KR20230084799A - Method for evaluating consultation quality - Google Patents

Abstract

A method for evaluating a voice consultation performed by a computing device including one or more processors is disclosed according to embodiments according to some embodiments of the present disclosure. The method includes the steps of acquiring voice counseling data generated in the process of performing a voice counseling session with the user terminal initiated in response to a call requested by the user terminal, and converting the voice counseling data into text. generating text counseling data and performing a first counseling quality evaluation process for evaluating the quality of counseling content through counseling words uttered by a counselor based on the text counseling data, and the voice counseling The method may include performing a second consultation quality evaluation process for evaluating the quality of a consultation attitude through at least one of consultation pronunciation, consultation accent, and consultation voice of the counselor based on the data.

Description

Method for Evaluating Consultation Quality {METHOD FOR EVALUATING CONSULTATION QUALITY}

The present disclosure relates to a method for improving consultation quality, and more particularly, to a method for evaluating consultation quality at a customer center.

A customer center (eg, a call center) of a company mainly receives customer complaints, inquiries, suggestions, and the like over the phone. Most of the consultations conducted in the call center are conducted through voice calls. Taking an insurance company as an example, due to the characteristics of the insurance company's business, the proportion of voice-based business processing through a call center is high, and professional counselors are conducting such voice-based consultation business for this purpose. During voice counseling, various types of counseling requests are received from various customers, and the quality of counseling experienced by customers may vary according to the counselor's own ability.

Due to the characteristics of an insurance company, the amount of business processing and consultation through the customer center is high, so it is an important factor to check the quality of counseling by a counselor to ensure that the correct business processing has been performed and to manage customer satisfaction with counseling. However, in general, since checking the quality of consultation requires a person to judge while listening to the contents of consultation, in reality, quality evaluation of sampled consultation is performed by selecting only a part of numerous customer consultation data. Furthermore, since this evaluation is also performed by a person, there is also a possibility that the objectivity of the evaluation may be degraded depending on the viewpoint and deviation of the evaluator.

Republic of Korea Patent Publication No. 2020-0005753

The present disclosure has been made in response to the above background art, and is to provide an automated counseling quality analysis algorithm capable of handling large-scale calls.

The technical problems of the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

According to some embodiments of the present disclosure to solve the above problems, a method for evaluating voice counseling performed by a computing device including one or more processors is disclosed. The method may include acquiring voice counseling data generated in a process of performing a voice counseling session with a user terminal initiated in response to a call requested by the user terminal; A first consultation quality evaluation process for generating text consultation data by performing text conversion on the voice consultation data and evaluating the quality of consultation contents through consultation words uttered by a counselor based on the text consultation data performing steps; and performing a second consultation quality evaluation process for evaluating the quality of the counseling attitude through at least one of counseling pronunciation, counseling intonation, and counseling voice of the counselor based on the voice counseling data.

In one embodiment, the first consultation quality evaluation process tokenizes the text consultation data into a plurality of tokens based on a natural language processing (NLP) algorithm and, based on the plurality of tokens, determines a value in the text consultation data. It is determined whether a specific keyword appears, and the tokenization may use a space or comma in the text counseling data as a stopword.

In one embodiment, it is determined whether or not a first predetermined keyword appears during a first time period in text counseling data, wherein the first keyword and the first time period may be mapped to each other and stored.

In an embodiment, the first consultation quality evaluation process may determine whether essential speech keywords appear in the text consultation data and determine whether a predefined negative keyword appears in the text consultation data. .

In an embodiment, the second consultation quality evaluation process may be performed when the first predetermined keyword does not appear during a first time period, when a required speech keyword does not appear in the text consultation data, or in the text consultation data. It may be characterized in that it is performed when a predefined negative keyword appears.

In an embodiment, the first consultation quality evaluation process may be performed when the quality of the counseling attitude of the counselor according to the second consultation quality evaluation process is less than a predetermined quality threshold.

In an embodiment, the second consultation quality evaluation process determines whether a quantitative value of counseling accent or a quantitative value of counseling voice of the agent during a first time period exceeds a threshold intensity, and determines whether the counseling voice exceeds the threshold intensity. By determining the duration of accent or counseling voice, the counselor's counseling attitude may be classified into one of one or more predetermined classes.

In one embodiment, the method further includes performing modulation on voice counseling data uttered by the counselor after the first time period when the classification result of the counseling attitude of the counselor is classified into a class related to deterioration in counseling quality. Further steps may be included.

In one embodiment, in the method, when the classification result of the counseling attitude of the counselor is classified into a class related to deterioration in counseling quality, the voice counseling data spoken by the counselor terminal that initiates the voice counseling session with the user terminal. The method may further include performing modulation on the voice counseling data before is transmitted to the user terminal.

In one embodiment, the step of modulating the voice consultation data may include modulation of the rhythm of the voice in the voice consultation data, modulation of the velocity of the voice in the voice consultation data, and tone in the voice consultation data. At least one of modulation of , modulation of the intensity of voice for each phrase in the voice counseling data, or modulation of the pitch of voice for each phrase in the voice counseling data may be performed.

In one embodiment, based on identification information of the user of the user terminal, a modulation method for the voice consultation data may be variably determined.

In one embodiment, the second consultation quality evaluation process evaluates the quality of the consultation pronunciation of the counselor based on a recognition score representing the recognition rate of the generated text consultation data, and the consultation attitude based on the quality of the pronunciation. quality can be assessed.

In one embodiment, the second counseling quality evaluation process evaluates the quality of the counseling pronunciation of the counselor based on a comparison between the generated text counseling data and the predetermined term data, and based on the quality of the pronunciation, the quality of the counseling pronunciation is evaluated. The quality of counseling attitudes can be evaluated.

In an embodiment, the second consultation quality evaluation process may evaluate the quality of the consultation attitude by obtaining at least one of a sigh, a predetermined negative onomatopoeia, or a predetermined negative intonation from voice consultation data.

In one embodiment, the second consultation quality evaluation process may include first voice consultation data uttered by a counselor and uttered by a user from the voice consultation data, based on the voice signal amplitude and the voice signal frequency of the voice consultation data. discriminating the second voice consultation data; identify an overlapping interval between the first audio consultation data and the second audio consultation data, based on a predetermined time frame of the audio consultation data; Further, based on the overlapping interval, it may be determined whether the counselor intervened in the speech process of the user of the user terminal.

In an embodiment, at least one of the first consultation quality evaluation process and the second consultation quality evaluation process sets a priority for one of the quality of the consultation content and the quality of the consultation attitude based on the past evaluation history of the counselor. A ranking may be determined, and based on the priorities, a consultation quality evaluation process may be performed.

In one embodiment, at least one of the first consultation quality evaluation process and the second consultation quality evaluation process extracts the speech content of the user of the user terminal from the voice consultation data, and obtains a predetermined positive value from the speech content of the user. It is possible to determine whether a keyword appears, and to perform a consultation quality evaluation process based on the predetermined positive keyword.

In one embodiment, the method may include generating feedback information to be provided to the counselor terminal based on at least one result of the result of the first consultation quality evaluation process and the result of the second consultation quality evaluation process. The feedback information includes a first evaluation related to counseling contents based on a result of the first consultation quality evaluation process and a second evaluation related to a consultation attitude based on a result of the second consultation quality evaluation process, The first evaluation is based on information on words that should have been uttered during the first time period, which is part of the course of the voice counseling session, information on words that should have been uttered during the course of the voice counseling session, and the voice counseling session. It includes information on words that should not have been uttered in the process of being performed, and the second evaluation is information about pronunciation of low quality and information about counseling accent whose quality is evaluated to be low in the process of performing the voice counseling session. , and information on counseling performance whose quality is evaluated as low.

A computer program stored in a computer readable storage medium according to an embodiment of the present disclosure is disclosed. The computer program includes instructions for causing a processor of a computing device to perform a method for evaluating voice counseling, the method comprising: a voice counseling session with a user terminal initiated in response to a call requested by the user terminal. acquiring voice counseling data generated during the process; A first consultation quality evaluation process for generating text consultation data by performing text conversion on the voice consultation data and evaluating the quality of consultation contents through consultation words uttered by a counselor based on the text consultation data performing steps; and performing a second consultation quality evaluation process for evaluating the quality of the counseling attitude through at least one of counseling pronunciation, counseling intonation, and counseling voice of the counselor based on the voice counseling data.

According to some embodiments of the present disclosure, an automated consultation quality analysis technique capable of handling large volume calls may be provided.

Effects obtainable in the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below. .

Various aspects are now described with reference to the drawings, wherein like reference numbers are used to collectively refer to like elements. In the following embodiments, for explanation purposes, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. However, it will be apparent that such aspect(s) may be practiced without these specific details.
1 illustratively illustrates a block diagram of a computing device, network, and user terminal in accordance with some embodiments of the present disclosure.
2 is a schematic diagram illustrating a network function according to some embodiments of the present disclosure.
3 exemplarily illustrates a method for evaluating voice counseling according to some embodiments of the present disclosure.
4 illustratively illustrates operation of a voice counseling process and a counseling quality evaluation model for evaluating voice counseling according to some embodiments of the present disclosure.
5 schematically illustrates a consultation quality evaluation model according to some embodiments of the present disclosure.
6 schematically illustrates a voice modulation model according to some embodiments of the present disclosure.
7 depicts a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Various embodiments are now described with reference to the drawings. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is apparent that these embodiments may be practiced without these specific details.

The terms “component,” “module,” “system,” and the like, as used herein, refer to a computer-related entity, hardware, firmware, software, a combination of software and hardware, or an execution of software. For example, a component may be, but is not limited to, a procedure, processor, object, thread of execution, program, and/or computer running on a processor. For example, both an application running on a computing device and a computing device may be components. One or more components may reside within a processor and/or thread of execution. A component can be localized within a single computer. A component may be distributed between two or more computers. Also, these components can execute from various computer readable media having various data structures stored thereon. Components may be connected, for example, via signals with one or more packets of data (e.g., data and/or signals from one component interacting with another component in a local system, distributed system) to other systems and over a network such as the Internet. data being transmitted) may communicate via local and/or remote processes.

In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or”. That is, unless otherwise specified or clear from the context, “X employs A or B” is intended to mean one of the natural inclusive substitutions. That is, X uses A; X uses B; Or, if X uses both A and B, "X uses either A or B" may apply to either of these cases. Also, the term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the listed related items.

Also, the terms "comprises" and/or "comprising" should be understood to mean that the features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, elements, and/or groups thereof. Also, unless otherwise specified or where the context clearly indicates that a singular form is indicated, the singular in this specification and claims should generally be construed to mean "one or more".

In addition, the term “at least one of A or B” should be interpreted as meaning “when only A is included”, “when only B is included”, and “when A and B are combined”.

Those skilled in the art will further understand that the various illustrative logical blocks, components, modules, circuits, means, logics, and algorithm steps described in connection with the embodiments disclosed herein may be implemented using electronic hardware, computer software, or combinations of both. It should be recognized that it can be implemented as To clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, configurations, means, logics, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented in hardware or as software depends on the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application. However, such implementation decisions should not be interpreted as causing a departure from the scope of this disclosure.

The description of the presented embodiments is provided to enable any person skilled in the art to use or practice the present invention. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure. The general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present invention is not limited to the embodiments presented herein. The present invention is to be accorded the widest scope consistent with the principles and novel features set forth herein.

1 is a block diagram of a computing device 100, a network 300, and a user terminal 200 for providing a counseling service according to an embodiment of the present disclosure.

The term "voice counseling" used in the present disclosure may refer to a service including any form of counseling that provides an answer to a part desired by a customer in connection with a call originating from a customer. For example, voice counseling may include counseling with customers in a call center. Hereinafter, a counseling service related to insurance will be described as an example, but the scope of such counseling is not limited to insurance, and counseling related to various domains may also be included within the scope of the present disclosure.

1 is a block diagram of a computing device 100, a network 300, and a user terminal 200 for providing or assisting a counseling service according to an embodiment of the present disclosure.

The configuration of the computing device 100 shown in FIG. 1 is only a simplified example. In one embodiment of the present disclosure, the computing device 100 may include other components for performing a computing environment of the computing device 100, and only some of the disclosed components may constitute the computing device 100. The computing device 100 in the present disclosure may refer to any type of device having processing capability. As an example, the computing device 100 may be a call center (eg, customer center) server. As another example, the computing device 100 may be a counselor terminal. As another example, the computing device 100 may be a customer terminal.

Computing device 100 may include any type of computer system or computer device, such as, for example, a microprocessor, mainframe computer, digital processor, portable device or device controller, and the like. This computing device 100 may include any type of server and/or user terminal. For example, the user terminal may include a PC (personal computer), a notebook (note book), a mobile terminal, a smart phone (smart phone), a tablet PC (tablet pc), etc. owned by the user, , can include all types of terminals that can access wired/wireless networks.

The computing device 100 may include a processor 110 , a storage unit 120 , and a communication unit 130 .

The processor 110 may include one or more cores, and includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU) of a computing device. unit), or any type of processor for data processing, data analysis, and/or deep learning. The processor 110 reads the computer program stored in the storage unit 120 to evaluate the quality of voice counseling according to an embodiment of the present disclosure and/or to provide feedback on the quality evaluation of voice counseling. can be performed. According to an embodiment of the present disclosure, the processor 110 may perform an operation for learning a neural network. The processor 110 processes input data for learning in machine learning or deep learning (DL: preprocessing such as denoising of input data), extracts features from input data, calculates errors, and backpropagates. Calculations for neural network learning, such as weight update of the neural network using backpropagation, may be performed. At least one of the CPU, GPGPU, and TPU of the processor 110 may process learning of the network function. For example, the CPU and GPGPU can process learning of network functions and data classification using network functions. In addition, in an embodiment of the present disclosure, the learning of a network function and data classification using a network function may be processed by using processors of a plurality of computing devices together. In addition, the computer program executed in the computing device 100 according to an embodiment of the present disclosure may be a CPU, GPGPU or TPU executable program.

For example, the processor 110 of the present disclosure may remove ambient noise (noise) that may occur on a call from voice counseling data and perform tokenization to divide the voice counseling data into a plurality of tokens. and a pre-processing operation of performing one-hot encoding on each of the tokenized tokens. The processor 110 may obtain a result of evaluating the quality of the voice counseling data by inputting the preprocessed input data to the input layer.

According to an embodiment of the present disclosure, the storage unit 120 may store any type of information generated or determined by the processor 110 and any type of information received by the communication unit 150. In the present disclosure, the storage unit 120 may be used interchangeably with a memory. For example, the storage unit 120 may store voice counseling data, text counseling data, counseling content quality evaluation data, counseling attitude quality evaluation data, counselor past history data, and/or counselor identification information.

According to an embodiment of the present disclosure, the storage unit 120 may be a flash memory type, a hard disk type, a multimedia card micro type, or a card type memory (eg SD or XD memory, etc.), RAM (Random Access Memory, RAM), SRAM (Static Random Access Memory), ROM (Read-Only Memory, ROM), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Memory) Read-Only Memory), a magnetic memory, a magnetic disk, and an optical disk may include at least one type of storage medium. The computing device 100 may operate in relation to a web storage that performs the storage function of the storage unit 120 on the Internet. The above description of the storage unit 120 is only an example, and the present disclosure is not limited thereto.

The network 300 according to an embodiment of the present disclosure may include any wired/wireless communication network capable of transmitting and receiving data and signals of any type and may be included in the network 300 represented in the present disclosure.

The communication unit 130 according to an embodiment of the present disclosure may communicate with any type of other computing device (eg, the user terminal 200A) through the network 300 .

The user terminal 200A according to an embodiment of the present disclosure is a PC (personal computer), a notebook (note book), a mobile terminal, a smart phone (smart phone), a tablet PC (tablet pc) owned by the user. ), etc., and may include all types of terminals capable of accessing wired/wireless networks.

The counselor terminal 200B according to an embodiment of the present disclosure may include a PC, a laptop computer, a mobile terminal, a smartphone, a tablet PC, etc. corresponding to the counselor, and performs functions corresponding to the user terminal 200A described above. can do. Additionally, the counselor terminal 200B may have additional functions not provided by the user terminal 200A through communication with the computing device 100 . For example, feedback information on counseling quality may be delivered from the computing device 100 to the counselor terminal 200B.

2 is a schematic diagram illustrating a network function according to an embodiment of the present disclosure.

Throughout this specification, artificial intelligence-based models, artificial intelligence models, computational models, neural networks, network functions, and neural networks may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes. Nodes (or neurons) constituting neural networks may be interconnected by one or more links.

Before input data is input to an input layer of an artificial intelligence model, preprocessing techniques according to an embodiment of the present disclosure may be performed. For example, the computing device 100 may perform pre-processing of tokenizing text counseling data. Such tokenization may use a space or comma in the text counseling data as a stopword. In another example, the computing device 100 may perform preprocessing in a manner of removing noise from voice counseling data. In another example, the computing device 100 may perform preprocessing of a method of distinguishing a speaker in voice counseling data.

In a neural network, one or more nodes connected through a link may form a relative relationship of an input node and an output node. The concept of an input node and an output node is relative, and any node in an output node relationship with one node may have an input node relationship with another node, and vice versa. As described above, an input node to output node relationship may be created around a link. More than one output node can be connected to one input node through a link, and vice versa.

In a relationship between an input node and an output node connected through one link, the value of data of the output node may be determined based on data input to the input node. Here, a link interconnecting an input node and an output node may have a weight. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. An output node value may be determined based on the weight.

As described above, in the neural network, one or more nodes are interconnected through one or more links to form an input node and output node relationship in the neural network. Characteristics of the neural network may be determined according to the number of nodes and links in the neural network, an association between the nodes and links, and a weight value assigned to each link. For example, when there are two neural networks having the same number of nodes and links and different weight values of the links, the two neural networks may be recognized as different from each other.

A neural network may be composed of a set of one or more nodes. A subset of nodes constituting a neural network may constitute a layer. Some of the nodes constituting the neural network may form one layer based on distances from the first input node. For example, a set of nodes having a distance of n from the first input node may constitute n layers. The distance from the first input node may be defined by the minimum number of links that must be passed through to reach the corresponding node from the first input node. However, the definition of such a layer is arbitrary for explanation, and the order of a layer in a neural network may be defined in a method different from the above. For example, a layer of nodes may be defined by a distance from a final output node.

An initial input node may refer to one or more nodes to which data is directly input without going through a link in relation to other nodes among nodes in the neural network. Alternatively, in a relationship between nodes based on a link in a neural network, it may mean nodes that do not have other input nodes connected by a link. Similarly, the final output node may refer to one or more nodes that do not have an output node in relation to other nodes among nodes in the neural network. Also, the hidden node may refer to nodes constituting the neural network other than the first input node and the last output node.

In the neural network according to an embodiment of the present disclosure, the number of nodes in the input layer may be the same as the number of nodes in the output layer, and the number of nodes decreases and then increases again as the number of nodes progresses from the input layer to the hidden layer. can In addition, the neural network according to another embodiment of the present disclosure may be a neural network in which the number of nodes of the input layer may be less than the number of nodes of the output layer and the number of nodes decreases as the number of nodes increases from the input layer to the hidden layer. there is. In addition, the neural network according to another embodiment of the present disclosure is a neural network in which the number of nodes in the input layer may be greater than the number of nodes in the output layer, and the number of nodes increases as the number of nodes increases from the input layer to the hidden layer. can A neural network according to another embodiment of the present disclosure may be a neural network in the form of a combination of the aforementioned neural networks.

A deep neural network (DNN) may refer to a neural network including a plurality of hidden layers in addition to an input layer and an output layer. Deep neural networks can reveal latent structures in data. That is, pictures, texts, videos, voices, textual data of voices, subliminal structure of music (e.g. what objects are in a picture, what is the content and emotion of the text, what is the content and emotion of the voice, etc. ) can be identified. Deep neural networks include convolutional neural networks (CNNs), recurrent neural networks (RNNs), auto encoders, generative adversarial networks (GANs), and restricted boltzmann machines (RBMs). machine), a deep belief network (DBN), a Q network, a U network, a Siamese network, a Generative Adversarial Network (GAN), and the like. The description of the deep neural network described above is only an example, and the present disclosure is not limited thereto.

In one embodiment of the present disclosure, the network function may include an autoencoder. An autoencoder may be a type of artificial neural network for outputting output data similar to input data. An auto-encoder may include at least one hidden layer, and an odd number of hidden layers may be disposed between input and output layers. The number of nodes of each layer may be reduced from the number of nodes of the input layer to an intermediate layer called the bottleneck layer (encoding), and then expanded symmetrically with the reduction from the bottleneck layer to the output layer (symmetrical to the input layer). Autoencoders can perform non-linear dimensionality reduction. The number of input layers and output layers may correspond to dimensions after preprocessing of input data. In the auto-encoder structure, the number of hidden layer nodes included in the encoder may decrease as the distance from the input layer increases. If the number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) is too small, a sufficient amount of information may not be conveyed, so more than a certain number (e.g., more than half of the input layer, etc.) ) may be maintained.

The neural network may be trained using at least one of supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Learning of the neural network may be a process of applying knowledge for the neural network to perform a specific operation to the neural network.

A neural network can be trained in a way that minimizes output errors. In the learning of the neural network, the learning data is repeatedly input into the neural network, the output of the neural network for the training data and the error of the target are calculated, and the error of the neural network is transferred from the output layer of the neural network to the input layer in the direction of reducing the error. It is a process of updating the weight of each node of the neural network by backpropagating in the same direction. In the case of teacher learning, the learning data in which the correct answer is labeled is used for each learning data (ie, the labeled learning data), and in the case of comparative teacher learning, the correct answer may not be labeled in each learning data. That is, for example, learning data in the case of teacher learning regarding data classification may be data in which each learning data is labeled with a category. Labeled training data is input to a neural network, and an error may be calculated by comparing an output (category) of the neural network and a label of the training data. As another example, in the case of comparative history learning for data classification, an error may be calculated by comparing input learning data with a neural network output. The calculated error is back-propagated in a reverse direction (ie, from the output layer to the input layer) in the neural network, and the connection weight of each node of each layer of the neural network may be updated according to the back-propagation. The amount of change in the connection weight of each updated node may be determined according to a learning rate. The neural network's computation of input data and backpropagation of errors can constitute a learning cycle (epoch). The learning rate may be applied differently according to the number of iterations of the learning cycle of the neural network. For example, a high learning rate may be used in the early stage of neural network training to increase efficiency by allowing the neural network to quickly obtain a certain level of performance, and a low learning rate may be used in the late stage to increase accuracy.

In neural network learning, generally, training data can be a subset of real data (ie, data to be processed using the trained neural network), and therefore, errors for training data are reduced, but errors for real data are reduced. There may be incremental learning cycles. Overfitting is a phenomenon in which errors for actual data increase due to excessive learning on training data. For example, a phenomenon in which a neural network that has learned a cat by showing a yellow cat does not recognize that it is a cat when it sees a cat other than yellow may be a type of overfitting. Overfitting can act as a cause of increasing the error of machine learning algorithms. Various optimization methods can be used to prevent such overfitting. To prevent overfitting, methods such as increasing the training data, regularization, inactivating some nodes in the network during learning, and using a batch normalization layer should be applied. can

A computer readable medium storing a data structure according to an embodiment of the present disclosure is disclosed.

Data structure can refer to the organization, management, and storage of data that enables efficient access and modification of data. Data structure may refer to the organization of data to solve a specific problem (eg, data retrieval, data storage, data modification in the shortest time). A data structure may be defined as a physical or logical relationship between data elements designed to support a specific data processing function. A logical relationship between data elements may include a connection relationship between user-defined data elements. A physical relationship between data elements may include an actual relationship between data elements physically stored in a computer-readable storage medium (eg, a persistent storage device). The data structure may specifically include a set of data, a relationship between data, and a function or command applicable to the data. Through an effectively designed data structure, a computing device can perform calculations while using minimal resources of the computing device. Specifically, the computing device can increase the efficiency of operation, reading, insertion, deletion, comparison, exchange, and search through an effectively designed data structure.

The data structure can be divided into a linear data structure and a non-linear data structure according to the shape of the data structure. A linear data structure may be a structure in which only one data is connected after one data. Linear data structures may include lists, stacks, queues, and decks. A list may refer to a series of data sets in which order exists internally. The list may include a linked list. A linked list may be a data structure in which data are connected in such a way that each data is connected in a single line with a pointer. In a linked list, a pointer can contain information about connection to the next or previous data. A linked list can be expressed as a singly linked list, a doubly linked list, or a circular linked list depending on the form. A stack can be a data enumeration structure that allows limited access to data. A stack can be a linear data structure in which data can be processed (eg, inserted or deleted) at only one end of the data structure. The data stored in the stack may be a LIFO-Last in First Out (Last in First Out) data structure. A queue is a data listing structure that allows limited access to data, and unlike a stack, it can be a data structure (FIFO-First in First Out) in which data stored later comes out later. A deck can be a data structure that can handle data from either end of the data structure.

The nonlinear data structure may be a structure in which a plurality of data are connected after one data. The non-linear data structure may include a graph data structure. A graph data structure can be defined as a vertex and an edge, and an edge can include a line connecting two different vertices. A graph data structure may include a tree data structure. The tree data structure may be a data structure in which one path connects two different vertices among a plurality of vertices included in the tree. That is, it may be a data structure that does not form a loop in a graph data structure.

Throughout this specification, computational model, neural network, network function, and neural network may be used interchangeably. Hereinafter, a neural network is unified and described. The data structure may include a neural network. And the data structure including the neural network may be stored in a computer readable medium. The data structure including the neural network may also include preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and neural network It may include a loss function for learning of . A data structure including a neural network may include any of the components described above. That is, the data structure including the neural network includes preprocessed data for processing by the neural network, data input to the neural network, weights of the neural network, hyperparameters of the neural network, data obtained from the neural network, activation function associated with each node or layer of the neural network, and neural network. It may be configured to include all or any combination thereof, such as a loss function for learning of . In addition to the foregoing configurations, the data structure comprising the neural network may include any other information that determines the characteristics of the neural network. In addition, the data structure may include all types of data used or generated in the computational process of the neural network, but is not limited to the above. A computer readable medium may include a computer readable recording medium and/or a computer readable transmission medium. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network includes one or more nodes.

The data structure may include data input to the neural network. For example, the data structure may include data obtained as a result of one-hot encoding. A data structure including data input to the neural network may be stored in a computer readable medium. Data input to the neural network may include training data input during a neural network learning process and/or input data input to a neural network that has been trained. Data input to the neural network may include pre-processed data and/or data subject to pre-processing. Pre-processing may include a data processing process for inputting data to a neural network. Accordingly, the data structure may include data subject to pre-processing and data generated by pre-processing. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure may include the weights of the neural network. (In this specification, weights and parameters may be used in the same meaning.) Also, a data structure including weights of a neural network may be stored in a computer readable medium. A neural network may include a plurality of weights. The weight may be variable, and may be changed by a user or an algorithm in order to perform a function desired by the neural network. For example, when one or more input nodes are interconnected by respective links to one output node, the output node is set to a link corresponding to values input to input nodes connected to the output node and respective input nodes. A data value output from an output node may be determined based on the weight. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

As a non-limiting example, the weights may include weights that are varied during neural network training and/or weights for which neural network training has been completed. The variable weight in the neural network learning process may include a weight at the time the learning cycle starts and/or a variable weight during the learning cycle. The weights for which neural network learning has been completed may include weights for which learning cycles have been completed. Accordingly, the data structure including the weights of the neural network may include a data structure including weights that are variable during the neural network learning process and/or weights for which neural network learning is completed. Therefore, it is assumed that the above-described weights and/or combinations of weights are included in the data structure including the weights of the neural network. The foregoing data structure is only an example, and the present disclosure is not limited thereto.

The data structure including the weights of the neural network may be stored in a computer readable storage medium (eg, a memory or a hard disk) after going through a serialization process. Serialization can be the process of converting a data structure into a form that can be stored on the same or another computing device and later reconstructed and used. A computing device may serialize data structures to transmit and receive data over a network. The data structure including the weights of the serialized neural network may be reconstructed on the same computing device or another computing device through deserialization. The data structure including the weights of the neural network is not limited to serialization. Furthermore, the data structure including the weights of the neural network is a data structure for increasing the efficiency of operation while minimizing the resource of the computing device (for example, B-Tree, Trie, m-way search tree, AVL tree, Red-Black Tree). The foregoing is only an example, and the present disclosure is not limited thereto.

The data structure may include hyper-parameters of the neural network. Also, the data structure including the hyperparameters of the neural network may be stored in a computer readable medium. A hyperparameter may be a variable variable by a user. Hyperparameters include, for example, learning rate, cost function, number of learning cycle iterations, weight initialization (eg, setting the range of weight values to be targeted for weight initialization), hidden unit number (eg, the number of hidden layers and the number of nodes in the hidden layer). The foregoing data structure is only an example, and the present disclosure is not limited thereto.

3 illustratively illustrates a method for evaluating consultation quality according to some embodiments of the present disclosure. The flowchart shown in FIG. 3 is shown for illustrative purposes, and additional steps may be present or some of the steps included in the flowchart shown in FIG. 3 may be omitted.

The method of FIG. 3 may be performed by, for example, the computing device 100 of the present disclosure.

The computing device 100 may acquire voice counseling data generated in the process of performing a voice counseling session with the user terminal 200A initiated in response to a call requested by the user terminal (310).

For example, the computing device 100 may recognize an incoming call from the user terminal 200A. A counseling session for the recognized call may be initiated between the user terminal 200A and the computing device 100 or between the user terminal 200A and the counselor terminal 200B. The expression that a counseling session starts may mean that voice counseling between the counselor terminal 200B or the computing device 100 and the user terminal 200A is started. For example, initiation of voice counseling between the computing device 100 and the user terminal 200A means that a guide on ARS is being delivered to the user terminal 200A after the call is connected and before actual consultation with the counselor. can mean As another example, the start of voice counseling between the counselor terminal 200B and the user terminal 200A may mean that voice counseling between the user and the counselor of the customer center is started after the guidance of the automatic response system (ARS) ends. there is.

In one embodiment of the present disclosure, the computing device 100 may obtain voice data of a user and/or a counselor uttered during a voice counseling process. For example, the computing device 100 may acquire voice counseling data uttered during a voice counseling process by transmitting voice data of the user and/or the counselor to the computing device 100 .

In one embodiment of the present disclosure, the computing device 100 performs text conversion on the acquired voice counseling data to generate text counseling data and, based on the text counseling data, converts counseling words spoken by a counselor. Through this, a first consultation quality evaluation process for evaluating the quality of consultation contents may be performed (320).

The computing device 100 may generate text counseling data from the acquired voice counseling data using a speech to text (STT) technique. In performing the STT technique, a speech recognition technique including a process of converting the contents into text after a computer interprets speech language may be utilized. As an example of a speech recognition technology, a language model based on a Hidden Markov Model (HMM) may be considered.

The computing device 100 may evaluate the quality of counseling content through counseling words uttered by a counselor. The evaluation of the counseling content may include quality evaluation of the content uttered during the voice counseling process, such as whether a specific word was uttered at a specific point in the counselor's uttered voice, whether required words were uttered, and/or whether negative words were uttered. can

Based on the voice counseling data, the computing device 100 may perform a second counseling quality evaluation process for evaluating the quality of counseling attitude through at least one of the counselor's pronunciation, counseling accent, and counseling voice (330).

Based on the voice counseling data (or based on the text counseling data), the computing device 100 determines the counselor's pronunciation, the counselor's intonation, the counselor's volume, the pitch of the counselor's voice by phrase, the strength of the counselor's voice by phrase, The quality of the counselor's counseling attitude in the voice counseling process, such as whether the counselor intervenes and/or the tone of the counselor's voice, can be evaluated.

In a further embodiment of the present disclosure, the first consultation quality evaluation process and/or the second consultation quality evaluation process may be further based on feedback words uttered from a user terminal conducting a consultation session with a counselor. For example, when positive words (eg, "thank you very much", "it was very helpful", "thank you for your kind consultation", "the best", etc.) are included in the words spoken from the user terminal, the appearance of these positive words may be used as a positive factor for the counselor's content or attitude toward voice counseling in the first consultation quality evaluation process and/or the second consultation quality evaluation process. As another example, when negative words (eg, "worst", "not so good", etc.) are included in words uttered from the user terminal, the appearance of these negative words determines whether the first consultation quality evaluation process and/or the second consultation In the quality evaluation process, it can be used as a negative factor for the counselor's content or attitude toward voice counseling.

In a further embodiment of the present disclosure, the first consultation quality evaluation process and the second consultation quality evaluation process may be performed dependently on each other. For example, the second consultation quality evaluation process may be performed when the quality of the consultation contents in the first consultation quality evaluation process is less than a specific threshold value. For example, the first consultation quality evaluation process may be performed when the quality of the consultation attitude in the second consultation quality evaluation process is less than a specific threshold value. Through this method, since both the first and second counseling quality evaluations can be performed only for voice counseling that does not meet a specific level, computing resources can be efficiently utilized in evaluating counseling quality.

4 illustratively illustrates operation of a voice counseling process and a counseling quality evaluation model for evaluating voice counseling according to some embodiments of the present disclosure.

The voice counseling process 400A in FIG. 4 may be performed between the user terminal 200A and the counselor terminal 200B (or the computing device 100). The computing device 100 may recognize that a counseling session in the voice counseling process 400A is initiated (410). In the voice counseling session, voices spoken from the counselor terminal 200B and the user terminal 200A may flow in (420).

The consultation quality evaluation process model 400B may refer to an algorithm that may be used by or exists in the computing device 100 .

In FIG. 4, the first consultation quality evaluation process 460A and the second consultation quality evaluation process 460B within the consultation quality evaluation model 400B are shown as separate processes for convenience of explanation, but are integrated into one process. It will be clear to those skilled in the art that it can be implemented as a process. Furthermore, the STT process of converting the voice consultation data 440 into the text consultation data 450 may also be included in the first consultation quality evaluation process 460A and/or the second consultation quality evaluation process 460B.

The computing device 100 may analyze the incoming voice consultation data 440 using the consultation quality evaluation model 400B. Voice counseling data 440 may be converted into text counseling data 450 using the STT technique. The consultation quality evaluation model 400B may refer to a model that generates consultation content quality evaluation data 470A and consultation attitude quality evaluation data 470B using the voice consultation data 440 and the text consultation data 450. there is.

In one embodiment, the consultation quality evaluation model 400B may be a speech recognition model. In this example, the consultation quality evaluation model 400B is an 'acoustic model' for automatic speech recognition including a Hidden Markov model and/or a Gaussian Mixture model, and a BERT model, GPT series model (GPT-2, GPT-3 etc.), a 'language model' such as an SLM-based model and/or a Transformer model. As an example, the first consultation quality evaluation process 460A may output the consultation content quality evaluation data 470A from the text consultation data 450 using at least one of the exemplary models described above. As such, the speech recognition model may be implemented as a combination of an acoustic model and a language model. Textualization of a recognized voice may be implemented through a combination of an acoustic model and a language model. Additionally, the computing device 100 may evaluate the counselor's counseling attitude by performing intonation analysis such as timbre, pitch, and/or strength of each phrase using the acoustic model.

The above-described acoustic model can be used when generating text counseling data 450 from voice counseling data 440 . Also, the acoustic model may be used in the second consultation quality evaluation process for evaluating the quality of the consultation attitude. For example, the acoustic model may extract acoustic features of an incoming voice using a feature extraction technique such as Wav2Vec or SincNet, and then derive a result based on the extracted features.

In one embodiment, the consultation quality evaluation model 400B may include a language model based on natural language processing (NLP). In this example, the consultation quality evaluation model 400B may include a language model such as a BERT model, a GPT series model (GPT-2, GPT-3, etc.), an SLM-based model, and/or a Transformer model. As an example, the first consultation quality evaluation process 460A may output the consultation content quality evaluation data 470A from the text consultation data 450 using at least one of the exemplary models described above.

The consultation content quality evaluation data 470A includes, for example, an evaluation of whether a specific keyword appears at a specific time point, an evaluation of whether an essential speech keyword appears, an evaluation of whether a negative keyword appears, and an evaluation of whether a mandatory speech keyword appears. Evaluation of the number of occurrences, evaluation of whether negative keywords appear, evaluation of whether positive words appear and count of occurrences in feedback from user terminals, and/or appearances of negative words in feedback from user terminals It may include data related to evaluation of existence and number of appearances.

In one embodiment, the first consultation quality evaluation process 460A may perform pre-processing on the text consultation data 450 . For example, the first consultation quality evaluation process 460A may perform tokenization, which is an operation of dividing the text consultation data 450 into units called tokens. A unit of a token may be a morpheme unit, a word unit, a word unit, and/or a sentence unit tokenization, and may be different according to an implementation mode. For example, when the token unit is set in units of words (keywords), the computing device 100 can easily determine whether essential keywords, positive keywords, and/or negative keywords appear in a voice counseling process. The text counseling data 450 may include temporal information of spoken voice data. For example, it is assumed that voice counseling data 440, "Insurance claim, a beneficiary's identification card, and a copy of the beneficiary's bankbook are required" is converted into text counseling data 450. In this assumption, there may be a time interval between the first expression "at the time of insurance claim", the second expression "insurance claim", the third expression "beneficiary's identification card", and the fourth expression "copy of the beneficiary's bankbook". . The computing device 100 may generate text counseling data 450 from the voice counseling data 440 by reflecting this temporal interval. On the text counseling data 450, such temporal intervals may be expressed as separators such as spaces, semicolons, and/or commas. Therefore, when the token unit is a word unit, the text counseling data 450 may be divided based on the above-mentioned classifiers in the tokenization process, and tokenization may be implemented in a word or keyword unit.

In a further embodiment, tokenization of units of morphemes may be considered. Morphemes can be largely divided into independent morphemes that can be used on their own regardless of affixes, endings, or postpositions, and dependent morphemes that are used in combination with other morphemes. A free morpheme may itself be a word, and may include, for example, nouns, pronouns, and interjections. Dependent morphemes may include affixes, endings, and particles. Accordingly, the computing device 100 may tokenize the independent morphemes into tokens after removing dependent morphemes, such as affixes, endings, and postpositions, extracting only independent morphemes, by performing tokenization in units of morphemes.

In a further embodiment, computing device 100 may perform vectorization on tokens, such as words or morphemes, by applying one-hot encoding to tokenized text consultation data 450 . Vectors generated according to one-hot encoding may be used for quality evaluation of the content of voice counseling within the counseling quality evaluation model 400B.

In one embodiment, based on the consultation content quality evaluation data 470A generated through the first consultation quality evaluation process 460A and the consultation attitude quality evaluation data 470B generated through the second consultation quality evaluation process 460B. Thus, feedback data 480 may be generated. The feedback data 480 is data for providing feedback on the counseling quality of the counselor to the counselor, information on words to be uttered during the first time period, which is a part of the course of the voice counseling session, and information about the voice counseling session being performed. It includes information on words that should have been uttered during the course and information about words that should not have been uttered during the course of the voice counseling session, and information on low-quality pronunciation and low-quality pronunciation during the course of the voice counseling session. Information on the evaluated counseling accent and information on the counseling voice whose quality is evaluated as low may be included.

In one embodiment, the feedback data 480 may be delivered to the counselor terminal 200B performing (or completing) the voice counseling process 400A. Through this feedback data 480, the voice counseling process 400A proceeds with feedback-based voice counseling (430), and accordingly, the quality of the voice counseling can be improved.

5 schematically illustrates a consultation quality evaluation model 400B according to some embodiments of the present disclosure.

The first consultation quality evaluation process 460A and the second consultation quality evaluation process 460B may use at least one of the voice consultation data 510 and the text consultation data 520 as an input.

The computing device 100 determines whether a specific keyword appears 530 at a specific point in time through the first consultation quality evaluation process 460A, whether a required speech keyword appears 540, and whether a negative keyword appears 550. can output

The computing device 100 may perform an analysis of volume and/or intonation (560), an analysis of pronunciation quality (570), and an analysis of interruption (580) through the second consultation quality evaluation process (450B). there is.

In one embodiment, the computing device 100 tokenizes the text consultation data 520 into a plurality of tokens based on a natural language processing (NLP) algorithm in the first consultation quality evaluation process 460A, the plurality of tokens Based on these, whether a specific keyword appears in the text counseling data 520 can be determined. Tokenization here can be achieved by using a space or comma in the text counseling data 520 as a stopword.

In one embodiment, a specific keyword and specific time information may be mapped to each other and stored. The computing device 100 may determine whether a specific keyword in the text consultation data 520 appears within a specific time period in the first consultation quality evaluation process 460A. Through this, in the first consultation quality evaluation process 460A, whether a specific keyword is uttered at a specific time may be evaluated (530).

In one embodiment, the computing device 100 evaluates 540 whether essential speech keywords appear in the text counseling data 520 in the first counseling quality evaluation process 460A, and in the text counseling data 520 It may be evaluated 550 whether a predefined negative keyword has appeared. The computing device 100 may determine whether and how many positive keywords or negative keywords uttered by the customer appear in the text consultation data 520 in the first consultation quality evaluation process 460A, and based on this, the contents of the consultation are determined. quality can also be assessed.

In one embodiment, the first consultation quality evaluation process 460A may quantitatively express the quality of the counselor's counseling contents by combining result values for each of the above factors.

In an embodiment, the computing device 100 may perform the second consultation quality evaluation process 460B according to the result of the first consultation quality evaluation process 460A. The second consultation quality evaluation process 460B may be performed when the result of the first consultation quality evaluation process 460A does not meet a specific condition. For example, when a specific keyword does not appear during the first time period, when a required speech keyword does not appear in the text counseling data 520, or when a negative keyword appears in the text counseling data 520, second counseling A quality assessment process 460B may be performed. As another example, when the quantitative value of the quality of the consultation contents output in the first consultation quality evaluation process 460A does not satisfy a predetermined criterion, the second consultation quality evaluation process 460B may be performed.

In an embodiment, the computing device 100 may perform the first consultation quality evaluation process 460A according to the result of the second consultation quality evaluation process 460B. The first consultation quality evaluation process 460A may be performed when the result of the second consultation quality evaluation process 460B does not meet a specific condition. For example, the first consultation quality evaluation process 460A may be performed when the quality of the counselor's consultation attitude according to the second consultation quality evaluation process 460B is less than a predetermined quality threshold.

In one embodiment, the second consultation quality evaluation process 460B determines whether a quantitative value of counseling accent or a quantitative value of counseling voice of the agent during the first period of time exceeds a threshold intensity, and determines whether the threshold intensity is exceeded. By determining the duration of the counseling accent or the counseling voice, the counseling attitude of the counselor may be classified into one of one or more predetermined classes. For example, when it is determined that the counselor's voice exceeds a specific reference value for 10 seconds or longer, the counselor's counseling attitude may be classified as "aggressive". As an example, the class of the present disclosure may mean any qualitative data for classifying counseling attitudes, and may include aggressive, passive, cynical, critical, stable, authentic, comfortable, hurried, and the like.

In the present disclosure, intonation includes a state of pitch or a type of variation of uttered voice data, and may indicate a speaker's attitude other than the original meaning of uttered voice. For example, in a phrase, the melody of voice, which is indicated by the rise and fall of voice, the stress of voice (eg, the alternation of long and short sounds or the alternation of words with and without stress), the speed of voice indicating the speed and slowness of speech, the emotion or A tone that represents the color of the voice giving psychological nuance can be a component of intonation.

The "quantitative value of counseling accent" in the present disclosure may mean the degree of a class of counseling attitude appearing in the aforementioned accent. For example, when the speech rate is high, it may be classified as an “unstable” class, and a result of “the quantitative value for the unstable class is 80 points” may be generated according to the degree of the speech rate. The aforementioned classes may be divided into positive and negative classes, for example. The positive class may correspond to a class in which counseling quality is good, and the negative class may correspond to a class related to deterioration in counseling quality. If the counselor's voice counseling content corresponds to the negative class, the degree of the negative class or the quality of the counseling attitude may be determined by comparing the quantitative value with a specific threshold, and the same applies to the case of the positive class.

The second counseling quality evaluation process 460B may finally classify the counseling attitude of the counselor based on the quantitative value of the counseling accent and the duration of the quantitative value of the counseling accent. For example, if it is determined through the quantitative value of counseling intonation that “unstable class 70 points” lasted for 10 seconds and “stable class 60 points” lasted 1 minute, the class for the voice counseling was finally classified as stable class. can be classified as

The second consultation quality evaluation process 460B may finally classify the counselor's counseling attitude based on the quantitative value of the counseling voice and the duration of the quantitative value of the counseling voice. For example, if it is determined through the quantitative value of counseling voice that “excitement class 50 points” lasted 30 seconds and “stable class 60 points” lasted 10 seconds, the class for the voice counseling is finally excited class can be classified as

In one embodiment, the computing device 100 may evaluate the quality of the counseling pronunciation of the counselor through the second counseling quality evaluation process 460B. The quality of counseling pronunciation can also be related to the quality of counseling attitude. The second counseling quality evaluation process 460B may evaluate the quality of the counseling pronunciation of the counselor based on the recognition score representing the recognition rate of the text counseling data. As such, the quality of counseling attitude can be evaluated based on the quality of pronunciation. For example, if recognition of predetermined keywords such as "subscription to insurance" and "thousands of insurance money" are not accurate in the text counseling data converted through STT, it may be determined that the counselor's pronunciation is incorrect. The quality of the counselor's pronunciation may be determined based on a ratio of correctly recognized words to all spoken words. That is, the second consultation quality evaluation process 460B evaluates the quality of the consultation pronunciation of the counselor based on the comparison between the generated text consultation data and the predetermined term data, and determines the quality of the consultation attitude based on the quality of the pronunciation. can be evaluated.

In one embodiment, the computing device 100 obtains at least one of a sigh, a predetermined negative onomatopoeia, or a predetermined negative intonation from voice consultation data through the second consultation quality evaluation process 460B, thereby determining the quality of the consultation attitude. can be evaluated. For example, whether or not the sigh is uttered may be obtained by comparing frequency information and/or volume information on the timbre and/or intensity corresponding to the sigh sound with frequency information and/or volume information of the voice counseling data. . As another example, whether or not a negative intonation that may affect the other party's emotions may be determined based on the strength information for each phrase. As another example, whether a negative onomatopoeia appears may be determined by comparing a voice data pattern for onomatopoeia related to profanity and uttered voice counseling data.

In one embodiment, the computing device 100 may analyze the voice counseling data to determine whether the counselor's utterance is interrupted by the customer's utterance and determine the degree of interruption. In the second consultation quality evaluation process 460B, the computing device 100 may distinguish a speaker from voice consultation data. The computing device 100 may distinguish first voice consultation data uttered by a counselor and second voice consultation data uttered by a user from the voice consultation data based on the voice signal amplitude and the voice signal frequency of the voice consultation data. there is. In an additional embodiment, the computing device 100 extracts a speaker's unique voice characteristics from voice counseling data based on an artificial intelligence algorithm, a method of recognizing a speaker using a formant of vowels in voice counseling data, and the like. Various speaker recognition or speaker discrimination algorithms can be used.

The computing device 100 may identify an overlapping section between the first voice consultation data and the second audio consultation data based on a predetermined time frame of the voice consultation data. The computing device 100 may generate voice consultation data for each speaker and align the voice consultation data for each speaker based on time information allocated to the voice consultation data for each speaker. The computing device 100 determines a portion of the first voice consultation data and the second voice consultation data having overlapping speech times, based on the speech time information of the first voice consultation data and the speech time information of the second voice consultation data. parts can be identified. If it is determined that both the first voice consultation data and the second voice consultation data have been uttered at a specific time point or a specific time period (eg, 0.5 second, 1 second, 2 second, etc.), the corresponding time point or the corresponding time period is in the overlapping section. can be matched. The computing device 100 may determine whether the counselor intervenes in the user's utterance process of the user terminal based on the overlapping section. In addition, the computing device 100 may determine the number of times the counselor interrupts the user's utterance process of the user terminal based on the overlapping intervals (eg, based on the number of overlapping intervals).

The computing device 100 may evaluate the quality of the counselor's counseling attitude based on the number of times the counselor intervenes in the user's speech process or whether or not the counselor intervenes in the user's speech process.

In one embodiment of the present disclosure, the computing device 100 determines the counseling content based on the counselor's past evaluation history through at least one of the first consultation quality evaluation process 460A and the second consultation quality evaluation process 460B. You can prioritize either the quality of the counseling or the quality of the counseling attitude. For example, in the case of a specific counselor, there may be a past history of having a poor counseling attitude, even though he received good feedback on counseling contents because he satisfactorily adheres to predetermined counseling rules. In the case of such a counselor, the first consultation quality evaluation process 460A, which evaluates the quality of consultation contents, may have a lower importance than the second consultation quality evaluation process 460B, which evaluates the quality of the consultation attitude. As another example, it is assumed that a specific counselor receives good feedback about the counseling attitude (accent, etc.), but receives feedback that the quality of the counseling content is poor, such as not uttering a specific keyword. In this example, when counseling with a corresponding counselor is in progress, the computing device 100 performs the first consultation quality evaluation process 460A instead of the second consultation quality evaluation process 460B based on past counseling history information of the corresponding counselor. ) can be determined. The past evaluation history includes past counseling quality evaluation results of the counselor, for example, poor counseling content, good counseling content, poor counseling attitude, good counseling attitude, poor pronunciation, frequent interruptions, aggressive accent, and/or Alternatively, evaluation results according to a combination of at least one factor used in the counseling quality evaluation process, such as a high frequency of utterance of a negative keyword, may be included.

As such, the computing device 100 determines whether to perform the first consultation quality evaluation process 460A or the second consultation quality evaluation process when the counselor proceeds with the consultation by referring to the past evaluation history included in the identification information of the corresponding counselor. (460B) can be determined. In this way, when the counseling quality evaluation process to be applied to the counselor is adaptively changed based on the counselor's identification information or past counseling evaluation history information, computing resources required for the counselor's counseling quality evaluation can be efficiently managed.

In one embodiment of the present disclosure, the computing device 100 performs at least one of the first consultation quality evaluation process 460A or the second consultation quality evaluation process 460B based on the counselor's past evaluation history. , priorities of a plurality of evaluation items in the first consultation quality evaluation process 460A or the second consultation quality evaluation process 460B may be determined. Based on these priorities, the computing device 100 may perform a consultation quality evaluation process. The computing device 100 checks the counselor's past evaluation history and determines priorities for specific evaluation items or specific evaluation factors included in the evaluation history, thereby performing the first consultation quality evaluation process 460A or the second consultation When performing the quality evaluation process 460B, weights may be placed on the corresponding evaluation items. For example, it is assumed that the content that a specific counselor uttered negative keywords several times is included in the past evaluation history. During the counseling of the corresponding counselor, the computing device 100 refers to the past evaluation history, and when performing the first consultation quality evaluation process 460A, only determines whether a negative keyword appears or sets a higher weight when a negative keyword appears. Thus, the quality evaluation score of the consultation content can be output. As described above, when the counseling quality evaluation process is performed, the computing device 100 prioritizes a specific evaluation item based on the counselor's past evaluation history, so that a quality evaluation process more suitable for the counselor can be performed.

6 schematically illustrates a voice modulation model 600 according to some embodiments of the present disclosure.

According to an embodiment of the present disclosure, the computing device 100 may perform modulation on the agent's (or customer's) spoken voice using the voice modulation model 600 . For example, if the counseling attitude of the counselor is classified into a class related to counseling quality deterioration (eg, negative class), the computing device 100 may modify voice counseling data uttered by the counselor thereafter. can

As another example, when the counselor's counseling attitude is classified as a class related to counseling quality deterioration (eg, negative class), the computing device 100 uses a voice uttered from the counselor terminal 200B initiating a voice counseling session. Modulation may be performed on voice counseling data before the counseling data is transmitted to the user terminal 200A. In this case, immediate class classification is performed on the voice data uttered by the counselor, and modulation of the voice counseling data can be performed in real time.

Regarding voice modulation, while a voice counseling session is being performed, the computing device 100 may input voice data spoken from the counselor terminal 200B to the voice modulation model 600 .

The voice modulation model 600 in the present disclosure may refer to any type of model that modulates voice data according to certain conditions. For example, the voice modulation model 600 may modulate voice data with pre-stored voice data. As another example, the voice modulation model may modulate voice data according to specific conditions using an artificial intelligence-based model.

In one embodiment of the present disclosure, when inputting voice data to the voice modulation model 600, the computing device 100 may input voice data from which noise is removed. For example, when voice data consists of a user's voice and background sound (eg, ambient sound), the background sound has a spectrum different from that of the user's voice, such as a multiple of the fundamental frequency in the frequency domain, unlike the user's voice. can have The computing device 100 may improve the performance of the voice modulation model by removing voice data having a different spectrum from the user's voice as noise. As another example, the computing device 100 may remove noise from voice data based on parameters related to the user's voice, such as frequency corresponding to voice, type (voiced or unvoiced, etc.), sound velocity, and pitch.

In one embodiment, the computing device 100 recognizes that a predefined emotional word (eg, profanity, profanity, or negative keyword) appears in the voice counseling data, or a quantitative value of intonation exceeding a threshold strength in the voice counseling data When this is detected, it may be determined to modulate the voice counseling data. Emotional words in the present disclosure may refer to predefined words that contain exasperated emotions, and may have a psychological effect on agents or customers, such as, for example, profanity and/or profanity. may contain words. As described above, intonation in the present disclosure includes a state of pitch or a type of variation of uttered voice data, and may indicate a speaker's attitude other than the original meaning of uttered voice. For example, in a phrase, the melody of voice, which is indicated by the rise and fall of voice, the stress of voice (eg, the alternation of long and short sounds or the alternation of words with and without stress), the speed of voice indicating the speed and slowness of speech, the emotion or A tone that represents the color of the voice giving psychological nuance can be a component of intonation.

In one embodiment of the present disclosure, computing device 100 may quantify the aforementioned intonation according to a predefined technique, and may determine the agent's or customer's emotional level based on the quantified value of the intonation. In another embodiment of the present disclosure, computing device 100 may classify intonations into a predetermined number of classes. In this example, computing device 100 may classify the intonation into a class, such as “angry,” “excited,” “normal,” and/or “calm.” That is, through the second consultation quality evaluation process 460B, the computing device 100 may assign the speaker's accent to a predetermined class.

In one embodiment of the present disclosure, when the computing device 100 modulates the voice counseling data, for example, the modulated voice counseling data in such a manner as to convert emotion words in the voice counseling data to silence or other representations. can create For example, the computing device 100 may perform voice modulation in a manner of changing emotional words into non-emotional words. The non-emotional words may refer to words in which at least a part of emotion words including profanity and/or profanity are changed into other expressions. Non-emotion words may include any type of word for reducing the emotion level of an emotion word.

In another embodiment, when the computing device 100 modulates the voice counseling data, for example, the rhythm, speed, timbre, intensity of voice for each phrase, and/or pitch of voice for each phrase in the voice counseling data. It is possible to generate modulated voice counseling data in a manner of changing. The computing device 100 modulates the intensity and/or pitch of the voice so that a high-quality counseling attitude can be delivered to other users who will listen to the voice counseling data by mitigating the intonation of the user's voice. can be allowed

As described above, the technique according to an embodiment of the present disclosure includes a technical feature of additionally modulating the counselor's voice while evaluating the counseling quality of the counselor in an efficient manner and providing feedback to the counselor according to the evaluation result. Hereinafter, an operating algorithm of the voice modulation model 600 in a situation where a customer expresses emotion toward a counselor is described. Accordingly, in the following description, the modulation of the voice spoken by the customer to the counselor is taken as an example, but the modulation of the voice spoken by the counselor to the customer may also be included according to implementation aspects.

In a further embodiment of the present disclosure, the computing device 100 has a voice modulation model so that modulated voice counseling data is output instead of voice counseling data uttered by a customer at the counselor terminal 200B performing a voice counseling session. (600) can also be used. As a result, direct or indirect damage to counselors caused by customers' spoken voices including profanity, verbal abuse, and emotional accents can be reduced, and a fundamental solution to counselors' emotional labor can be presented.

The voice modulation model 600 may include a first module 610 and a second module 620 . The voice modulation model 600 shown in FIG. 6 is presented for illustrative purposes, and additional components may be added or some of these components may be excluded or incorporated into one or the other depending on the implementation.

In an embodiment of the present disclosure, the first module 610 may analyze voice data spoken by a user or input to the voice modulation model 600 to determine whether to modulate the voice data. The second module 620 may be connected to the output of the first module 610, and when the first module 610 determines to modulate the voice data, it may generate modulated voice data from the voice data. Additionally, the operations performed in the first module 610 may also be performed in the consultation quality evaluation model 400B in FIGS. 4 and 5 .

In one embodiment, the first module 610 may include any of the forms of natural language processing (NLP) algorithms described above for analysis on speech data.

The first module 610 may determine whether and/or the number of predefined emotional words appear in the text. The first module 610 may determine to perform modulation on the first voice data when a predefined emotion word appears or when the number of occurrences of the predefined emotion word exceeds a threshold number. Emotional words in the present disclosure may refer to words that can affect the other person's emotions or express the speaker's emotions.

In one embodiment, the first module 610 counts the number of occurrences of the sentiment word during a predetermined period of time (eg, 5 seconds, 10 seconds, 30 seconds, 1 minute, etc.), such that after the predetermined period of time It is possible to determine whether to modulate the voice data of the user who is uttered. For example, when the number of utterances of emotion words is 3 times for 10 seconds, the user's voice data uttered after 10 seconds may be modulated and transmitted to the counselor terminal. As another example, when the emotion word appears once for 5 seconds, the user's voice data may be modulated and transmitted to the counselor terminal after 5 seconds or after recognizing that the emotion word appears.

In another embodiment, the first module 610 may perform the voice modulation function in real time. For example, when the first module 610 recognizes that a predefined emotion word appears in the first voice data, the first module 610 performs modulation on the emotion word before the appeared emotion word is transmitted to the counselor terminal. can In this case, since the counselor can perform tasks without hearing emotional words from the user, the counseling efficiency of the counselor can be increased.

In one embodiment of the present disclosure, the first module 610 may include an accent analysis algorithm, and may perform accent analysis on uttered voice data based on the algorithm.

For example, the first module 610 determines the rhythm of a voice in the first voice data, the speed of a voice in the first voice data, the timbre of the first voice data, and each phrase in the first voice data. The intonation of the first voice data may be determined based on at least one of voice strength and voice pitch for each phrase in the first voice data. For example, the first module 610 determines the rhythm of a voice in the first voice data, the speed of a voice in the first voice data, the timbre of the first voice data, and each phrase in the first voice data. The method of inputting each of a plurality of factors including the intensity of voice or the height of voice for each phrase in the first voice data as input data to an artificial intelligence-based model and obtaining an output of the artificial intelligence model A class classification or quantitative value of intonation for the first voice data may be obtained.

In an embodiment, the first module 610 may determine an amount of time or utterance duration during which a quantitative value of intonation in the first speech data exceeds a threshold intensity during a particular time period. For example, if it is determined that the total duration of the intonation classified into the class of “anger” for a specific time period of 10 seconds or the total duration of the intonation in a state where the quantitative value of the intonation exceeds 70 points is 5 seconds, the first Module 610 may determine to modulate the first voice data uttered (or recognizing an utterance) after the time period of 10 seconds has elapsed.

In one embodiment, when the first module 610 determines that the quantitative value of intonation exceeds a threshold strength or that the class of intonation corresponds to a specific class, before the first voice data including the intonation is transferred to the counselor terminal Modulation of the first voice data may be determined. In this way, the first module 610 can implement real-time modulation of voice data.

In an embodiment of the present disclosure, the first module 610 may differently apply a criterion for determining whether or not to modulate based on identification information of a counselor who initiates a voice counseling session with a user terminal. For example, the counselor identification information may include the counselor's career, the counselor's gender, the counselor's age, and/or the counselor's pre-entered information (eg, counselor's condition level information, etc.). In the present disclosure, the criterion for determining whether or not to have been tampered with may include the type of emotion word, a threshold value for the number of appearances of the emotion word, and a threshold strength of the intonation. A high criterion for determining whether or not to be altered means that the number of words judged as emotional words is reduced, and/or the threshold value according to the number of occurrences of emotional words is increased, and/or the intonation This may mean that the threshold strength is increased.

For example, counselor A is a counselor with a lot of experience, and counselor B is a relatively less experienced counselor than A. In this case, agent B may be more vulnerable to the customer's emotional words than agent A. In the case of Counselor B, there is a need to reduce the possibility of Counselor B's emotional labor by lowering the criteria for determining falsification compared to Counselor A. Accordingly, according to an embodiment of the present disclosure, the computing device 100 differently determines a criterion for determining whether or not to be tampered with based on the counselor's identification information, thereby enabling customized emotion management for each counselor. In an additional embodiment of the present disclosure, the computing device 100 inputs counseling condition information from a counselor terminal conducting a voice counseling session with the user terminal through the counselor terminal, eg, daily, weekly, or at a specific point in time. can receive According to the counseling condition information, the computing device 100 differently applies the modulation criterion of the first module 410, thereby reducing the possibility of emotional labor according to the counselor's current condition.

In one embodiment of the present disclosure, the second module 620, in response to determining that the predefined sentiment word occurring in the first speech data is to be modulated, by muting the predefined sentiment word; or Modulation may be performed on the first voice data in a manner of changing to a predefined non-emotion word.

In one embodiment of the present disclosure, the second module 620, in response to determining a modulation for an intonation that exceeds a threshold strength in the first voice data, determines a modulation for a portion containing the intonation in the first voice data. Modulation of at least one of voice intensity and voice pitch may be performed.

In one embodiment of the present disclosure, the second module 620 modulates the rhythm of voice in the first voice data, modulates the velocity of voice in the first voice data, modulates the rhythm of voice in the first voice data. Modulation of the first voice data by at least one of modulation of timbre, modulation of the intensity of each phrase in the first voice data, or modulation of the pitch of each phrase in the first voice data can be performed.

In a further embodiment of the present disclosure, the second module 620 may perform modulation on the first voice data in a different manner based on agent identification information initiating a voice counseling session with the user terminal. . For example, it is assumed that counselor C has greater stress on intonation than stress on the appearance of emotion words. In this situation, the second module 420 may perform modulation focusing on modulation of intonation rather than modulation of emotional words for agent C. For example, the variation of the modulation scheme for each counselor may be performed based on information pre-input from the counselor.

For example, the second module 620 prioritizes the modulation of the intonation over the modulation of the emotion words for a first counselor terminal, and prioritizes the modulation of the pitch of the voice for a second counselor terminal. Priority is given to the modulation of the intensity of the voice, and for the third counselor terminal, modulation of the first voice data can be performed with a voice of a gender matching the gender of the user of the third counselor terminal. there is.

In a further embodiment of the present disclosure, the voice modulation model 600 includes a generation network that is trained to receive voice data and output modulated voice data, and receive the modulated voice data that is an output of the generation network to perform the It may be a pre-trained artificial intelligence model based on a discriminator network that determines whether input data is modulated data.

7 depicts a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

Although the present disclosure has been described above as being generally embodied by a computing device, those skilled in the art will understand that the present disclosure may be combined with computer-executable instructions and/or other program modules that may be executed on one or more computers and/or may be implemented in hardware and software. It will be appreciated that it can be implemented as a combination.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those skilled in the art will understand that the methods of the present disclosure can be applied to single-processor or multiprocessor computer systems, minicomputers, mainframe computers as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like ( It will be appreciated that each of these may be implemented with other computer system configurations, including those that may be operative in connection with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer readable media. Computer readable media can be any medium that can be accessed by a computer, including volatile and nonvolatile media, transitory and non-transitory media, removable and non-transitory media. Includes removable media. By way of example, and not limitation, computer readable media may include computer readable storage media and computer readable transmission media. Computer readable storage media are volatile and nonvolatile media, transitory and non-transitory, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. includes media Computer readable storage media may include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device. device, or any other medium that can be accessed by a computer and used to store desired information.

A computer readable transmission medium typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism. Including all information delivery media. The term modulated data signal means a signal that has one or more of its characteristics set or changed so as to encode information within the signal. By way of example, and not limitation, computer readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above are also intended to be included within the scope of computer readable transmission media.

An exemplary environment 1100 implementing various aspects of the present disclosure is shown including a computer 1102, which includes a processing unit 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 couples system components, including but not limited to system memory 1106 , to processing unit 1104 . Processing unit 1104 may be any of a variety of commercially available processors. Dual processor and other multiprocessor architectures may also be used as the processing unit 1104.

System bus 1108 may be any of several types of bus structures that may additionally be interconnected to a memory bus, a peripheral bus, and a local bus using any of a variety of commercial bus architectures. System memory 1106 includes read only memory (ROM) 1110 and random access memory (RAM) 1112 . A basic input/output system (BIOS) is stored in non-volatile memory 1110, such as ROM, EPROM, or EEPROM, and is a basic set of information that helps transfer information between components within computer 1102, such as during startup. contains routines. RAM 1112 may also include high-speed RAM, such as static RAM, for caching data.

The computer 1102 may also include an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA) - the internal hard disk drive 1114 may also be configured for external use within a suitable chassis (not shown). Yes—a magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to a removable diskette 1118), and an optical disk drive 1120 (e.g., a CD-ROM) for reading disc 1122 or reading from or writing to other high capacity optical media such as DVDs). The hard disk drive 1114, magnetic disk drive 1116, and optical disk drive 1120 are connected to the system bus 1108 by a hard disk drive interface 1124, magnetic disk drive interface 1126, and optical drive interface 1128, respectively. ) can be connected to The interface 1124 for external drive implementation includes at least one or both of USB (Universal Serial Bus) and IEEE 1394 interface technologies.

These drives and their associated computer readable media provide non-volatile storage of data, data structures, computer executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the description of computer readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art can use zip drives, magnetic cassettes, flash memory cards, cartridges, etc. It will be appreciated that other tangible media readable by the computer, such as the like, may also be used in the exemplary operating environment and any such media may include computer executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored on the drive and RAM 1112, including an operating system 1130, one or more application programs 1132, other program modules 1134, and program data 1136. All or portions of the operating system, applications, modules and/or data may also be cached in RAM 1112. It will be appreciated that the present disclosure may be implemented in a variety of commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1102 through one or more wired/wireless input devices, such as a keyboard 1138 and a pointing device such as a mouse 1140. Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, and the like. Although these and other input devices are often connected to the processing unit 1104 through an input device interface 1142 that is connected to the system bus 1108, a parallel port, IEEE 1394 serial port, game port, USB port, IR interface, may be connected by other interfaces such as the like.

A monitor 1144 or other type of display device is also connected to the system bus 1108 through an interface such as a video adapter 1146. In addition to the monitor 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, and the like.

Computer 1102 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 1148 via wired and/or wireless communications. Remote computer(s) 1148 may be a workstation, computing device computer, router, personal computer, handheld computer, microprocessor-based entertainment device, peer device, or other common network node, and generally includes It includes many or all of the components described for, but for simplicity, only memory storage device 1150 is shown. The logical connections shown include wired/wireless connections to a local area network (LAN) 1152 and/or a larger network, such as a wide area network (WAN) 1154 . Such LAN and WAN networking environments are common in offices and corporations and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to worldwide computer networks, such as the Internet.

When used in a LAN networking environment, computer 1102 connects to local network 1152 through wired and/or wireless communication network interfaces or adapters 1156. Adapter 1156 may facilitate wired or wireless communications to LAN 1152, which also includes a wireless access point installed therein to communicate with wireless adapter 1156. When used in a WAN networking environment, computer 1102 may include a modem 1158, be connected to a communicating computing device on WAN 1154, or establish communications over WAN 1154, such as over the Internet. have other means. A modem 1158, which may be internal or external and a wired or wireless device, is connected to the system bus 1108 through a serial port interface 1142. In a networked environment, program modules described for computer 1102, or portions thereof, may be stored on remote memory/storage device 1150. It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between computers may be used.

Computer 1102 is any wireless device or entity that is deployed and operating in wireless communication, eg, printers, scanners, desktop and/or portable computers, portable data assistants (PDAs), communication satellites, wireless detectable tags associated with It operates to communicate with arbitrary equipment or places and telephones. This includes at least Wi-Fi and Bluetooth wireless technologies. Thus, the communication may be a predefined structure as in conventional networks or simply an ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) makes it possible to connect to the Internet without wires. Wi-Fi is a wireless technology, such as a cell phone, that allows such devices, eg, computers, to transmit and receive data both indoors and outdoors, i.e. anywhere within coverage of a base station. Wi-Fi networks use a radio technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in the unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band) .

Those skilled in the art will understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols and chips that may be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields s or particles, or any combination thereof.

Those skilled in the art will understand that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein are electronic hardware, (for convenience) , may be implemented by various forms of program or design code (referred to herein as software) or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and the design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory devices (eg, EEPROM, cards, sticks, key drives, etc.), but are not limited thereto. Additionally, various storage media presented herein include one or more devices and/or other machine-readable media for storing information.

It is to be understood that the specific order or hierarchy of steps in the processes presented is an example of exemplary approaches. Based upon design priorities, it is to be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure. The accompanying method claims present elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

The description of the presented embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of this disclosure, and the general principles defined herein may be applied to other embodiments without departing from the scope of this disclosure. Thus, the present disclosure is not to be limited to the embodiments presented herein, but is to be interpreted in the widest scope consistent with the principles and novel features presented herein.

Claims (20)

A method for evaluating voice counseling performed by a computing device comprising one or more processors, comprising:
obtaining voice counseling data generated in the course of performing a voice counseling session with the user terminal initiated in response to a call requested by the user terminal;
A first consultation quality evaluation process for generating text consultation data by performing text conversion on the voice consultation data and evaluating the quality of consultation contents through consultation words uttered by a counselor based on the text consultation data performing steps; and
performing a second consultation quality evaluation process for evaluating the quality of the counseling attitude through at least one of counseling pronunciation, counseling intonation, and counseling voice of the counselor based on the voice counseling data;
including,
method.
According to claim 1,
The first consultation quality evaluation process,
Tokenize the text counseling data into a plurality of tokens based on a natural language processing (NLP) algorithm, determine whether a specific keyword appears in the text counseling data based on the plurality of tokens, and tokenize the text counseling data. uses a space or comma in the text consultation data as a stopword,
method.
According to claim 2,
determining whether a first predetermined keyword appears during a first time period in the text counseling data, wherein the first keyword and the first time period are mapped to each other and stored;
method.
According to claim 2,
The first consultation quality evaluation process,
Determining whether an essential speech keyword has appeared in the text counseling data and determining whether a predefined negative keyword has appeared in the text counseling data;
method.
According to claim 3 or 4,
The second consultation quality evaluation process,
Performed when the first predetermined keyword does not appear during the first time period, when a required speech keyword does not appear in the text counseling data, or when a predefined negative keyword appears in the text counseling data characterized in that,
method.
According to claim 3 or 4,
The first consultation quality evaluation process,
Characterized in that it is performed when the quality of the counselor's counseling attitude according to the second counseling quality evaluation process is less than a predetermined quality threshold,
method.
According to claim 1,
The second consultation quality evaluation process,
determining whether a quantitative value of counseling accent or quantitative value of counseling voice of the counselor during a first period of time exceeds a threshold intensity and determining a duration of the counseling accent or voice volume exceeding the threshold intensity, so that the counselor classifying the counseling attitude into one of one or more pre-determined classes,
method.
According to claim 7,
modulating voice counseling data uttered by the counselor after the first time period, when a classification result of the counseling attitude of the counselor is classified into a class related to deterioration in counseling quality;
Including more,
method.
According to claim 7,
When the classification result of the counseling attitude of the counselor is classified into a class related to deterioration in counseling quality, before the voice counseling data spoken by the counselor terminal initiating the voice counseling session with the user terminal is transmitted to the user terminal. performing modulation on the voice counseling data;
Including more,
method.
According to claim 8 or 9,
The step of performing modulation on the voice consultation data,
Modulation of voice rhythm in the voice counseling data, modulation of voice speed in the voice counseling data, modulation of timbre in the voice counseling data, and intensity of voice for each phrase in the voice counseling data Performing at least one of modulation or modulation of the pitch of the voice for each phrase in the voice consultation data,
method.
According to claim 8 or 9,
Based on the identification information of the user of the user terminal, the modulation method for the voice consultation data is variably determined.
method.
According to claim 1,
The second consultation quality evaluation process,
Evaluating the quality of the counseling pronunciation of the counselor based on a recognition score representing the recognition rate of the generated text counseling data, and evaluating the quality of the counseling attitude based on the quality of the pronunciation.
method.
According to claim 1,
The second consultation quality evaluation process,
Evaluating the quality of the counseling pronunciation of the counselor based on a comparison of the degree of agreement between the generated text counseling data and predetermined term data, and evaluating the quality of the counseling attitude based on the quality of the pronunciation;
method.
According to claim 1,
The second consultation quality evaluation process,
Evaluating the quality of the counseling attitude by obtaining at least one of a sigh, a predetermined negative onomatopoeia, or a predetermined negative intonation from the voice counseling data,
method.
According to claim 1,
The second consultation quality evaluation process,
distinguishing first voice consultation data spoken by an agent and second voice consultation data spoken by a user from the voice consultation data based on the voice signal amplitude and voice signal frequency of the voice consultation data;
identify an overlapping interval between the first audio consultation data and the second audio consultation data, based on a predetermined time frame of the audio consultation data; and
Based on the overlapping interval, determining whether the counselor intervened in the user's speech process of the user terminal;
method.
According to claim 1,
At least one of the first consultation quality evaluation process and the second consultation quality evaluation process,
determine a priority for either the quality of the counseling content or the quality of the counseling attitude based on the past evaluation history of the counselor; and
Based on the priority, performing a consultation quality evaluation process,
method.
According to claim 1,
At least one of the first consultation quality evaluation process and the second consultation quality evaluation process,
Based on the counselor's past evaluation history, priorities for a plurality of evaluation items in the first consultation quality evaluation process or the second consultation quality evaluation process are determined; and
Based on the priority, performing a consultation quality evaluation process,
method.
According to claim 1,
At least one of the first consultation quality evaluation process and the second consultation quality evaluation process,
Extracting the contents of the user's speech of the user terminal from the voice consultation data;
Determining whether a predetermined positive keyword appears from the user's utterance, and
Based on the predetermined positive keyword, performing a consultation quality evaluation process,
method.
According to claim 1,
generating feedback information to be provided to the counselor terminal based on at least one result of the first consultation quality evaluation process and the second consultation quality evaluation process;
Including more,
The feedback information includes a first evaluation related to counseling contents based on a result of the first consultation quality evaluation process and a second evaluation related to a consultation attitude based on a result of the second consultation quality evaluation process,
The first evaluation is based on information on words that should have been uttered during a first time period, which is part of the course of the voice counseling session, information on words that should have been uttered during the course of the voice counseling session, and the voice counseling session. contains information about words that should not have been uttered in the process of being performed, and
The second evaluation includes information on pronunciation of low quality, information on counseling accent whose quality is evaluated low, and information on counseling voice whose quality is evaluated low in the course of the voice counseling session.
method.
A computer program stored on a computer readable storage medium,
The computer program includes instructions for causing a processor of a computing device to perform a method for evaluating voice counseling, the method comprising:
obtaining voice counseling data generated in the course of performing a voice counseling session with the user terminal initiated in response to a call requested by the user terminal;
A first consultation quality evaluation process for generating text consultation data by performing text conversion on the voice consultation data and evaluating the quality of consultation contents through consultation words uttered by a counselor based on the text consultation data performing steps; and
performing a second consultation quality evaluation process for evaluating the quality of the counseling attitude through at least one of counseling pronunciation, counseling intonation, and counseling voice of the counselor based on the voice counseling data;
including,
A computer program stored on a computer readable storage medium.

Images