KR102237319B1 - An artificial intelligence system providing customized goods - Google Patents

Abstract

The present invention relates to a customized commercial information artificial intelligence search system, and more particularly, the present invention collects shopping information of online and offline stores in real time using a crawler to convert into big data, and based on this big data product information DB Users showing similar preferences while analyzing the user's personal preferences by analyzing the product search terms used by users for product search mainly through images or voices, product search information, which is the search result, and purchase product information through the artificial intelligence server. It relates to a customized commercial information artificial intelligence search system that categorizes into user groups grouped by individual, provides and recommends real-time product information optimized for each individual or group.

Description

Customized commercial information artificial intelligence search system {An artificial intelligence system providing customized goods}

The present invention relates to a customized commercial information artificial intelligence search system, and more particularly, the present invention collects shopping information of online and offline stores in real time using a crawler to convert into big data, and based on this big data product information DB Users showing similar preferences while analyzing the user's personal preferences by analyzing the product search terms used by users for product search mainly through images or voices, product search information, which is the search result, and purchase product information through the artificial intelligence server. It relates to a customized commercial information artificial intelligence search system that categorizes into user groups grouped by individual, provides and recommends real-time product information optimized for each individual or group.

Recently, O2O (Online-To-Offline), which is receiving a lot of attention in the mobile commerce market, can be cited. O2O is a method of inducing consumers offline by attracting customers through online based on ICT (Information & Communication Technology) technology. For example, by using Near Field Communication (NFC) technology or location based (LB) technology, discount coupons are sent to consumers near offline stores to attract potential customers to offline stores. In addition, it is possible to derive effective marketing methods by analyzing the data accumulated through such attracted customers. It is attracting attention as a means of increasing the profits of offline stores as it is possible to implement sophisticated customer management, such as providing more benefits to loyal customers based on the information of visiting customers.

Existing O2O services were at the level of listing simple location-based store information. For this reason, most consumers had to purchase products by considering the image, rating, average delivery time, and price of the product in a very complex manner. Therefore, there is a need for a function that minimizes such annoying choices and automatically selects products from more products. In addition, since all the data in the existing O2O is non-real-time data, it is not possible to convert sales according to real-time sales situation for each store (when hourly sales are insufficient, discount sales function is insufficient), and there is no way to inform consumers.

On the other hand, as part of an effort to supplement the above problems, there is a chat-type ordering method using artificial intelligence. Delivery Hero's chatbot'Hungry Bot' is ordered in a natural language conversation. A restaurant or menu is recommended to the user by referring to past order history data in the corresponding dialog. If you repeatedly order a specific pizza company or a specific menu through the Facebook messenger window, the details of the order are remembered and reflected in the next pizza order.

Kakao and Naver are also offering chat-type orders like chatbots. Kakao provides'Order Kakao Talk', and Naver provides a chat-type service called'Naver Talk Talk'. Naver's own artificial intelligence speaker will be linked with the'people of delivery' of elegant brothers. The artificial intelligence speaker is a service that converts human speech into text and finds the product. Kakao also provides ordering, taxi calls, and recommended music services using a speaker called Kakao Mini.

The recommendation function of the above technologies is merely showing the latest categories. That is, the current product recommendation is limited to simple product recommendation that considers the user's recently viewed category or age. This level of recommendation can be implemented by simply changing the search conditions of the program without the introduction of artificial intelligence. However, it is necessary to introduce artificial intelligence in order to comprehensively judge the user's actual purchase product and purchase pattern. That is, in order to quantify and analyze the location of the user's point of sale and the image of the product that the user viewed the most, it is necessary to introduce artificial intelligence with a function capable of selecting comprehensive products such as humans. For example, if a user frequently visits a specific place in a specific region by analyzing purchase data such as purchase product information, the product may be recommended by guessing the user's moving route. This is based on the user's purchase product information, and if the user's main purchase product is clothing, and among the clothing, if a pink dress with a short skirt, the image of the product is quantified and the most similar product in the store on the moving path of the user's visited area is selected. Intuitive product recommendation of the recommended function becomes possible. This intuitive product recommendation is particularly useful when recommending local products of the O2O business.

Domestic registered patent 10-1635090 (Registration date 2016.06.24) Korean Patent Publication 10-2018-0016180 (Publication date 2018.02.14) Domestic registered patent 10-1918358 (Registration date 2018.11.07) Domestic registered patent 10-0340036 (Registration date 2002.05.27) Domestic registered patent 10-1180371 (Registration date 2012.08.31)

An object of the present invention for overcoming the above problems is to provide a big data search system and intuitive image search and voice recognition through big data conversion of product information to obtain desired information.

Another purpose is to provide a function that recommends products tailored to the user's taste by utilizing an artificial intelligence server. This is to recommend similar products by storing the user's product log as data through an artificial intelligence server that is linked with the big data product information DB collected by the crawler, and analyzing the user's preferences on a specific date.

Another purpose is to enable even interactive search orders using voice recognition technology. When the user speaks in natural language through the interface module, the corresponding voice file is automatically converted into text, and a product similar to the information is searched for and provided to the user.

Another purpose is to use image search to capture the product currently being viewed in a street show window, a large mart, etc. with a camera through the interface module of the user terminal. It is to search for similar products and provide them to users.

Another object is to provide a product information search and reception system that converts online and offline store and product information into big data using a crawler, and enables image search and voice search through an interface module of a user terminal.

The present invention for achieving the above object is a big data product information DB for storing product information; A plurality of offline seller terminals that provide store and product information to the big data product information DB; An interface module installed in a plurality of user terminals to input a product search word using any one of an image, a voice, and a text, and receive the searched product search information; An image search module for providing the product search information calculated by searching the image or text of the product search term in the big data product information DB to the interface module; Voice search in which the product search information in text form, calculated by searching the voice of the product search term in the big data product information DB through a text conversion API (Application Programming Interface), is provided as voice to the interface module through the voice conversion API module; A crawler that collects data of on/offline store and product information, including the product search term, the product search information, and purchase product information of a user through the product search term, and stores the data in the big data product information DB; A data collection unit for collecting the product search terms collected through the crawler, the log information of the interface module including the product search information and the purchased product information as user-specific data, and classifying the plurality of user-specific data for each product A product category unit, a preference analysis unit that analyzes the classification of each product based on the purchased product information to identify preferences for each product for each user, a user group unit that groups individual users according to the preferences of each user, and the user group An artificial intelligence server including a product recommendation unit that provides product information optimized for each; It characterized in that it consists of.

In addition, the voice search module may include a Cloud Speech API for converting a voice file, which is a natural language input through the interface module, into text data; A text extracting unit for extracting a search word and a command by simplifying the text data for each word; A response message unit for comparing the extracted command with a preset command DB, searching for the search word as the extracted command in the big data product information DB through the artificial intelligence server, and responding to the result as a message; Android TTS API for converting the text of the message into a voice file; It characterized in that it consists of.

In addition, the voice search module may further include a material text unit for texting to the user through the response message unit when the command extracted through the text extracting unit is not registered in the command DB.

Further, provided with an NFC (Near Field Communication) sticker that provides information to the crawler and stores the store-specific ID of the seller terminal, wherein the user terminal having location-based information is tagged on the NFC sticker, or the product search information With the permission of the store's'follow' in the product information details of, it is automatically paired to any one of Wi-Fi, beacon, and NFC installed in the offline store to receive product information, event information, discounts or free coupons of the offline store. do.

The present invention having the above-described configuration has the advantage that first, if a favorite product is photographed in a street or a store using image search, a product similar to the same can be searched and purchased one-stop.

Second, there is an advantage of having no difficulty in searching and purchasing products even for the disabled or the elderly who are physically inconvenient through voice search. Touch-oriented smartphones may have restrictions on use for handicapped people with disabilities or poor eyesight, but they can also order and pay for products with only voice, which is very helpful to the socially weak.

Third, there is an advantage that only products optimized for users can be recommended. In busy daily life, the function that automatically selects my preferences can reduce the stress of selection for the user. The present invention uses artificial intelligence to analyze the user's product search information and purchase product information to recommend an optimized product, so that the user can easily and conveniently select and purchase a product.

1 is a conceptual diagram of a product information retrieval system for an on/offline store according to the present invention.
Figure 2 is an overall configuration diagram for the present invention customized commercial information artificial intelligence search system service.
3 is a conceptual diagram of product search of the image search module of the present invention.
Figure 4 is a structural diagram of speech recognition of the voice search module of the present invention.
5 is a flowchart of the voice search module of the present invention.
Figure 6 is a product classification learning flow chart of the present invention artificial intelligence server.
Figure 7 is a configuration diagram of the present invention artificial intelligence server.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, since the accompanying drawings show preferred embodiments of the present invention, the scope of the present invention is not limited by these drawings. In addition, even if the configuration is essential for carrying out the present invention, a detailed description will be omitted for a configuration that belongs to a known technology or can be easily implemented by a person skilled in the art.

As shown in Fig. 1, the present invention collects shopping information of online and offline stores in real time using a crawler 7 to convert it into big data, and based on this big data product information DB 1, the user mainly uses images or voices. User groups grouped by users showing similar preferences by analyzing the product search terms used for product search, product search information, and purchase product information through the artificial intelligence server (8). It categorizes as, and features a customized commercial information artificial intelligence search system 10 that provides and recommends real-time product information optimized for each individual or group.

As shown in Figs. 1 and 2, the customized commercial information artificial intelligence search system 10 of the present invention having the above characteristics includes a seller terminal 2, a user terminal 3, a big data product information DB 1, It consists of a crawler 7 and an artificial intelligence server 8 that collects information in real time.

The plurality of seller terminals 2 include both online and offline stores that provide store information and product information to the big data product information DB 1. In the case of an online store, store and product information registered online is automatically collected by the crawler 7 and stored in the big data product information DB 1. In addition, in the case of an offline store, store and product information is provided to the crawler 7 and/or the big data product information DB 1.

In addition, the user terminal 3 having location-based information may be connected to the seller terminal 2 through online or offline. First, it is possible to connect by tagging the user terminal 3 to an NFC sticker in which the ID of each store of the seller terminal 2 installed in the offline store is stored. In addition, it is possible to connect to the offline store with the permission of'following' of the store in the product information details of the product search information provided to the user terminal 3 through the big data product information DB 1. As described above, when the user terminal 3 is connected to the seller terminal 2, the user terminal 3 is automatically paired or connected to the interface module 4 of the user terminal 3 to any one of Wi-Fi, beacon, and NFC installed in the store. Through the automatic connection, it is possible to receive product information, event information, discounts or free coupons of the offline store in real time.

Interface modules are installed in the plurality of user terminals 3. The preferred user terminal 3 is a smartphone, but is not limited thereto. The consumer operates the interface module 4 to take any one of an image, voice, or text through photo taking as an input value, and communicates with the big data product information DB 1 to search results according to product search terms and product search terms. Receive product search information. At this time, the big data product information DB 1 is provided with on/offline store and product information collected in real time by the crawler 7. And the artificial intelligence server 8 is a product search word input through the interface module 4 of the user terminal 3, and the product information or purchase intention that the user selects and purchases through the product search information that is the search result. Product search information optimized for the entered product search word is calculated in consideration of high-ranked product information, etc., and customized information is provided to individual users. Of course, at this time, the artificial intelligence server 8 not only features of individual users to provide customized information, but also product search words, product search information, purchased product information, and wished product information entered through the interface module 4 of other users. May be used to calculate customized information of individual users.

3 is a conceptual diagram of a product search when product information is searched by image using the interface module 4 of the user terminal 3.

The image search module 5 operates when an image search is selected by operating the interface module 4 of the user terminal 3. When a product is photographed in a show window or a large mart through the interface module 4, the image file is transmitted to the big data product information DB 1. The product image file searches the indexed data in the big data product information DB 1 through the LIRE (Lucene Image Retrieval) module, and transmits the product most similar to the image captured by the user to a list. Here, except for the LIRE module in the description of the image search module 5, the process of product search information by text search is the same.

4, the voice search module 6 executes the interface module 4 in the user terminal 3 so that the user clicks the personality recognition button or activates the voice recognition function through voice to speak in natural language. If so, the corresponding voice file is converted into text, and a product similar to the corresponding information is searched for and displayed. Using such voice recognition technology, even an interactive search order may be possible.

The voice search module 6 converts a voice file, which is a query word, into text data using Google's Cloud Speech Application Programming Interface (API) 61. In addition, the text data is simplified for each word by the text extracting unit 62 to extract search words and commands. The search word extracted by the text extracting unit 62 and the command are compared whether or not the search word and the command are recognizable in the command DB 66 preset by the query word processing unit 63. The query recognized by the query word processing unit 63 retrieves the search word as the command from the big data product information DB 1 through the artificial intelligence server 8, and sends the result to the user in the response message unit 64. You will respond with a message. Here, since the response message is text, it is provided as product search information to the user in the form of a voice file through the Android Text-To-Speech Application Programming Interface (Android TTS API) 65 that converts it into a voice file.

In addition, as shown in FIG. 5, the query word processing unit 63 is for learning a language recognition function in order to provide a corresponding answer by determining for what purpose a query language, which is a variety of natural languages used by a user, is used. It may be composed of a unit 63a, a comparison unit 63b, a similar word determination unit 63c, an induction query unit 63d, and a synonym processing unit 63f. In this way, since the voice search module 6 has the same language recognition function as the query word processing unit 63, the voice search module 6 even functions of a chatbot to complete an order for a product while exchanging a conversation with a user. Becomes feasible.

The number replacement unit 63a determines a similar word as a probability using a pre-learning algorithm for the user's query word. In the comparison unit 63b determined by the number replacing unit 63a, the query word is compared with the pre-learning query word stored in the command DB 66 set in advance. The comparison result with the command DB 66 set in advance by the comparison unit 63b is transmitted to the similar word determination unit 63c, and the similar word determination unit 63c determines whether the query word of the user is a learned query word. At this time, the query words recognized as similar words through the similar business panel and the query words re-answered through the derived query unit 63d, which will be described below, are identified by the recognition confirmation unit 63e. Through the artificial intelligence server 8, the big data product information DB 1 is searched and the result is provided to the user as product search information through a response message unit 64. However, for a query word that is not recognized by the similar word determination unit 63c, the induction query unit 63d makes a derivation query to the user through the response message unit 64 as an induction sentence having a high similarity. And, if the answer re-answered through the derived query is a query language that can be recognized as a learned query language of the command DB 66 previously set through the comparison unit 63b and the similar word determination unit 63c, the initial query word and A synonym processing unit 63f that processes the derived query word as a synonym is processed as a synonym.

The crawler 7 includes the product search word input by the user into the interface module 4 from the user terminal 3, the product search information as a result of the search through the big data product information DB 1, and the product search word. And through the product search information, the big data product information DB (1) is collected by collecting data of on/offline store and product information in real time, as well as information on products that the user actually purchased and wished products reflecting the intention to purchase. Save it.

6 and 7, the artificial intelligence server 8 is based on the product search word input through the interface module 4 of the user terminal 3, among the information stored in the big data product information DB 1 It performs a function of calculating product search information optimized for the user and transmitting it to the interface module 4.

Here, the optimized product search information reflects the gender, age, and region of the user, as well as the user's product preference type, which is identified based on the product search term previously searched through the interface module 4. This is because the crawler 7 collects and stores the user's product search terms, product search information, purchased product information, and favorite products in real time, so that the artificial intelligence server 8 Product search information optimized for each individual can be calculated and provided to the user through the interface module 4. This means that even with the same search word, the artificial intelligence server 8 provides customized product search information specialized for each user through learning.

The artificial intelligence server 8 having the above functions includes a data collection unit 81, a product category unit 82, a preference analysis unit 83, a user group unit 84, an optimal product recommendation unit 85, and It includes a purchase induction unit (86).

First, the data collection unit 81 stores all log information of the interface module 4 including the product search terms, product search information, purchased product information, and favorite products collected through the crawler 7 by user. To collect. In addition, the user-specific data collected as user-specific data in the data collection unit 81 is not of one user, but a plurality of user-specific data, which are multiple users. The product category unit 82 is set by classifying each product. And, through the product category unit 82, for each individual user, the preference analysis unit 83 is set by determining which type of product a specific user prefers from a specific product. Next, a similar preference range of individual users for a specific product is set through the preference analysis unit 83, and a user group unit 84 for grouping individual users according to the preference setting range is set. In this way, when individual users are grouped according to their preferences, the optimal product recommendation unit 85 is a product selected by many other users belonging to the same group as the specific user when a specific user inputs a product search term through the interface module 4. The search information is listed and provided as product search information of the specific user.

In addition, the artificial intelligence server 8 has not actually made a purchase among the product search information provided by the optimal product recommendation unit 85 through the user's product search term. For products, if a product similar to the user's wished product is registered in the offline store of the visited area by tracking the user's visited area, product information, event information, discounts or free coupons of the offline store with similar products are sent to the user. A purchase induction unit 86 that induces the purchase of a favorite product may be configured.

10: artificial intelligence search system 1: big data product information DB
2: seller terminal 3: user terminal
4: Interface module 5: Image search module
6: Voice search module 61: Cloud Speech API
62: text extraction unit 63: query word processing unit
63a: numeric substitution unit 63b: comparison unit
63c: similar word judgment unit 63d: derived query unit
63e: recognition verification unit 63f: synonym processing unit
64: Response message part 65: Android TTS API
66: Command DB 7: Crawler
8: artificial intelligence server 81: data collection unit
82: product category unit 83: preference analysis unit
84: user group unit 85: optimal product recommendation unit
86: purchase induction department

Claims (5)

Big data product information DB for storing product information;
A plurality of seller terminals that provide store and product information to the big data product information DB;
An interface module installed in a plurality of user terminals having location-based information to input a product search term using any one of image, voice, and text, and to receive the searched product search information;
The product search information calculated by searching the image or text of the product search term in the big data product information DB, the image being indexed in the big data product information DB using a LIRE (Lucene Image Retriewal) module. An image search module for providing the interface module to the interface module;
Cloud Speech API (Application Programming Interface) for converting a voice file, which is a query word input through the interface module, into text data, a text extraction unit for extracting search words and commands by simplifying the text data for each word, and the extracted search words and commands A query word processing unit that compares a preset command DB, a response message unit that searches the search word with the extracted command from the big data product information DB through the artificial intelligence server and responds with a message, and voices the text of the message. Android TTS API for converting into a file, a voice search module providing voice to the interface module through;
A crawler that collects data of on/offline store and product information in real time, including the product search term and product search information and user's purchase product information through the product search term, and stores it in the big data product information DB;
A data collection unit for collecting the product search terms collected through the crawler, the log information of the interface module including the product search information and the purchased product information as user-specific data, and classifying the plurality of user-specific data for each product A product category unit, a preference analysis unit that analyzes the classification of each product based on the purchased product information to determine preferences for each product for each user, a user group unit that groups individual users according to the preferences of each user, and the user group An artificial intelligence server including an optimal product recommendation unit that provides product information optimized for each; Customized commercial information artificial intelligence search system, characterized in that consisting of.
The method of claim 1,
The artificial intelligence server provides product information and event information of an offline store holding the wished product according to the user's visited region, in order to induce the purchase of a wished product reflecting the user's purchase intention among the product search information provided through the optimal product recommendation unit. , Customized commercial information artificial intelligence search system, characterized in that it further comprises a purchase induction unit for sending a discount or free coupon.
delete The method of claim 1,
The query word processing unit may include a numeric replacement unit for determining a similar word as a probability using a pre-learning algorithm for the query word; A comparison unit comparing the user's query word with a pre-learning query word stored in the command DB according to the result of the number replacement unit; A similar word determination unit that determines whether or not similarity is based on a result of the comparison unit; A derivation query unit that performs a derivation query to a user through the response message unit with an induction sentence having a high similarity level, which is not recognized by the similar word determination unit; A recognition confirmation unit that responds to a user through the response message unit recognized through the similar word determination unit and the induction query unit; A synonym processing unit that processes the initial query word and the derived query as a synonym when the answer re-answered through the derived query is a query word recognizable through the comparison unit and the similar word determination unit; Customized commercial information artificial intelligence search system, characterized in that consisting of.
The method of claim 1,
Provides information to the crawler and has an NFC (Near Field Communication) sticker storing the ID of each offline store of the seller terminal,
The user terminal with location-based information is automatically paired to any one of Wi-Fi, beacon, and NFC installed in the offline store by tagging on the NFC sticker or with the'follow' permission of the store in the product information details of the product search information. Customized commercial information artificial intelligence search system, characterized in that receiving product information, event information, discount or free coupon of the offline store.

Images