products through the UI provided to the user terminal 110 . For example, the user may write a review on the synergistic effect of using cosmetics A and B together through the UI provided on the user terminal 110 . The synergistic effect may indicate a positive or negative effect that is multiplied as a plurality of products are combined and used.

In the present disclosure, a review evaluating a combination of these two products is defined as a "synergy review (or affinity review)". Meanwhile, a review evaluating one product is defined as a "general review". "Review" may be used to encompass synergy reviews and general reviews.

In other words, the server 100 may provide the user terminal 110 with a UI for inputting information for the user to create a synergy review.

The server 100 may analyze synergy reviews written by users and provide the analysis result to the user terminal 110 .

The server 100 may provide this analysis result as recommendation information (for products). For example, the server 100 may provide the user terminal 110 with an analysis result of analyzing a synergistic effect of a combination of a product searched for by a user (eg, product A) and another product (product B). there is. The other product may be at least one product that can be used in combination with the product searched for by the user. Such a synergistic effect may be provided to the user terminal 110 as an analysis result according to analysis of synergy reviews for a combination of the product and other products (ie, a combination of product A and product B) written by other users.

A method of creating and analyzing a synergy review and providing specific analysis results will be described in more detail with reference to FIGS. 2 to 15 to be described later.

In the detailed description to be described later, at least some of the operations performed by the server 100, which is a computer system that performs the review management method of the embodiment, may be performed by the user terminal 110. Redundant descriptions in this regard will be omitted.

2 is a block diagram showing a user's terminal for creating synergy reviews and outputting analysis results for synergy reviews and a server for providing analysis results for synergy reviews, according to an embodiment.

The server 100 is a device that provides a UI for writing user reviews (general reviews and synergy reviews) to the user terminal 110 and generates analysis results by analyzing reviews written by users, and includes at least one computing device. can include As shown, the server 100 may include a processor 140 and a communication unit 130 . Although not shown, the server 100 may further include a memory and an input/output interface. Accordingly, the server 100 may include a computer system that performs the review management method of the embodiment. Hereinafter, the server 100 may be referred to as the computer system 100 .

Computer system 100 may constitute a platform for writing and analyzing reviews, or may be at least part of such a platform.

For example, the computer system 100 provides an analysis result of synergy review analysis to a user accessing a specific website (eg, shopping mall, etc.) or a specific service (eg, service provided through a mobile application). and the analysis result may be provided as recommendation information according to the user's search request.

The computer system 100 may provide a UI for writing a review for the terminal 110, may provide analysis results obtained by analyzing synergy reviews, store data necessary for providing them, and process required operations. can Computer system 100 may include or communicate with a database for storing the necessary data. Computer system 100 may be composed of a plurality of servers or other computing devices unlike those shown.

The memory of the computer system 100 is a computer-readable recording medium and may include a random access memory (RAM), a read only memory (ROM), and a permanent mass storage device such as a disk drive. . Here, the ROM and the non-perishable mass storage device may be separated from the memory and included as a separate permanent storage device. Also, an operating system and at least one program code may be stored in the memory. These software components may be loaded from a computer-readable recording medium separate from memory. The separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD/CD-ROM drive, and a memory card. In another embodiment, the software components may be loaded into the memory through the communication unit 130 rather than a computer-readable recording medium.

The processor 140 may be configured to process commands of a computer program by performing basic arithmetic, logic, and input/output operations. Instructions may be provided to the processor 140 by memory or the communication unit 130 . For example, processor 140 may be configured to execute instructions received according to program code loaded into memory.

The processor 140 may include software and/or hardware module(s) as components not shown. A review management method to be described later may be executed through the components of the processor 140 .

The communication unit 130 may be a component for the computer system 100 to communicate with other devices (such as the user terminal 110 or other servers). In other words, the communication unit 130 is a hardware module or network such as an antenna, data bus, network interface card, network interface chip, and networking interface port of the computer system 100 that transmits/receives data and/or information to/from other devices. It may be a software module such as a device driver or networking program.

The input/output interface may be a means for interfacing with an input device such as a keyboard or mouse and an output device such as a display or speaker.

Also, in other embodiments, computer system 100 may include more components than those shown.

The user terminal 110 is a terminal used to receive and output information/data provided from the computer system 100 and write a user review, and may be a terminal in which applications or programs for executing these tasks are installed.

In the illustrated embodiment, the user may write a review through the UI provided from the computer system 100 through the user terminal 110, and may receive and output analysis results for synergy reviews from the computer system 100. there is.

The user terminal 110 is an electronic device, and may be a smartphone or similar device capable of installing and executing an application or program for writing reviews and outputting analysis results for synergy reviews. In addition, the user user terminal 110 may be a personal computer (PC), a laptop computer, a laptop computer, a tablet, an Internet Of Things (IoT) device, or a wearable computer. etc.

The user terminal 110 may include a communication unit 230 , a processor 220 and a display unit 210 .

The communication unit 230 may be a device for the user terminal 110 to communicate with other devices (eg, other terminals or the computer system 100). In other words, the communication unit 230 is a hardware module or network device driver such as a network interface card, network interface chip, and networking interface port of the user terminal 110 that transmits/receives data and/or information to/from other electronic devices or servers ( driver) or a software module such as a networking program.

The processor 220 may manage components of the user terminal 110 and may execute a program or application used by the user terminal 110 . For example, the user terminal 110 may be configured to process analysis results for synergy reviews received or information/data input through a UI provided for writing a review.

In addition, the processor 220 may process operations required for execution of programs or applications and processing of data. The processor 220 may be at least one processor of the user terminal 110 or at least one core within the processor.

The display unit 210 may include a display as a device for outputting information or data input by a user. The display unit 210 may output various types of information including a UI for writing a review, an analysis result of synergy reviews, and a product search result by a user. The display unit 210 may include a touch panel. At this time, the display unit 210 can also function as an input unit capable of receiving an input from a user.

Meanwhile, although not shown, the user terminal 110 may include a storage unit as a device for storing data or information. Storage may include any memory or storage device. The storage unit may store programs or applications executed by the processor 220 or information/data related to writing reviews and outputting analysis results for synergy reviews.

A review management method (that is, a method of creating synergy reviews and analyzing synergy reviews) by the user terminal 110 and the computer system 100 will be described in more detail with reference to FIGS. 3 to 15 to be described later.

Since descriptions of the technical features described above with reference to FIG. 1 may be applied as they are to FIG. 2 , duplicate descriptions will be omitted.

In the detailed description to be described later, steps and operations of the review management method are described as being performed by the computer system 100 for convenience of description, but may be performed by components included in the computer system 100.

Also, at least some of the steps or operations of the computer system 100 may be performed by the user terminal 110 . Redundant descriptions in this regard will be omitted.

3 is a flow diagram illustrating a method for managing reviews (ie, generic reviews and synergy reviews created for a combination of products), according to one embodiment.

The review management method, which will be described later, will cover a method in which the computer system 100 provides a UI so that a user can write a synergy review, and a method in which the computer system 100 analyzes reviews written by users. can

In step 310, the computer system 100 may provide a review writing user interface (UI) for writing a review. The computer system 100 may allow a user to write a review by providing a review writing user interface (UI) to the user terminal 110 .

For example, the computer system 100 may allow the user to write a synergy review by providing the user terminal 110 with a synergy review writing user interface (UI) for writing the aforementioned synergy review.

Each synergy review of synergy reviews written by users managed by the computer system 100 (ie, subject to analysis by the computer system 100) is a user terminal 110 of the user who wrote each synergy review. ) Information input through the synergy review writing UI provided to, may include information on the first product and the second product constituting the combination of products and text evaluating the combination of the first product and the second product. . A combination of products input through the review writing UI (ie, product information for evaluation through reviews) may include three or more products.

In step 320, the computer system 100 may receive input information for writing a review (synergy review) from the user terminal 110. Input information for creating a synergy review may include information about the first product and the second product constituting the combination of products and text for evaluating the combination of the first product and the second product. When the evaluation target is a combination of three or more products, information on three or more products may be included in the input information.

In this regard, FIGS. 8A to 8C show a user interface (UI) for creating a synergy review, according to an example.

8a to 8c show a synergy review writing UI.

As shown, information on products (810, 820) to create a synergy review may be input through the user terminal 110. In the example, product A 810 is input, and a product to be combined with product A 810 can be selected or searched through a search synergy review product 820 button. As shown in FIG. 8B , product B 822 may be selected as a combination for product A 810 . Product A 810 and product B 822 are products that can be used in combination, and may be different cosmetics.

The synergy review writing UI may provide a UI for inputting a comprehensive rating for a combination of product A 810 and product B 822 (ie, a combination of products subject to evaluation). For example, the user may input a rating (or star rating) of 0 to 10 through selecting a star rating (comprehensive rating 830).

Meanwhile, the synergy review writing UI may provide a UI for inputting individual ratings for attributes of individual products of a combination of product A 810 and product B 822 . For example, the user selects a star score for each of skin trouble, spreadability, and staying power as attributes of product A 810 and product B 822, which are cosmetics, and scores from 0 to 10 for each attribute. can be entered (individual ratings (852 to 856)). The number of attributes for which individual ratings 852 to 856 can be input may be configured differently according to the type of product or a combination of products.

The synergy review writing UI may provide a UI for inputting text for evaluating the combination of product A 810 and product B 822 . As shown, as the user uses a combination of product A 810 and product B 822 through the corresponding UI, the user can input satisfactory points, regrettable points, honey tips (useful tips), etc. as text, and also , A photo associated with a combination of product A 810 and product B 822 may be uploaded.

As the preparation of the input items 810 to 856 is completed, the preparation of synergy review may be completed, and the completed synergy review may be managed by the computer system 100 . In this way, the computer system 100 may configure a comprehensive synergy review management platform that provides writing and analysis functions for synergy reviews.

Therefore, in the embodiment, the user can evaluate a combination of products by creating a "synergy review" considering not only the individual product purchased or used by the user, but also the product used in combination with the purchased or used product.

Meanwhile, a predetermined reward (point, etc.) may be provided to the user (user's account) who wrote the synergy review.

Below, a method of analyzing the created synergy reviews and outputting the analysis result (eg, to the user terminal 110, etc.) will be described in detail.

At step 330, the computer system 100 may manage synergy reviews written by users.

At step 332, the computer system 100 may analyze the synergy reviews written for a combination of at least two products of the plurality of products using the analysis model. The analysis model is an artificial intelligence (AI)-based analysis model, and may be an analysis model using an artificial neural network. For example, the analysis model may be a deep learning model that predicts/infers an analysis result by analyzing synergy reviews.

This analysis model can be trained to output a first analysis result including positive or negative values for a combination of products and a second analysis result including attribute information associated with the first analysis result. A positive or negative degree can indicate the compatibility (degree) of a combination of products. A 'attribute' may indicate a property or characteristic of a product or a combination of products.

In step 334, the computer system 100 may output a first analysis result and a second analysis result for synergy reviews according to the analysis by the analysis model. In other words, the computer system 100 may analyze the synergy reviews written by users and output, as a first analysis result, a degree indicating whether a combination of products represented by the synergy reviews is positive or negative, and furthermore, the first analysis result. (That is, the positive or negative degree) information on the attribute of the product associated with the product may be output as the second analysis result. The user checks the first analysis result and the second analysis result provided from the computer system 100 through the user terminal 110, so that a combination of specific products (eg, a combination of the product searched by the user and other products) is determined by the user. It is possible to check whether or not a positive synergistic effect is produced, and it is possible to easily grasp what attributes of a product are related to the synergistic effect.

Below, a method of outputting the first analysis result and the second analysis result will be described in detail.

The first analysis result and the second analysis result may be based on a result of analyzing text included in synergy reviews written by users using the analysis model.

For example, the analysis model may include a first analysis model learned to output a first analysis result for a combination of products by analyzing text included in synergy reviews. In addition, the analysis model may include a second analysis model learned to output a second analysis result for a combination of products based on the text included in the synergy reviews and the first analysis result. As such, the analysis model includes a first analysis model for providing a primary analysis result for synergy reviews and a second analysis result for synergy reviews using the analysis result by the first analysis model. A second analysis model may be included.

The first analysis result by the first analysis model may include data obtained by quantifying the positive or negative degree of a combination of products. The digitized data may represent the positive (or negative) of a combination of specific products as a value within a predetermined range. These values may be output to the user terminal 110 in the form of scores, ratings, star ratings, and the like determined as values within a predetermined range (eg, -1 to 1). Accordingly, the user can intuitively grasp the positive or negative degree of a combination of specific products by checking the first analysis result through the user terminal 110 .

On the other hand, the second analysis result by the second analysis model is attribute information associated with the first analysis result, and includes at least one attribute representing the attribute of the product associated with the positive or negative degree indicated by the first analysis result by the first analysis model. Can contain attribute keywords. The "attribute keyword" may be a keyword (representative keyword) indicating a product attribute. For example, in the case where a product constituting a combination of products is cosmetics, attribute keywords may be skin troubles, application properties, and holding power as keywords representing each attribute of cosmetics.

The second analysis result by the second analysis model may include an attribute keyword indicating a property of a product that affects the positive or negative degree of the combination of products determined by the first analysis result. For example, the attribute keyword included in the second analysis result may indicate a product attribute that contributes to a positive or negative degree for a combination of products represented by the first analysis result. This maximum contributing product attribute is the attribute of the product (i.e., representative attribute) that has the greatest influence (i.e., is the most dominant) on the positive or negative degree of the combination of products represented by the second analytical model. ) can mean.

Meanwhile, the second analysis result may include a plurality of attribute keywords, and each of the attribute keywords may correspond to a product attribute. The attribute keywords may be presented to the user terminal 110 in an order that affects the positive or negative degree of the combination of products indicated by the first analysis result. Accordingly, the user can intuitively grasp the attribute(s) of the product that affects the positive or negative degree of the combination of products.

In step 335, the computer system 100 may output, as a result of the first analysis by the first analysis model, a comprehensive score digitizing the positive or negative degree of the combination of products, and in step 336, The computer system 100 may output attribute keywords representing attributes of the product as a result of the second analysis by the second analysis model.

In step 337, the computer system 100 may further output an attribute score associated with the attribute of the product indicated by the attribute keyword. The attribute score is a numerical value of the positive or negative degree of the attribute of the product (ie, the attribute indicated by the attribute keyword) determined by the second analysis model, and the first analysis model is the second analysis result by the second analysis model. Based on , it is possible to estimate (determine) the attribute score.

In this regard, FIG. 7 shows a method of outputting analysis results for synergy reviews using an analysis model according to an example.

In FIG. 7 , an example of an analysis result for synergy reviews output to the user terminal 110 is shown.

As shown, the analysis result of the synergy reviews may be presented in the user terminal 110 as a "synergy effect between products" (meaning a synergy effect on a combination of specific products). The illustrated example shows analysis results for product A (705) and product B (710).

The comprehensive score 720 may be a result of a first analysis by a first analysis model, which quantifies the positive or negative degree of the combination of the products 705 and 710 .

Each of attribute 1 732 and attribute 2 734 may be an attribute keyword indicating a product attribute as a result of the second analysis by the second analysis model. Attribute 1 732 may indicate a positive or negative degree of a combination of the products 705 and 710 , that is, a property of the product that has the greatest influence on the overall score 720 . Attribute 2 (734) may indicate a product attribute that has the second largest influence on the overall score (720) after attribute 1 (732).

The attribute score 1 742 may be a numerical value of positive or negative with respect to the attribute 1 732 . Attribute score 2 (744) may be a numerical value of positive or negative with respect to attribute 2 (734).

Through the analysis results of the synergy reviews provided through the user terminal 110, the user can determine the overall synergistic effect of the combination of the product he/she has searched for and other products, and the attribute(s) of the product that causes the synergistic effect. ) can be easily identified.

Meanwhile, a detailed method for learning the first analysis model and the second analysis model, and a more detailed method for outputting the first analysis result and the second analysis result using the first analysis model and the second analysis model will be described later in FIG. 4 . It will be described in more detail with reference to FIG. 15 .

Descriptions of the technical features described above with reference to FIGS. 1 and 2 may be applied to FIGS. 3, 7, and 8 as they are, and thus duplicate descriptions are omitted.

4 is a flowchart illustrating a method of constructing a first analysis model according to an example. 5 is a flowchart illustrating a method of constructing a second analysis model according to an example.

Referring to FIGS. 4 and 5 , a method of learning/constructing an analysis model that provides analysis results for the aforementioned synergy reviews will be described in more detail.

The first analysis model may be learned to output the first analysis result, and may be learned by steps 410 to 440 shown in FIG. 4 . In other words, a learning model built according to learning by steps 410 to 440 may be used as the first analysis model.

In step 410, the computer system 100 may extract first tokens composed of words indicating positive and second tokens composed of words indicating negative from texts of reviews pre-written by users. Pre-written reviews may include synergy reviews and/or general reviews. For example, the computer system 100 performs natural language processing on texts including pre-written reviews, and performs morphological analysis, so that first tokens composed of words indicating positive and words indicating negative are converted from the text. The configured second tokens may be extracted.

As an example, a first set of tokens consisting of words representing affirmation may be [like, well, bought, shipped, done, satisfied, too, great, true, a lot] and a second set of words representing negative The set of tokens can be [not, shipped, bought, too, not, dissatisfied, excited, absent, dry]. In determining the set of tokens, characters determined to be unnecessary, such as surveys, may be excluded, and words whose frequency appears in reviews (from the viewpoint of improving the learning speed) less than or equal to a predetermined value may also be excluded.

Meanwhile, unnecessary special characters or broken characters (Korean) in the text of the review may also be removed through preprocessing.

At step 420, the computer system 100 may remove tokens of overlapping words and/or tokens of words corresponding to attributes of the product from the first tokens and the second tokens.

At step 430, the computer system 100 may generate a training data set for the first analysis model according to the processing at step 420. This training data set may be a data set for (initial) training of the first analysis model.

For example, the aforementioned first set of tokens [like, well, purchase, delivery, did, satisfaction, too, great, true, much] and the second set of tokens [not, delivery, purchase, not too, not, dissatisfaction, Exciting, absent, dry], overlapping tokens “delivery” and “purchase” may be excluded from the learning data. In addition, “dry”, which is a word indicating a property of a product, may be excluded from the learning data.

Meanwhile, in determining the first set of tokens composed of words indicating affirmation and the second set of tokens composed of words indicating negation, a pre-constructed positive word dictionary and/or a negative word dictionary may be used. That is, words to be included in the first set of tokens and the second set of tokens may be determined using the pre-constructed positive word dictionary and/or negative word dictionary.

Alternatively/additionally, in determining the first set of tokens and the second set of tokens, a separate classification model built to classify positive or negative words may also be used. That is to say, words to be included in the first set of tokens and the second set of tokens can be determined using this classification model.

Alternatively/additionally, in determining the first set of tokens and the second set of tokens, a rating input by the user at the time of writing the review may be used. For example, a first set of tokens may be determined from texts of reviews whose ratings input by the user are greater than or equal to a predetermined value, and a second set of tokens may be determined from texts of reviews whose ratings input by the user are less than a predetermined value. .

At step 440, the computer system 100 may use the generated training data set to train a first analytic model to estimate positive or negative ratings for reviews written by users (synergy reviews). . The first analysis model may estimate how positive or negative a review/synergy review including text is through learning using the generated training data set. The result of estimation can be determined as a digitized value.

Accordingly, a first learning model capable of estimating a positive or negative degree (corresponding to a synergistic effect) of a combination of products indicated by the corresponding synergy review may be constructed by analyzing synergy reviews written by users.

In relation to this, FIGS. 11A and 11B illustrate a method of constructing a first analysis model and numerically estimating a synergistic effect (positive or negative) of a combination of products through the first analysis model according to an example. indicate

As shown in FIG. 11A, the computer system 100 may generate a learning data set for learning the first analysis model corresponding to the AI analysis model (1110), and through machine learning using the learning data set ( 1120), and a first analysis model capable of estimating the degree of positivity or negativity representing the synergistic effect between products may be constructed (1130).

As illustrated, the set of tokens with positive meaning corresponding to the training data set can be [perfect, famous, good, :), tone up, simple, good, safe, life item, real, recommended, calm] and the set of tokens with negative meanings can be [herpes, dermatologist, lack, waste of money, no, caution, dead skin, trouble, ㅠㅠ]. Each of these positive meaning tokens and negative meaning tokens may be named an “emotion token”.

As shown in FIG. 11B, the computer system 100 uses the first analysis model built according to FIG. 11A to quantify the synergy for the combination of products represented by the synergy review (ie, the combination of Product A and Product B). effect can be output. For example, as a synergy review for a combination of products (that is, a combination of product A and product B) is created by the user (1140), the computer system 100 analyzes the corresponding synergy review using the first analysis model. In operation 1150, a measurement result (0.85) of the positive or negative degree of the combination of products (that is, the combination of product A and product B) may be output as an analysis result (first analysis result) (1160).

The computer system 100 measures the positive or negative degree of the combination of product A and product B (0.85 ) can be updated.

Meanwhile, the second analysis model may be learned to output the second analysis result, and may be learned by steps 510 and 520 shown in FIG. 5 . In other words, a learning model built according to learning by steps 510 and 520 may be used as the second analysis model.

In step 510, the computer system 100 may extract words representing product properties from reviews previously written by users. Pre-written reviews may include synergy reviews and/or general reviews. For example, the computer system 100 may extract tokens composed of words representing attributes of a product from the text by performing natural language processing and morphological analysis on texts included in pre-written reviews. These tokens may be named factor (or attribute) tokens. Words (tokens of words) representing attributes of the extracted product may become a learning data set for learning the second analysis model. Meanwhile, words whose frequency of appearance in reviews is less than a predetermined value (from the viewpoint of improving learning speed) may be excluded from the training data set.

In step 520, the computer system 100 determines whether each review of the previously written reviews is affirmative according to the result of analysis by the first analysis model on the text of the previously written reviews among the extracted words (ie, attribute tokens). A second analytic model may be trained to determine words representing attributes of products associated with degree or negativity. In other words, the second analytic model may be configured to determine a word representing an attribute of a product associated with the positive or negative rating of each review determined by the first analytic model. The second analysis model may determine a word representing an attribute that has the greatest influence on the positive or negative rating of a review, or may determine words representing high-ranking attributes having a large influence on the positive or negative rating of a review.

The computer system 100 may determine “attribute keywords” representing attributes associated with the word based on the word determined by the second analysis model.

The "attribute keyword" may be a word itself (representing a property) extracted from reviews analyzed by the analysis model, or may be a keyword previously set to indicate a property corresponding to the word. For example, when words such as "sticky", "sticky", and "sticky" are determined by the second analysis model as a column representing a property of a product associated with a positive or negative degree, "sticky" or "sticky" A word such as may be determined as an attribute keyword. As described above, these attribute keywords may be output as the second analysis result.

Accordingly, a second analysis model capable of estimating the attribute that greatly affected the positive or negative degree (corresponding to the synergistic effect) of a combination of specific products by analyzing synergy reviews written by users can be built. there is.

In relation to FIGS. 12A and 12B , according to an example, a second analysis model is built, and a factor associated with a synergistic effect (positive or negative) of a combination of products is determined through the second analysis model. show how to do it

As shown in FIG. 12A, the computer system 100 may generate a learning data set for learning a second analysis model corresponding to the AI analysis model (1210), and through machine learning using the learning data set ( 1220), and a second learning model capable of estimating factors related to the synergistic effect may be constructed (1230).

As illustrated, a set of factor tokens corresponding to the learning data set may be fragrance, trouble, adhesion, coverage, moisture, dryness, coloring, moisturizing power, mildness, color, and durability]. In addition, learning includes the result of analysis by the first analysis model and/or the above-mentioned emotion tokens - for example, a set of positive meaning tokens [perfect, famous, good good, :), tone up, simple, good, safe , life item, real, recommended, soothing], and a set of tokens with negative meanings [herpes, dermatology, lack, waste of money, no, caution, dead skin, trouble, ㅠㅠ] - can be used more. As shown in FIG. 12B, computer system 100 uses the second analysis model built according to FIG. Regarding the synergistic effect, factors (attributes) associated with the synergistic effect may be determined and output. For example, as a synergy review for a combination of products (that is, a combination of product A and product B) is written by the user (1240), the computer system 100 uses the first analysis model and the second analysis model to generate corresponding The synergy review may be analyzed (1250), and as a result of the first analysis, positive or negative for the combination of products represented by the corresponding synergy review is provided, but as a result of the second analysis, a factor (attribute) associated with the positive or negative degree Information may be provided (1160). For example, as shown, the attribute that most affected the positiveness of the combination of products (that is, the combination of product A and product B) was moistness, and the attribute that most affected the irregularity of the combination of products was stickiness. can be determined as “Moisture” and “stickiness” may be attribute keywords included in the aforementioned second analysis result.

The computer system 100 measures the positive or negative degree of the combination of product A and product B (0.85 ) may be updated, and an attribute keyword (ie, the second analysis result) associated with the updated measurement result may be determined.

The aforementioned analysis models (the first analysis model and the second analysis model) may include learning models built using Recurrent Neural Networks (RNNs)-based Long Short-Term Memory Models (LSTMs).

In this regard, FIGS. 13A and 13B show a method of building an AI-based analysis model according to an example.

Figure 13a shows the structure of an RNN-LSTM cell included in the analysis model, and Figure 13b shows the structure of an LSTM network included in the analysis model.

The illustrated analysis model may be configured to output desired output values (ie, the above-described first analysis result and second analysis result) using the aforementioned training data set as input data.

For example, the second analysis model of the embodiment may be built using an LSTM model based on token data of the aforementioned training data set and a learning review data set corresponding to an analysis result of the first analysis model. The second analysis model extracts tokens of attribute words from pre-written reviews, analyzes the corresponding tokens, and has the most influence on the first analysis result (positivity or negativity) (relativity) by the first analysis model. The quasi-properties can be determined to generate a list of properties. The TextRank keyword extraction algorithm can be used to extract tokens and determine attributes.

Meanwhile, learning of the analysis model described above may be performed through a server separate from the computer system 100 . Thus, learning of the analysis model may not affect synergy review management and service provision provided by the computer system 100 . The analytics model can provide a Restful API.

Below, an example of a method for providing a first analysis result and a second analysis result, which are analysis results for synergy reviews, will be described in more detail.

The first analysis result of the combination of specific products by the first analysis model may be determined based on the positive or negative degree estimated for each synergy review of synergy reviews written for the combination of corresponding products.

That is, the first analysis model may analyze each of the synergy reviews representing a combination of specific products, estimate positive or negative values for each synergy review, and estimate positive or negative values for each synergy review. (First compatibility degree) may be synthesized to determine a positive or negative degree (second compatibility degree) for a combination of specific products. For example, the second degree of affinity may be determined as an average value or a median value of the first degree of affinity.

Meanwhile, the second analysis result for the combination of the specific products may be determined based on an attribute keyword indicating a product attribute associated with the positive or negative degree (ie, the first compatibility degree) estimated for each synergy review. there is.

In other words, the second analysis model, for the first affinity for each synergy review determined by the first analysis model, is an attribute of a product associated with the corresponding first affinity (eg, having the greatest influence on the first affinity). (or an attribute keyword indicating a corresponding attribute) may be determined, and a second analysis result for a combination of specific products may be determined by synthesizing the determined attributes (or attribute keywords). For example, an attribute indicated by an attribute keyword included in the second analysis result may be an attribute corresponding to the most frequent value among the attributes (or attribute keywords) (determined with respect to the first affinity degrees).

Attributes corresponding to these modes and attribute keywords thereof may correspond to the aforementioned maximum contribution product attribute and keywords representing the maximum contribution product attribute.

Alternatively, an attribute indicated by an attribute keyword included in the second analysis result may be determined as an attribute most frequently mentioned in synergy reviews as a result of analysis of synergy reviews using an analysis model. Alternatively, the attribute indicated by the attribute keyword included in the second analysis result is the most mentioned attribute among the synergy reviews whose positivity/rating is equal to or greater than a predetermined value according to the analysis result of the synergy reviews by the analysis model and the first analysis result. may be determined. The attribute determined in this way may be the aforementioned maximum contribution product attribute. As for the method of determining the most mentioned attribute, the above-described method of determining the attribute corresponding to the most frequent value may be similarly applied.

A similar explanation can also be applied to indefiniteness, and redundant explanations are omitted.

Meanwhile, the attribute scores 742 and 744 described above with reference to FIGS. 3 and 7 are estimated positive or negative values for synergy reviews representing the attributes 732 and 734 associated with the corresponding attribute scores 742 and 744. can be determined based on

For example, the overall score 710 indicating the total synergy effect may be determined by analyzing all synergy reviews written for the combination of product A and product B 705 and 710 by the first analysis model. That is, the overall score 710 may be determined (eg, an average value, a median value, etc.) based on the first affinity degrees of all synergy reviews. In contrast, the first attribute score 742 is based on the first affinity degrees in which the first attribute 732 has the greatest influence according to the analysis result of the second analysis model on synergy reviews. (eg, an average value, a median value, etc.) may be determined. Similarly, the second attribute score 744 is based on the first affinity degrees in which the first attribute 734 has the greatest influence according to the analysis result of the second analysis model for synergy reviews. (eg, average value, median value, etc.).

Therefore, through the embodiment, not only the overall score 710 for the combination of product A and product B (705 and 710), but also the attribute score (742, 744) for each more detailed attribute is also the user terminal 110 can be provided in

On the other hand, the analysis model of the embodiment further uses the overall rating 830 and the individual ratings 852 to 856 input by the user in the analysis in determining the affinity (positive or negative) for the combination of products. can

For example, the overall rating 830 input by the user can be used (subsidiarily) to determine the overall rating 830 of FIG. 7 . For example, synergy reviews having an overall score 830 of a predetermined value or less (eg, 2 points or less) may be excluded from analysis by the analysis model. Alternatively, the average value or the median value of the comprehensive scores 830 for the synergy reviews may be used to correct the value of the comprehensive score (match) 720 determined by the first analysis model. Accordingly, the gap between the comprehensive score 830 by the user and the comprehensive score 720 by the analysis model can be prevented.

Additionally, individual ratings 852 - 856 entered by the user may be used (subsidiarily) to determine attributes 732 and 734 and attribute ratings 742 and 744 of FIG. 7 . For example, attributes indicated by the individual scores 852 to 856 may be extracted as attribute words for analysis by the second analysis model. In addition, the individual ratings 852 to 856 can also be used to determine the attribute that has the greatest impact on each synergy review. Individual ratings 852 - 856 may also be used to calibrate the values of attribute ratings 742 and 744 . Accordingly, the gap between the comprehensive score 830 by the user and the comprehensive score 720 by the analysis model can be prevented.

Meanwhile, the attributes 732 and 734 provided with the attribute ratings 742 and 744 and the attributes indicated by the individual ratings 852 to 856 may be different from each other. That is, through the attribute ratings 742 and 744, it is possible to provide a degree of affinity for attributes that are not input by the user (eg, as a default).

The description of the technical features described above with reference to FIGS. 1 to 3, 7, and 8 may be applied to FIGS. 4, 5, and 11 to 13 as they are, so duplicate descriptions are omitted.

6 is a flowchart illustrating a method of providing recommendation information including a synergistic effect according to a combination of products according to a search request from a user according to an example.

Referring to FIG. 6 , a method of providing an analysis result obtained by analyzing the aforementioned synergy reviews as recommendation information will be described.

For example, the case where the combination of products to which the analysis result is provided includes at least a first product and a second product will be described.

In step 610, the computer system 100 may receive a search request for the first product from the user terminal 110 of the user.

In step 620, the computer system 100, in response to the search request in step 610, provides information on a second product, the first analysis result, and the second analysis result as recommendation information on the first product to the user. It can be provided to the terminal 110. Here, the second product may be a product determined to have the highest compatibility with the first product according to the analysis of synergy reviews using the above-described analysis model. The "synergy effect between products" shown in FIG. 7 may be an example of recommendation information for a combination of product A 705 and product B 710 provided as a user searches for product A 705 .

Also, this recommendation information may be presented together with a search result for the first product. For example, recommendation information may be provided in the form of FIG. 9 .

9 illustrates a method of comparing a first product and a second product according to an example.

In the example, the first product information area 910 represents an area displaying information on product C requested by the user to be searched for. The second information area 920 represents an area for exposing a target advertisement in response to a user's search request. That is, information on product D may be displayed as recommendation information in the second information area 920 . Product D may be a product determined to have the highest compatibility with product C according to the analysis of synergy reviews by the analysis model. Although not shown in FIG. 9 , information 720 to 744 corresponding to the first analysis result and the second analysis result may be further displayed in the second information area 920 as shown in FIG. 7 .

As such, the analysis result of the synergy reviews of the embodiment may be used for targeted advertisements for users. That is, the analysis result of the synergy reviews of the embodiment is not directly provided to the user terminal 110, but is delivered to the advertiser (advertiser server or external server) that provides targeted advertisements for the user, so that the advertiser can target the user. It can be used as marketing information to determine advertisements. In addition, the analysis results of these synergy reviews may be transmitted to the product manufacturer (manufacturer's server) and used for product development and marketing.

Meanwhile, a synergy review creation button 930 may be displayed on the screen for comparing the products of FIG. 9 . For example, if the user already has a use experience for the combination of product C and (recommended) product D, the user can create a synergy review for the combination of product C and product D by selecting the synergy review creation button 930. . When the synergy review writing button 930 is selected, the synergy review writing UI described above with reference to FIG. 8 may be provided to the user terminal 110 .

According to the illustrated example, the user can easily obtain information on other products that can be used in combination with the product requested for search, and can conveniently create a synergy review for a combination of products that has already been used.

Descriptions of the technical features described above with reference to FIGS. 1 to 5, 7, 8, and 11 to 13 may be applied to FIGS. 6 and 9 as they are, so duplicate descriptions are omitted.

10 illustrates a method of displaying a synergy review on a user terminal according to an example.

The illustrated screen represents a screen of the user terminal 110 displaying reviews written by users.

As shown, the general review 1010 and the synergy review 1020 of the embodiment are displayed on the same screen but may be displayed separately from each other. Meanwhile, the user can view additional synergy reviews (and general reviews) by selecting the more button 1030 .

The synergy review 1020 that is first exposed may be the latest review or a review that has received the most recommendations from other users.

Synergy review 1020 may be provided along with a general review of Product A or Product B. The synergy review 1020 may be provided with a button for writing a synergy review for a combination of corresponding products (product A and product B), and as the corresponding button is selected, the user can create a new synergy for product A and product B. You can write a review.

Text included in the illustrated general review 1010 and synergy review 1020 may be analyzed by the above-described analysis model.

Since the description of the technical features described above with reference to FIGS. 1 to 9 and FIGS. 11 to 13 can be applied to FIG. 10 as it is, duplicate descriptions will be omitted.

14 illustrates a method of constructing an analysis model for synergy review analysis in a system for synergy review management according to an example. 15 illustrates a method of configuring a system for management of synergy reviews according to an example.

In FIGS. 14 and 15 , a synergy review management platform (shown affinity review sharing platform) implemented through the computer system 100 and a method of configuring an analysis model (a compatibility review analysis AI model) are exemplified.

As shown in FIG. 14, the compatibility review sharing platform may configure a review writing UI, may include a pre-written review DB (or may be configured to communicate with an external DB), and include a product inquiry module. The product DB may be searched according to a user's search request/review writing request. Information about a product searched by the user or other products that can be combined with the product searched by the user included in the aforementioned recommendation information may be inquired from the product DB.

The affinity review analysis AI model tokenizes the text of the review using the written reviews as learning data, can query a preset attribute DB, and extracts sentiment words (corresponding to the above-mentioned sentiment words) and attribute words from the text. can be extracted. The affinity review analysis AI model can analyze the degree of affinity between products (result of the above-mentioned first analysis) based on the sentiment word, and the attribute (core attribute) related to the degree of affinity based on the word of attribute (result of the second analysis described above). ) can be analyzed.

As shown in FIG. 15, the compatibility review sharing platform will provide analysis results (the above-described first and second analysis results) on the compatibility between products to the user terminal 110 or the advertiser through the compatibility analysis result UI. can The analysis result for the degree of compatibility may be stored in the compatibility analysis DB, and the analysis result stored in the compatibility analysis DB may be inquired and provided to the user terminal 110 or the advertiser.

The affinity review analysis AI model can analyze the compatibility between products and the attributes (core attributes) related to the affinity by analyzing newly created synergy reviews.

Meanwhile, the affinity review analysis AI model may be updated through learning about synergy reviews being created.

Functions such as providing and updating a compatibility review for a combination of products can be provided through a Restful API.

Since descriptions of technical features described above with reference to FIGS. 1 to 13 may be applied to FIGS. 14 and 15 as they are, duplicate descriptions are omitted.

Although the “product” described in the embodiments is illustrated as cosmetics, it may be a product of a category different from cosmetics, such as foodstuffs and home appliances, which are not limited to cosmetics. That said, the computer system 100 of the embodiment may support the creation and management of synergy reviews for these different categories of products.

The devices described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA) , a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. can be embodied in Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

Claims (10)

In the review management method performed by a computer system,
Analyzing synergy reviews written for a combination of at least two products among a plurality of products using an analysis model - the analysis model includes a first analysis result including a positive or negative degree for the combination of the products and , learned to output a second analysis result including attribute information associated with the first analysis result; and
outputting the first analysis result and the second analysis result according to the analysis by the analysis model;
Including, review management method. According to claim 1,
Each synergy review of the synergy reviews is information input through a synergy review writing user interface (UI) provided to the user terminal of the user who wrote each synergy review, and the first product and the second product constituting the combination of the products. , Review management method including information about and text for evaluating a combination of the first product and the second product. According to claim 1,
The analysis model includes a first analysis model learned to analyze text included in the synergy reviews and output the first analysis result for the combination of the products,
The first analysis result includes data obtained by quantifying the positive or negative degree of the combination of the products. According to claim 3,
The first analysis model is learned by the following steps to output the first analysis result,
These steps are
extracting first tokens composed of words indicating positive and second tokens composed of words indicating negative from texts of reviews pre-written by users;
generating a learning data set for the first analysis model by removing overlapping tokens of words and tokens of words corresponding to product attributes from the first tokens and the second tokens; and
Training the first analysis model to estimate a positive or negative rating for a review written by a user using the training data set.
Including, review management method. According to claim 3,
The analysis model includes a second analysis model learned to output the second analysis result for the combination of products based on the text included in the synergy reviews and the first analysis result,
The second analysis result, as the attribute information, includes at least one attribute keyword indicating a property of a product associated with a positive or negative degree indicated by the first analysis result. According to claim 5,
The attribute keyword includes a keyword representing a product attribute that contributes to a maximum positive or negative degree for the combination of products indicated by the first analysis result. According to claim 5,
The second analysis model is learned by the following steps to output the second analysis result,
These steps are
Extracting words representing attributes of a product from reviews previously written by users; and
Among the extracted words, to determine a word representing a property of a product associated with a positive or negative rating of each review of the pre-written reviews according to an analysis result by the first analysis model on the text of the pre-written reviews Training a second analysis model
Including, review management method. According to claim 5,
The first analysis result for the combination of products is determined based on the estimated positive or negative rating for each synergy review of the synergy reviews written for the combination of products,
The second analysis result for the combination of the products is determined based on an attribute keyword representing a property of a product associated with a positive or negative degree estimated for each synergy review, review management method. According to claim 5,
The outputting step is
outputting, as a result of the first analysis, a comprehensive score obtained by quantifying positive or negative degrees of the combination of the products, and the attribute keyword as a result of the second analysis; and
Further outputting attribute ratings associated with attributes of products indicated by the attribute keywords.
Including, review management method. According to claim 1,
The combination of the products includes at least a first product and a second product,
The outputting step is
As a search request for the first product is received from the user's user terminal, in response to the search request, information on the second product, the first analysis result, and the first product as recommendation information on the first product Step 2 providing analysis results to the user terminal
Including, review management method.

Images