KR102613260B1 - System for Integrating and Providing E-Commerce Data - Google Patents

Abstract

The present disclosure provides an apparatus and method for providing integrated e-commerce data. According to an aspect of the present disclosure, provided are an apparatus and a method for providing integrated e-commerce data, which infer, label, and store empty data of a preset format from collected e-commerce data.

Description

{System for Integrating and Providing E-Commerce Data}

This disclosure relates to an e-commerce data integrated provision system.

The content described in this section simply provides background information and does not constitute prior art.

Businesses in the e-commerce market sell products not only through their own sales channels, but also through channels that collect and sell products from multiple businesses. Accordingly, the same product is being distributed and sold in various channels through different sales methods, such as different product names or detailed product information, different prices, or bundled sales. In addition, these sales channels are being developed in various types, such as temporary creation and disappearance depending on time and users. As a result, e-commerce businesses are having difficulty collecting and utilizing e-commerce data distributed across various channels.

Additionally, to increase revenue, e-commerce businesses can use various additional solutions, such as applying advertising tools to each sales channel or using more user behavior analysis tools. However, in order to effectively analyze advertising and/or user behavior, it is necessary to comprehensively collect and analyze e-commerce data generated from various channels, including these additional solutions. Additionally, there is a need to visualize and provide data collected from various channels in a meaningful way.

The purpose of this disclosure is to provide a system that integrates and provides various types of e-commerce data generated from various channels.

According to one aspect of the present disclosure, an apparatus and method can be provided for collecting and storing e-commerce data from one or more channels, but inferring and storing empty data.

According to one aspect of the present disclosure, it is possible to receive a query of a natural language text, infer a query statement for obtaining an answer to the query, and a method for outputting the answer, and provide result data corresponding to the query along with an output method.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood by those skilled in the art from the description below.

According to one aspect of the present disclosure, a data unit; a collection unit that collects e-commerce data from one or more channels and stores it in a preset format in the data unit; a classification unit for inferring and labeling empty data of the collected e-commerce data using a product classification model previously learned to infer empty data of the preset format and storing the label in the data unit; An input unit that receives a request in natural language form from a user terminal; And by using an interactive data analysis model that has already been learned to infer a method of receiving a query in a natural language text and outputting a query statement and an answer to obtain an answer to the query, result data corresponding to the request in natural language format and the An e-commerce data integration provision device is provided, including a provision unit that provides a method for outputting result data.

According to one aspect of the present disclosure, in a method for an e-commerce data integration providing device to integrate and provide e-commerce data,

A process of collecting e-commerce data from one or more channels and storing it in a preset format in the data unit;

A process of inferring and labeling empty data of the collected e-commerce data using a product classification model previously learned to infer empty data of the preset format and storing the data in the data unit, thereby configuring the data unit; A process of receiving a request in natural language form from a user terminal; It outputs query statements and answers corresponding to requests in natural language format by using an interactive data analysis model that has already been learned to infer how to input queries in natural language text and output query statements and answers to obtain answers to the questions. the process of inferring a method; and a process of providing result data and a method for outputting the result data by applying the inferred query statement to the configured data unit.

According to one aspect of the present disclosure, the above-described product classification model extracts essential data to be used for inference from e-commerce data stored in the data unit, tokenizes data corresponding to specific attributes, and combines the extracted essential data and Provides a method and/or device for providing integrated e-commerce data that extracts features from e-commerce data input using tokens and infers empty data from the generated features.

According to one aspect of the present disclosure, a computer-readable recording medium recording a program for performing the above-described e-commerce data integration method and/or a computer-readable recording medium for executing each process of the above-described method. Provides stored computer programs.

According to one aspect of the present disclosure, there is an effect of providing integrated e-commerce information distributed in one or more channels.

According to one aspect of the present disclosure, there is an effect of appropriately inferring empty attribute information among integrated e-commerce information.

According to one aspect of the present disclosure, there is an effect of receiving natural language as input and providing result data corresponding to the natural language input from integrated e-commerce information together with an appropriate method of outputting the result data.

The effects of the e-commerce data integration providing device and method according to various embodiments of the present disclosure are not limited to the effects mentioned above, and are effects that a person skilled in the art may consider as the effect of the dashboard system of the present disclosure. Includes.

1 is a conceptual diagram showing an e-commerce data integrated provision system according to an embodiment of the present disclosure.
Figure 2 is a block diagram showing an e-commerce data integration provision device according to an embodiment of the present disclosure.
Figure 3 is an exemplary diagram of a data storage format of the data unit according to an embodiment of the present disclosure.
Figure 4 is a flowchart showing a method for providing integrated e-commerce data according to an embodiment of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described using exemplary drawings, but these embodiments and drawings are only for illustrating the present disclosure and do not limit the scope of the claims. Meanwhile, in relation to the description of the present disclosure, detailed descriptions of well-known structures or functions may be omitted.

In addition, in describing the components, functions, effects, etc. of the present disclosure, terms such as first, second, a, b, A, B, 1), 2) may be used, but this refers to each component. It is intended for distinction and does not limit the nature, order, or order of the components in question.

In addition, terms such as 'interface', 'system', 'platform', 'device', 'part', and 'module' used in the specification refer to a unit that processes at least one function or operation, which is hardware or software. Alternatively, it may be implemented through a combination of hardware and software.

Meanwhile, various components, modules, units, devices, interfaces, systems, etc. of the present disclosure may be implemented as computer hardware, software, applications, and/or combinations thereof, and these various implementations may be executable on a programmable system. It may include implementation of the above computer programs.

The drawings and reference numerals of the present disclosure are for explaining embodiments of the present disclosure and do not represent the only embodiments in which the present disclosure may be practiced.

The e-commerce data integration providing method of the present disclosure is executed by an e-commerce data integration providing apparatus, and the e-commerce data integration providing apparatus is executed on a computing device. The e-commerce data integration provision device performs each function by one or more processors available to the computing device, and includes computer-readable storage connected to the processor and having instructions stored therein. In this case, ‘inclusion’ may mean not only physical connection between components, but also network connection through communication means.

1 is a conceptual diagram showing an e-commerce data integrated provision system according to an embodiment of the present disclosure.

The e-commerce data integrated provision system according to an embodiment of the present disclosure includes a channel 10 for collecting e-commerce data (e.g., product-related data, order form, payment results, advertising indicators, etc.), from the channel 10. A user obtains corresponding result information by transmitting a specific request to the e-commerce data integration provision device 20 and the e-commerce data integration provision device 20, which collects e-commerce data, integrates, analyzes and stores the data. It may consist of a terminal 30. At this time, the connection between the channel 10, the e-commerce data integration providing device 20, and the user terminal 30 is not only connected through one or more networks with each as an endpoint, but also physically connected by an electrical method. It may be.

Channel 10 includes files stored and managed by the user, channels for selling the user's products (e.g. open market platform, company platform, e-commerce platform in store, etc.), and information applied by the user to manage/use the channel. tools (e.g. order collection solutions, easyadmin solutions, first party data analytics solutions, etc.), and/or channels/tools used by users to maximize revenue (e.g. advertising channels, social media, retail media solutions, display advertising solutions, etc.) , search advertising solutions, etc.). The e-commerce data integration provision device 20 collects e-commerce data in various formats according to a method suitable for each channel. For example, data can be collected using methods such as API connection, crawling, and scraping, and the format of the collected e-commerce data depends on the method provided by each channel, such as JSON, Text, and API response.

When fetching data in various formats from multiple channels, the e-commerce data integration providing device 20 can preprocess and store the collected e-commerce data in a way that allows storage and data processing. A formalization engine can be used for this preprocessing. For example, a formalization engine may be an engine that processes and stores data in various formats into a tabular format that can be stored in a table composed of column data and row data. Alternatively, the formalization engine may be an engine that processes and stores data in various formats, for example, in key-value format, document format, wide-column format, and graph format.

Meanwhile, the user terminal 30 receives data/information resulting from the integration of e-commerce data from the e-commerce data integration provision device 20. At this time, the e-commerce data integration providing device 20 may provide appropriate format information for visualizing the integrated result data, allowing the user terminal 30 to output the result data in an appropriate manner. The e-commerce data integration providing device 20 processes a request in natural language format input by the user terminal 30 or generated by the user terminal 30 to provide appropriate information without the user directly entering a query. You can check the resulting data. Here, a request in natural language format may be a request that includes information such as natural language text or natural language voice.

Figure 2 is a block diagram showing an e-commerce data integration provision device according to an embodiment of the present disclosure.

The e-commerce data integrated provision device 20 according to an embodiment of the present disclosure includes all or part of a data unit 200, a collection unit 220, a classification unit 240, an input unit 260, and a provision unit 280. Includes. The collection unit 220 collects e-commerce data from each channel and stores it in the data unit 200, and the classification unit 240 infers empty data from the stored e-commerce data and uses it as information on the corresponding e-commerce data. Save. The e-commerce data integration provision device 20 receives a request in natural language format from the user terminal through the input unit 260, and the provision unit 280 obtains result data for the request from the data in the data unit 200. , together obtain a method for outputting the result data. According to various embodiments of the e-commerce data integration providing device 20 of the present disclosure, some of the components shown in FIG. 2 may be changed or deleted, or new components may be added.

Meanwhile, the e-commerce data integration providing device 20 shown in FIG. 2 is not limited to a device and may be implemented as a software module or processor that performs the functions of each component 200 to 280.

The collection unit 220 collects e-commerce data from one or more channels and stores the e-commerce data in the data unit 200 in a preset format.

The classification unit 240 infers empty data of the collected e-commerce data using a product classification model that has been previously learned to infer empty data in a preset format. The classification unit 240 labels the corresponding e-commerce data as all or part of the inferred empty data and stores it in the data unit 200. This labeling may be labeling with attribute information for obtaining result data that a user (e.g., an e-commerce business operator) wants to receive from integrated e-commerce data. For example, an e-commerce business may want to integrate e-commerce data to provide indicators by product group (e.g. sales, contribution margin, etc.) or indicators by brand. Accordingly, the product classification model can infer a ‘product group’ and/or a ‘brand’ as empty data based on the inferred feature(s), and label the inferred product group and/or brand in the corresponding e-commerce data.

Meanwhile, the preset format when e-commerce data is first stored in the data unit 200 may not include the attribute to be labeled. That is, in the product classification model, the classification unit 240 can infer (newly create) attribute information to be labeled from the extracted features. In this case, the classification unit 240 can cause the data unit 200 to newly create an attribute to be labeled and store the inferred attribute information. As a result, the e-commerce data integrated provision device 20 according to an embodiment of the present disclosure can label e-commerce data with appropriate attributes according to information requested by the user.

The product classification model extracts essential data to be used for inference from the e-commerce data stored in the data unit 200, tokenizes data corresponding to specific attributes, and inputs e-commerce using the extracted essential data and tokens. Features can be extracted from merch data and empty data can be inferred from the generated features.

Figure 3 is an exemplary diagram of a data storage format of the data unit according to an embodiment of the present disclosure.

Figure 3 shows an example of inferring empty column data of e-commerce data using column data and row data when the data unit 200 stores e-commerce data in tabular format. In this case, the product classification model can be used to extract features by tokenizing heat data as data corresponding to specific attributes.

Referring to FIG. 3, 6 row data and 4 column data can be confirmed in FIG. 3. Each row data can have four attribute information as column data.

In order to infer the empty data of the fifth row data ((a) and (b) in Figure 3), the product classification model uses the column data of the 'brand_name' attribute and the column data of the 'category' attribute at the same time or at the same time as the token. In this way, each empty data can be inferred.

Alternatively, the product classification model can extract row data with the same ‘company_name’ among the fifth row data as essential data and tokenize the column data of the extracted row data to infer empty data. In the case of the fifth row data in Figure 3, since it has 'A Company' as 'company_name', the product classification model extracts all row data corresponding to 'A Company' and adds 'brand_name' to the extracted row data. Tokenize the corresponding ‘Atrang’ and ‘Beauty Feminine’. The product classification model uses all or part of the column data such as ‘company_name’, ‘brand_name’, ‘channel_name’, ‘product_name’, and ‘category’ to infer the ‘brand_name’ value of the fifth row data. The specific inference method may vary depending on how the product classification model was learned and the attribute information contained in the row data. In the example of Figure 3, based on the similarity between ‘channel_name’ and ‘product_name’, ‘brand_name’ can be inferred as ‘Beauty Feminine’.

Considering the similarity of 'channel_name', 'product_name', and 'brand_name', the 'category' column data ((b) and (c) of Figure 3) of the fifth and sixth row data of Figure 3 is, Inferring from the column data of 'Company A' row data may be less accurate. In other words, the required data extraction method for inferring empty data in the product classification model may vary depending on the attribute to which the empty data belongs.

The input unit 260 receives a request in natural language format from the user terminal. This request in natural language format includes natural language text/voice input by the user terminal from the user, or natural language text generated by the user terminal based on user input (e.g., command selection, arrangement of no-code objects, etc.) can do.

The provider 280 receives a query in natural language text and uses a pre-trained interactive data analysis model to infer how to output a query statement and answer to obtain an answer to the query, and responds to the input request in natural language format. Provides corresponding result data and a method for outputting the result data. This provision may be provided within the e-commerce data integrated provision device 20 as well as provided to the user terminal.

The interactive data analysis model processes natural language text queries in natural language and outputs answers to first models and/or natural language text queries that have been previously learned to generate queries to obtain answers from data in the preset format. It may include a second model, which is a transformer model tuned using a labeled dataset. However, this second model is not necessarily based on the transformer model, and the second model of the present disclosure can be used using any network structure or any learning method to receive a natural language text query and infer an appropriate answer output method. there is.

The query statement generated by the first model may be an SQL statement or an UnQL statement, depending on the data storage method of the data unit 200. For example, if the data unit 200 has relational database characteristics, the first model may generate an SQL statement.

<Table 1> is an example of the dataset used to tune the second model.

Referring to <Table 1>, the second model, for example, for the query ‘How do customer demographics affect sales?’ or queries with a high degree of similarity, uses demographic information as the x-axis and sales information as the y-axis. It may be a transformer model tuned to use a scatter plot as a result data output method.

One
Q: What is the average sales per region?
A: {type: 'bar-chart', x-axis: 'region', y-axis: 'average_sales'}
2
Q: How have website visits changed over time?
A: {type: 'line-chart', x-axis: 'date', y-axis: 'website_visits'}
3
Q: How do customer demographics affect sales?
A: {type: 'scatter-plot', x-axis: 'demographic', y-axis: 'sales'}
4
Q: How do different marketing strategies impact customer acquisition?
A: {type: 'bar-chart', x-axis: 'marketing_strategy', y-axis: 'customer_acquisition'}
5
Q: What is the distribution of expenses across different departments?
A: {type: 'pie-chart', x-axis: 'department', y-axis: 'expenses'}
6
Q: What is the correlation between employee satisfaction and productivity?
A: {type: 'scatter-plot', x-axis: 'employee_satisfaction', y-axis: 'productivity'}

The first model may be a model learned to process not only Korean but also foreign language text queries. The first model may be a model learned to process typos, inscriptions, new words, anomalous words, and/or special characters.

The interactive data analysis model consists of an evaluation unit and an answer generation unit and may be a competitively and iteratively learned model. For example, the answer generation unit may generate a query sentence and an output method from a natural language text query, and the evaluation unit may score and evaluate the answers generated by the answer generation unit. Based on the evaluation results (scores), the answer generation unit can perform learning again in the direction of improving the evaluation results of the evaluation unit. Learning of the answer generation unit can be performed by repeating the cycle until the evaluation result of the evaluation unit reaches a threshold (e.g., 99% accuracy).

With the introduction of this interactive data analysis model, users can easily obtain integrated e-commerce data from the e-commerce data integration providing device 20 according to an embodiment of the present disclosure without having to learn commands. In addition, the user can obtain meaningful information by obtaining e-commerce data visualized in an appropriate manner according to the requested data using the e-commerce data integration providing device 20 according to an embodiment of the present disclosure.

Figure 4 is a flowchart showing a method for providing integrated e-commerce data according to an embodiment of the present disclosure.

The e-commerce data integrated provision device collects e-commerce data from one or more channels and stores it in a preset format in the data unit (S400).

The e-commerce data integration providing device uses a product classification model that has been previously learned to infer empty data in a preset format, infers and labels empty data of the collected e-commerce data and stores it in the data unit, Construct wealth (S410).

The e-commerce data integration provision device receives a request in natural language format from the user terminal (S420).

The e-commerce data integration provision device receives a query in natural language text and uses an interactive data analysis model that has been previously learned to infer how to output the query statement and answer to obtain the answer to the query, The query statement corresponding to the request and the method of outputting the answer are inferred (S430).

The e-commerce data integration providing device provides result data and a method of outputting the result data by applying the inferred query statement to the data unit configured in step S410 (S440).

Meanwhile, Figure 4 illustrates an exemplary operation method of the e-commerce data integrated provision method, and the e-commerce data integrated provision method is not necessarily executed sequentially according to the steps of Figure 4. For example, each step in FIG. 2 may be performed in parallel, in a different order, or with some steps omitted, as long as it does not conflict with the essence of the method for providing integrated e-commerce data of the present disclosure.

Various embodiments of the above-described device or method of the present disclosure may include being implemented with one or more computer programs. Such a computer program is executed by a computer system including one or more input devices, one or more output devices, memory, and one or more programmable processors coupled to transmit and receive data and instructions between the memory, the input devices, and the output devices, It can be driven, operated, and operated. A computer system may be, for example, but not limited to, a server, network device, set-top box, embedded device, computer expansion module, personal computer, laptop, personal data assistant (PDA), cloud computing system, or mobile device. Computer programs (e.g. firmware, middleware, OS, software, applications, source code) include instructions for driving or controlling a programmable processor and are stored in a computer-readable recording medium.

Here, computer-readable recording media includes all types of recording devices that store data that can be read by a computer system. Computer-readable recording media may be, for example, ROM, magnetic tape, floppy disk, memory card, hard disk, magneto-optical disk, storage device, etc., but are clearly not limited thereto. Additionally, the computer-readable recording medium may be distributed in a computer system connected to a network, and the computer-readable code may be stored and executed in a distributed manner.

The above is a description of various embodiments of the present disclosure, and various modifications are possible depending on various configurations, materials, means, methods, and knowledge that can be employed by those skilled in the art. These embodiments are only for illustrating the technical idea of the present disclosure and do not limit the technical idea or scope of the present disclosure.

20: E-commerce data integration provision device
200: data unit
220: Collection department
240: Classification department
260: input unit
280: Provider

Claims (6)

data department;
a collection unit that collects e-commerce data from a plurality of channels and stores it in a preset format in the data unit;
a classification unit for inferring and labeling empty data of the collected e-commerce data using a product classification model previously learned to infer empty data of the preset format and storing the label in the data unit;
An input unit that receives a request in natural language form from a user terminal; and
By using an interactive data analysis model that is pre-trained to receive a query in natural language text and infer how to output a query statement and answer to obtain an answer to the query, result data corresponding to the request in natural language format and the result Includes a provision unit that provides a data output method,
The plurality of channels include at least two of the following: files stored and managed by the user, channels for selling the user's products, tools used by the user to manage the channel, and channels used by the user to maximize profits. Including,
The product classification model extracts essential data to be used for inference from the e-commerce data stored in the data unit, tokenizes data corresponding to specific attributes, and inputs e-commerce data using the extracted essential data and tokens. Extracting features from data, inferring the empty data and attribute information for labeling from the generated features,
The interactive data analysis model includes a first model that is pre-trained to generate a query to obtain an answer from data in the preset format by processing natural language text queries in natural language, and a method for outputting answers to natural language text queries. Includes a second model, which is a transformer model tuned using a labeled dataset,
The first model is a model that has been pre-trained to be able to process Korean and foreign language text queries, and to be able to process typos, inscriptions, new words, anomalous words, and special characters included in the query,
The classification department,
If the format for storing data in the data unit does not include an attribute related to the attribute information, the attribute related to the attribute information is newly created in the data unit, and the empty data and the attribute information are stored in the data unit. to save,
E-commerce data integration provider. delete delete delete delete delete

Images