KR102621683B1 - Customer response system using artificial intelligence - Google Patents

Abstract

The present invention relates to a customer response system using artificial intelligence, which includes a customer service management server that builds an artificial intelligence-based chatbot model and provides an automated response service when customer inquiries about a specific product.

Description

Customer response system using artificial intelligence}

The present invention relates to a customer response system using artificial intelligence, and more specifically, to customer response using artificial intelligence that can provide an automated response service when customer inquiries about a specific product are built by building an artificial intelligence-based chatbot model. It's about the system.

Today, with the development of computers and other information and communication technologies, artificial intelligence (AI) technology has also gradually developed and is currently being applied to various fields. One of the technologies that applies artificial intelligence (AI) is chatbot.

A conventional chatbot is an interactive messenger that provides answers through artificial intelligence (AI), which talks to people in everyday language based on big data analysis, when you enter a question as if chatting on a corporate messenger. Competition among IT companies is intensifying because while providing corporate messenger services, natural language processing capabilities can be improved by analyzing usage patterns of business smartphones/PCs or collecting big data such as languages commonly used at work. Corporate messengers that adopt this chatbot function have the advantage of being able to use it as a platform to connect various functions because they can check and process information in the chat window without running a separate app.

Meanwhile, the market for small-package food delivery services is rapidly increasing due to the increase in single-person households, and in proportion to this, inquiries about products purchased or intending to purchase are also explosively increasing. However, conventional customer service services using human resources have limitations in terms of cost and time.

Meanwhile, the above-mentioned background technology is technical information that the inventor possessed for deriving the present invention or acquired in the process of deriving the present invention, and cannot necessarily be said to be known technology disclosed to the general public before filing the application for the present invention. .

Korean Patent No. 10-2445519

One aspect of the present invention provides a customer response system using artificial intelligence that can provide an automated response service when a customer inquires about a specific product by building an artificial intelligence-based chatbot model.

The technical problem of the present invention is not limited to the technical problem mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

A customer response system using artificial intelligence according to an embodiment of the present invention includes a customer service management server that builds an artificial intelligence-based chatbot model and provides an automated response service when a customer inquires about a specific product.

The customer service management server,

Provide a chatbot-based interactive interface to the consumer terminal owned by the customer who purchased the product, receive question data related to the product through the interactive interface, and

Extracting at least one keyword from the question data using a pre-trained artificial neural network,

The extracted keywords are sent to a network-connected external search engine and an external chatbot to collect response data from the external search engine and external chatbot, and response data is generated based on the plurality of collected response data to provide the above answers through an interactive interface. Data is transmitted to the consumer terminal.

The customer service management server,

The CS score is calculated for each response data using the following equation, and the response data with the highest calculated CS score is set as the response data.

[Equation]

Here, CS is the CS score, da_i is the distance between the embedding vector for the ith keyword among the plurality of keywords extracted from the response data and the embedding vector for the preset reference word, and w_i is different for each of the plurality of keywords extracted from the response data. The set weight (first weight), da_s is the average distance between embedding vectors for multiple keywords extracted from response data, t is the response delay time, a is the weight (second weight) set differently for each external search engine and external chatbot. weight), dq is the average distance between embedding vectors for keywords extracted from question data.

According to one aspect of the present invention described above, by building an artificial intelligence-based chatbot model, an automated response service can be provided when customer inquiries about a specific product are made, thereby reducing the cost and time required for customer service (CS). there is.

Figure 1 is a diagram illustrating the schematic configuration of a customer response system using artificial intelligence according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating the specific configuration of the customer service management server shown in FIG. 1.

The detailed description of the present invention described below refers to the accompanying drawings, which show by way of example specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different from one another but are not necessarily mutually exclusive. For example, specific shapes, structures and characteristics described herein may be implemented in one embodiment without departing from the spirit and scope of the invention. Additionally, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description that follows is not intended to be taken in a limiting sense, and the scope of the invention is limited only by the appended claims, together with all equivalents to what those claims assert, if properly described. Similar reference numbers in the drawings refer to identical or similar functions across various aspects.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

Figure 1 is a diagram illustrating the schematic configuration of a customer response system using artificial intelligence according to an embodiment of the present invention.

The customer response system using artificial intelligence according to the present invention reduces the cost and time required for customer service (CS) by building an artificial intelligence-based chatbot model and providing an automated response service when customer inquiries about a specific product. The purpose is to

Specifically, the customer response system using artificial intelligence according to the present invention includes a consumer terminal 100 and a customer service management server 200.

The consumer terminal 100 is a terminal owned by a user who wishes to receive chatbot services using the customer response system using artificial intelligence according to the present invention, and is capable of wired and wireless communication with external devices and capable of inputting, outputting, and processing information. It may be an electronic device such as a smartphone, PC, laptop, tablet PC, wearable device, etc.

For example, the consumer terminal 100 may be a terminal owned by a consumer who has purchased groceries, but is not limited thereto and may further include consumers who purchase various products in addition to groceries.

The customer service management server 200 builds an artificial intelligence-based chatbot model and provides an automated response service when customer inquiries about a specific product.

Figure 2 is a diagram showing the specific configuration of this customer service management server 200.

As shown, the customer service management server 200 includes a communication unit 210, a chatbot management unit 220, a keyword extraction unit 230, a score calculation unit 240, and an answer data generation unit 250.

First, the chatbot management unit 220 builds a chatbot-based conversational interface using the consumer terminal owned by the customer who purchased the product.

The communication unit 210 communicates with the consumer terminal through a chatbot-based conversational interface built by the chatbot management unit 220. In one embodiment, the communication unit 210 receives question data related to a product through the interactive interface.

The keyword extraction unit 230 extracts at least one keyword from the question data using a pre-trained artificial neural network.

In one embodiment, the keyword extractor 230 may learn training data using the Word2Vec algorithm to construct a neural network that extracts context information for input data in order to extract at least one keyword from data.

The Word2Vec algorithm may include a neural network language model (NNLM). A neural network language model is basically a neural network consisting of an input layer, projection layer, hidden layer, and output layer. Neural network language models are used to vectorize words. Since the neural network language model is a known technology, a more detailed description will be omitted.

The Word2vec algorithm is for text mining and is an algorithm that determines proximity by looking at the front and back relationships between each word. The Word2vec algorithm is an unsupervised learning algorithm. As the name indicates, the Word2vec algorithm can be a quantitative technique that expresses the meaning of words in vector form. The Word2vec algorithm can express each word as a vector in a space of about 200 dimensions. Using the Word2vec algorithm, you can obtain the vector corresponding to the word for each word.

The Word2vec algorithm can enable dramatic improvements in precision in the field of natural language processing compared to other conventional algorithms. Word2vec can learn the meaning of words using the relationships between words and adjacent words in sentences in the input corpus. The Word2vec algorithm is based on an artificial neural network and starts from the premise that words with the same context have close meanings. The Word2vec algorithm learns through text documents, and for one word, other words that appear nearby (about 5 to 10 words before or after it) are taught to the artificial neural network as related words. Because words with related meanings are likely to appear close together in a document, two words may have increasingly closer vectors during repeated learning.

The learning methods of the Word2vec algorithm include the CBOW (Continuous Bag Of Words) method and the skip-gram method. The CBOW method predicts the target word using the context created by surrounding words. The skip-gram method predicts words that may come nearby based on one word. The skip-gram method is known to be more accurate in large datasets.

Therefore, in the embodiment of the present invention, the Word2vec algorithm using the skip-gram method is used. For example, if learning is successfully completed through the Word2vec algorithm, similar words can be located nearby in a high-dimensional space. According to the Word2vec algorithm as described above, the closer the distribution of surrounding words in a learning document is, the more similar the calculated vector values can be, and words with similar calculated vector values can be considered similar. Since the Word2vec algorithm is a known technology, detailed descriptions related to vector value calculation will be omitted.

The keyword extraction unit 230 may input the question data received from the consumer terminal 100 into a neural network to infer the meaning of each word constituting the question data, and extract at least one keyword based on the inference result. You can.

The keyword extractor 230 may calculate the similarity between each word-specific embedding vector value and a plurality of reference vector values, and extract the reference vector value with the highest similarity to the evaluation result vector value among the plurality of reference vector values. At this time, Euclidean distance, cosine similarity, Tanimoto coefficient, etc. may be adopted as similarity calculation methods.

The keyword extraction unit 230 may list words in order of high similarity and extract the top n words as keywords.

Additionally, the keyword extractor 230 can train an artificial neural network, and can also use a trained artificial neural network. The processor can train or execute an artificial neural network stored in the memory, and the memory can store the trained artificial neural network. The electronic device that trains the artificial neural network and the electronic device that uses it may be the same, but may also be separate. Artificial intelligence is a computer system that implements some of the functions of the human brain and can learn, guess, and make decisions on its own. As learning progresses, the probability of extracting an answer may increase. Artificial intelligence can be composed of learning and elemental technologies using it. Artificial intelligence learning is an algorithmic technology that classifies and learns features based on input data, and elemental technologies may be technologies that implement some of the functions of the human brain using learning algorithms.

Artificial intelligence is a technology that makes it easy to approach problems that can have multiple answers probabilistically, and can logically and probabilistically infer the optimal cycle, method, and plan according to any input data. Artificial intelligence's inference technology can include judging input data, optimization predictions, knowledge and probability-based reasoning, and preference-based planning.

Artificial neural network is one of the learning algorithms in the machine learning field and is a program that implements the connection between neurons and synapses in the brain. Artificial neural networks can be created through a program to create a neural network structure and then learn it to have the desired function. Although there may be errors, it is possible to learn based on huge data and output appropriate output data with input data. It has the advantage of being able to obtain output data with statistically good results and being similar to human reasoning.

The keyword extraction unit 230 can build a query/metric dataset required for learning using an artificial intelligence algorithm built on big data, and may include a number of artificial neural networks that have been trained in advance for this purpose.

In this embodiment, the server may include a number of artificial neural networks that have been trained in advance to perform machine learning algorithms. With machine learning, you can output output data based on input data and use the results to learn on your own, which can improve your own data processing ability. Artificial neural networks can extract features based on input data, infer regularities, and output result data. As this process accumulates, the reliability of the result data increases.

In this embodiment, the artificial neural network may be an algorithm that outputs text data from at least one of the feature data of the shape, length, number, and elevation difference of an object recognized as text. Artificial neural networks can infer the best output data using big data as input data or after going through a processing process to organize unnecessary data.

ve Bayes Classifier, Hidden Markov Model, K-Means Clustering 등이 사용될 수 있다.Depending on the type of learning, artificial intelligence machine learning models include Super Viser Learning, UnSuper Viser Learning, Semisupervised Learning, and Reinforcement Learning. And machine learning algorithms include Decision Tree, K-Nearest Nearest Bor, Artificial Neural Network, Support Vector Machine, Ensemble Learning, Gradient Descent, Na

ve Bayes Classifier, Hidden Markov Model, K-Means Clustering, etc. can be used.

The artificial neural network may be pre-trained on various input values that may be included in the input data. An artificial neural network may be an artificial neural network that is learned according to reinforcement learning, one of the learning methods. Reinforcement learning is a method of gradually increasing the probability of obtaining the correct result by setting rewards and limits. Artificial neural networks can also be modeled based on Convelutional Neural Network (CNN) or Recurrent Neural Network (RNN).

In this way, the keyword extraction unit 230 can extract at least one keyword from question data and response data, which will be described later, using big data and artificial neural networks.

Afterwards, the communication unit 210 transmits the keywords extracted from the question data to an external search engine and an external chatbot connected to the network to collect response data from the external search engine and external chatbot, and collects response data based on the plurality of collected response data. Generates and transmits the answer data to the consumer terminal through an interactive interface.

To this end, the score calculation unit 240 calculates a CS score for each response data using Equation 1 below.

[Equation 1]

Here, CS is the CS score, da_i is the distance between the embedding vector for the ith keyword among the plurality of keywords extracted from the response data and the embedding vector for the preset reference word, and w_i is different for each of the plurality of keywords extracted from the response data. The set weight (first weight), da_s is the average distance between embedding vectors for a plurality of keywords extracted from the response data, t is the time taken from transmitting the keyword extracted from the question data to receiving the response data ( seconds), a is a weight (second weight) set differently for each external search engine and external chatbot, and dq is the average distance between embedding vectors for keywords extracted from question data.

For example, if ∑da_i of specific response data is 3, ∑w_i is 7, da_s is 1.5, t is 10, a is 5, and dq is 1, the CS score of the corresponding response data is calculated to be about 4.8.

In this way, the score calculation unit 240 calculates a CS score for each response data.

The answer data generator 250 collects response data from an external search engine and an external chatbot, generates response data based on a plurality of collected response data, and transmits the response data to the consumer terminal through an interactive interface. .

In one embodiment, the answer data generator 250 sets the response data with the highest calculated CS score as the answer data. Accordingly, the customer service management server 200 according to the present invention can provide an automated response service when a customer inquires about a specific product using the above-mentioned mathematical equation, and the reliability of the provided response can also be improved.

In some other embodiments, the customer service management server 200 determines whether the consumer terminal that transmitted the question data is a black consumer who intentionally generates malicious complaints or malicious question data after purchasing the product in order to take unfair advantage. By determining, answer data is not generated for question data received from a consumer terminal that is determined to be a black consumer.

To this end, the score calculation unit 240 calculates the black consumer index using Equation 2 below.

[Equation 2]

Here, U is the black consumer index, pw is the number of words included in the negative keyword list among words constituting the question data, negative p_i is the number of times the ith word included in the keyword list was used repeatedly in the question data, and c_i is negative. Weights are set differently for each word included in the keyword list, nw is the number of words included in the positive keyword list among words constituting the question data, y is the total number of words constituting the question data, k is the price of the product The constant set proportionally, d, is the difference between the number of words included in the negative keyword list and the number of words included in the positive keyword list among the words constituting the question data.

For example, the total number of words (y) constituting the response data is 30, among the words, 8 words are negative keywords, 3 words are positive keywords, When is 15, d is 5, and k is 6, the black consumer index U is calculated to be about 9 points, and the larger the value of u, the closer it can be judged to being a black consumer.

In this way, the customer service management server 200 calculates the black consumer index for each consumer terminal using the above-mentioned equation 2, thereby providing customer service by not replying to the question data received from the consumer terminal determined to be a black consumer. The cost and time required for (CS) can be dramatically reduced.

By reducing the reliability score of response data that is inappropriate for the answer, the reliability of the analysis results can be improved.

As such, the technology according to the present invention may be implemented as an application or in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., singly or in combination.

The program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable by those skilled in the computer software field.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. media), and hardware devices specifically configured to store and perform program instructions, such as ROM, RAM, flash memory, etc.

Examples of program instructions include not only machine language code such as that created by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform processing according to the invention and vice versa.

Although the above has been described with reference to embodiments, those skilled in the art will understand that various modifications and changes can be made to the present invention without departing from the spirit and space of the present invention as set forth in the following patent claims. You will be able to.

210: Department of Communications
220: Chatbot management department
230: Keyword extraction unit
240: Score calculation unit
250: Answer data generation unit

Claims (3)

In a customer response system using artificial intelligence, which includes a customer service management server that builds an artificial intelligence-based chatbot model and provides an automated response service when customer inquiries about a specific product,
The customer service management server,
Provide a chatbot-based interactive interface to the consumer terminal owned by the customer who purchased the product, receive question data related to the product through the interactive interface, and
Extracting at least one keyword from the question data using a pre-trained artificial neural network,
The extracted keywords are sent to a network-connected external search engine and external chatbot to collect response data from the external search engine and external chatbot, and response data is generated based on the plurality of collected response data to provide the above answers through an interactive interface. transmitting data to the consumer terminal,
The customer service management server,
The CS score is calculated for each response data using Equation 1 below, and the response data with the highest calculated CS score is set as the response data,
In order to determine whether the consumer terminal that transmitted the question data is a black consumer, the black consumer index for the question data is calculated using Equation 2 below, and the consumer is determined to be a black consumer based on the calculated black consumer index. A customer response system using artificial intelligence, characterized in that it does not respond to question data received from the terminal.

[Equation 1]

Here, CS is the CS score, da_i is the distance between the embedding vector for the ith keyword among the plurality of keywords extracted from the response data and the embedding vector for the preset reference word, and w_i is different for each of the plurality of keywords extracted from the response data. The set weight (first weight), da_s is the average distance between embedding vectors for multiple keywords extracted from response data, t is the response delay time, a is the weight (second weight) set differently for each external search engine and external chatbot. weight), dq is the average distance between embedding vectors for keywords extracted from question data.

[Equation 2]

Here, U is the black consumer index, pw is the number of words included in the negative keyword list among words constituting the question data, negative p_i is the number of times the ith word included in the keyword list was used repeatedly in the question data, and c_i is negative. Weights are set differently for each word included in the keyword list, nw is the number of words included in the positive keyword list among words constituting the question data, y is the total number of words constituting the question data, k is the price of the product The constant set proportionally, d, is the difference between the number of words included in the negative keyword list and the number of words included in the positive keyword list among the words constituting the question data. delete delete