CN111382234A - Reply providing method and device based on customer service - Google Patents
Reply providing method and device based on customer service Download PDFInfo
- Publication number
- CN111382234A CN111382234A CN201811513766.4A CN201811513766A CN111382234A CN 111382234 A CN111382234 A CN 111382234A CN 201811513766 A CN201811513766 A CN 201811513766A CN 111382234 A CN111382234 A CN 111382234A
- Authority
- CN
- China
- Prior art keywords
- question
- identified
- customer service
- feature vector
- semantic information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 230000004044 response Effects 0.000 claims abstract description 73
- 238000003909 pattern recognition Methods 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 46
- 230000008569 process Effects 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 23
- 238000010586 diagram Methods 0.000 description 12
- 230000011218 segmentation Effects 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 238000012423 maintenance Methods 0.000 description 7
- 238000007726 management method Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 5
- 210000004556 brain Anatomy 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000006854 communication Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012827 research and development Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
Abstract
The invention discloses a reply providing method and a reply providing device based on customer service, wherein the method comprises the following steps: receiving a first question input by a user; according to a preset pattern recognition rule, recognizing first entity words, first attribute words and first semantic information included in the first question, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing a response of the first question; therefore, the accuracy rate of response provided by the intelligent customer service robot is improved, and the user experience is further improved.
Description
Technical Field
The invention relates to the field of intelligent customer service, in particular to a method and a device for providing a reply based on customer service.
Background
The enterprise customer service is used as an interactive entrance of enterprise users and enterprise services and bears service businesses such as enterprise-related business consultation, business handling, question answering and complaint suggestion. With the development of intelligent customer service robots, more and more enterprises select the intelligent customer service robots to replace manual customer service to complete certain customer service business.
Although the intelligent customer service robots can complete service businesses timely, the completion process has many limitations. For example, the intelligent customer service robots can only recognize keywords in the service consultation questions with the plain text data, and then reply to the service consultation questions according to the keywords and the set question-answer matching rules. Because the semantics generated by the same keyword under different scenes are different, but the existing intelligent customer service robot cannot identify the different semantics carried in the same keyword, the response accuracy of the business consultation problem consulted by the user is not high, which causes the service of the user to the intelligent customer service robot to be not full, causes the quality of the customer service of an enterprise to be reduced, further greatly reduces the purchase willingness of the user to related products of the enterprise, even abandons the purchase of the products, and brings more negative effects, such as the reduction of the public praise of the enterprise.
Disclosure of Invention
The embodiment of the invention provides a response providing method and device based on customer service, which are used for solving the problem that in the prior art, the accuracy of an intelligent customer service robot for responding to a service problem of user consultation is low, so that the user experience effect is poor.
The embodiment of the invention provides a reply providing method based on customer service, which comprises the following steps:
receiving a first question input by a user;
according to a preset pattern recognition rule, recognizing first entity words, first attribute words and first semantic information included in the first question, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun;
forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question;
inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
Further, the training process of the recognition model is as follows:
aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a reply, and each question about the reply identifies a second entity word, a second attribute word and second semantic information included in the reply and each question in the first sample subset according to a preset pattern identification rule;
forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set;
and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding answer.
Further, the sample subset includes structured data represented by basic data types, unstructured data represented by pictures and voices, and semi-structured data obtained after parsing by using a corresponding processing tool.
Further, the method further comprises:
saving the first question and the response to the first question.
Further, the method further comprises:
determining each second sample subset according to each stored first question and the response aiming at the first question according to a set time interval, wherein each second sample subset comprises a response and each first question related to the response;
for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule;
forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set;
and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
The embodiment of the invention provides a reply providing device based on customer service, which comprises:
the receiving module is used for receiving a first question input by a user;
the customer service module is used for identifying first entity words, first attribute words and first semantic information included in the first question according to a preset pattern identification rule, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question;
and the determining module is used for inputting the first feature vector into a pre-trained recognition model, determining and providing the response of the first question.
Further, the apparatus further comprises:
the first training module is used for aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a response, and identifying the responses in the first sample subset and each question comprises a second entity word, a second attribute word and second semantic information according to a preset pattern identification rule; forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set; and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding answer.
Further, the apparatus further comprises:
and the storage module is used for storing the first question and the answer aiming at the first question.
Further, the apparatus further comprises:
the second training module is used for determining each second sample subset according to the stored first question and the answer aiming at the first question at a set time interval, wherein each second sample subset comprises one answer and each first question related to the answer; for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule; forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set; and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
Further, the customer service module includes: a routing unit and at least two customer service units,
the routing unit is used for determining the customer service unit with the minimum load according to the stored load information corresponding to each customer service unit and sending the first problem to the customer service unit with the minimum load;
each customer service unit is used for identifying a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, wherein the first entity word is a first noun included in the first question, and the first attribute word is an adjective word for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
Further, the customer service module includes:
a routing unit and at least two customer service units,
the routing unit is used for identifying a target problem type corresponding to the first problem according to the first problem, determining a target customer service unit corresponding to the target problem type according to the stored corresponding relation between the problem type and the customer service unit, and sending the first problem to the target customer service unit;
the target customer service unit is configured to identify a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, where the first entity word is a first noun included in the first question, and the first attribute word is an adjective word modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
The embodiment of the invention provides a reply providing method and a reply providing device based on customer service, wherein the method comprises the steps of receiving a first question input by a user; according to a preset pattern recognition rule, recognizing first entity words, first attribute words and first semantic information included in the first question, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
According to the embodiment of the invention, a first feature vector is formed according to a first entity word set, a first attribute word set, first semantic information and a first question in a first question input by a user; the first feature vector is input into a pre-trained recognition model, and the response of the first question is determined and provided, so that the accuracy of providing the response by the intelligent customer service robot is improved, and the user experience is further improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a response providing method based on customer service according to embodiment 1 of the present invention;
fig. 2 is a schematic structural diagram of a customer service-based response providing apparatus according to embodiment 4 of the present invention;
fig. 3 is a schematic process diagram of training a recognition model in an intelligent customer service robot according to embodiment 4 of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the attached drawings, and it should be understood that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1:
fig. 1 is a flowchart of a response providing method based on customer service according to an embodiment of the present invention, where the method includes:
s101: a first question input by a user is received.
The user can input a first question to be consulted in a consultation entrance on a corresponding business consultation webpage.
S102: according to a preset pattern recognition rule, recognizing first entity words, first attribute words and first semantic information included in the first question, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun.
After receiving the first question, in order to determine a response corresponding to the first question, in the embodiment of the present invention, a word segmentation process is performed on the first question, part of speech is identified based on each obtained word segmentation, each noun included in the first question is determined according to the part of speech of each identified word segmentation, the noun is used as an entity word, and for each noun, if an adjective modifying the noun exists, the adjective is determined as an attribute word. Since each participle of the first problem has been obtained by the participle processing, semantic information of the participle can be identified on a per-participle basis.
Specifically, the first entity word may be a noun related to the product information, such as a name of a purchased garment, and the first attribute word may be an adjective that modifies the first entity word, such as a word that indicates a color, a size, and the like of the garment, and the first semantic information may be obtained by processing through a corresponding semantic processing tool.
The specific processes of word segmentation processing, part of speech recognition and semantic analysis are prior art and are not described herein again.
S103: and forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question.
After each noun, i.e., each entity word, included in the first question is obtained through analysis, a set formed by all entity words is called a first entity word set, a set formed by all attribute words is called a first attribute word set, and a set formed by all semantic information is called a first semantic information set.
And forming a first feature vector according to the obtained first entity word set, the first attribute word set, the first semantic information set and the first question.
Specifically, the first feature vector including the weight value and each set may also be determined according to a preset weight value corresponding to each set.
S104: inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
The embodiment of the invention adopts a pre-trained recognition model to process the first feature vector, thereby completing the intelligent customer service question-answering process and the answer reasoning, further determining the answer of the first question and providing the answer to the user.
According to the embodiment of the invention, a first feature vector is formed according to a first entity word set, a first attribute word set, first semantic information and a first question in a first question input by a user; the first feature vector is input into a pre-trained recognition model, and the response of the first question is determined and provided, so that the accuracy of providing the response by the intelligent customer service robot is improved, and the user experience is further improved.
Example 2:
in order to improve the accuracy of the intelligent customer service robot in responding to the service questions consulted by the user, based on the above embodiment, the training process of the recognition model is as follows:
aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a reply, and each question about the reply identifies a second entity word, a second attribute word and second semantic information included in the reply and each question in the first sample subset according to a preset pattern identification rule;
forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set;
and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding answer.
The sample subset comprises structured data represented by basic data types, unstructured data not represented by the basic data types and semi-structured data obtained after analysis by adopting a corresponding processing tool, wherein the unstructured data not represented by the basic data types specifically comprises data represented by pictures and voice.
In order to improve the accuracy of the intelligent customer service robot in responding to the service questions consulted by the user, a training sample set is stored in the intelligent customer service robot, and the training sample set is formed by question and answer data collected by research personnel and normalized and integrated. The training sample set comprises a plurality of sample subsets, each sample subset comprises a response collected in advance and each question related to the response, and therefore the intelligent customer service robot can train the recognition model based on the responses in the sample subsets and each question.
The sample subset comprises structured data represented by basic data types, unstructured data represented by pictures and voices, and semi-structured data obtained after analysis by adopting a corresponding processing tool. Specifically, the structured data may be character type data, integer type data, and floating point type data, and these data may be used to store product information, customer information, and enterprise information; the unstructured data mainly takes pictures, voice and text data in the question answering data as main data; the semi-structured data is generally web page data, and the web page data needs to be parsed by a browser to obtain corresponding data, where the web page data may be html data or XML data, for example.
In order to improve the training efficiency of the recognition model in the intelligent customer service robot, before training the recognition module, research and development personnel need to effectively organize, extract and process the collected question and answer data, for example, classify the collected question and answer data into structured data, unstructured data and semi-structured data, and normalize and integrate the data, that is, integrate heterogeneous data to obtain integrated question and answer data, that is, the integrated data includes the three types of data, so as to construct a uniform data representation form and apply the uniform data representation form to the recognition model.
In order to train the recognition model, in the embodiment of the present invention, word segmentation processing is performed on the responses in each first sample subset and each question about the responses, part of speech recognition is performed based on each obtained word segmentation, each noun included in the responses in each first sample subset and each question is determined according to the part of speech of each recognized word segmentation, the noun is used as an entity word, and for each noun, if an adjective modifying the noun exists, the adjective is determined as an attribute word.
Specifically, according to a preset pattern recognition rule, second entity words, second attribute words and second semantic information included in the answers and each question in the first sample subset are recognized and analyzed; the second entity word may be a noun related to the product information, the second attribute word may be an adjective for modifying the second entity word, and the second semantic information may be obtained by processing through a corresponding semantic processing tool.
The specific processes of word segmentation processing, part of speech recognition and semantic analysis are prior art and are not described herein again.
After analyzing each noun, i.e. each entity word, included in the response and each question in the first sample subset, a set formed by all entity words is called a second entity word set, a set formed by all attribute words is called a second attribute word set, and a set formed by all semantic information is called a second semantic information set.
And forming a second feature vector according to the obtained second entity word set, the second attribute word set, the second semantic information and the set of each question.
Specifically, the second feature vector including the weight value and each set may also be determined according to a preset weight value corresponding to each set.
Specifically, when the recognition model is trained, for each first sample subset, a second entity word set, a second attribute word set, second semantic information and a question set are obtained, each set is respectively combined with a corresponding preset weight value, and sequentially arranged according to a sequence to form a second feature vector, each second feature vector and a response corresponding to each first sample subset are input into the recognition model, so that the recognition model is trained, and a specific formula is as follows:
Q＝{E:We；K:Wk；S:Ws；R:Wr}
where Q is the response in the first subset of samples, E is the second set of entity words, WeSet of weight values corresponding to the second set of entity wordsK is the second attribute word set, WkIs a weight value set corresponding to the second attribute word set, S is a second semantic information set, WsIs a weight value set corresponding to the second semantic information set, R is a problem set, WrAnd the weight value set corresponding to the question set.
According to the embodiment of the invention, aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a response, and each question of the response, according to a preset pattern recognition rule, recognizing the responses in the first sample subset and a second entity word, a second attribute word and second semantic information included in each question; forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set; and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding response, so that the accuracy of the intelligent customer service robot in responding to the service questions consulted by the user can be improved.
The embodiment of the invention also provides a customer service system framework provided based on the response of customer service, the system framework is composed of different technical components, and the technical components are deployed in a distributed cloud deployment mode in a container engine Docker or a container cluster management system Kubernets, so that the customer service question-answering method is realized. The communication between the technical components is completed through the interfaces on the technical components, and follows the communication rule of Docker or Kubernetes, and the process is the prior art and is not described herein again.
The roles of the key technical components are introduced from two major aspects of infrastructure and business application modules:
technical components that can efficiently perform data storage, message invocation, etc. operations can be understood as the infrastructure of the customer service system architecture. For example, the key-value pair storage database component Redis is used for storing cache information such as question-answer cache Session-cache data and mark information for identifying model training such as a training lock in each customer service question-answer process; the text search engine component ES is used for voice information retrieval in the question and answer process and dialogue record retrieval in the question and answer process of customer service each time; the distributed application program coordination service component ZooKeeper is used for managing a message queue of a dialogue record in the question and answer process; the message queue model component NSQ is used for assisting the ZooKeeper in managing the message queue; and the storage system component Minio is used for storing rich text data and is mainly used for storing the recognition model.
The technical components for processing different service businesses can be understood as business application modules in a customer service system framework, and are mainly used for responding to business questions consulted by users correspondingly. For example, the browser webpage updates the Web-adapter component, which is used for adapting to the message format of the webpage Web end, converting the question consulted by the user at the Web end into a standard user message format, and converting the standard reply into reply data at the Web end, wherein the conversion of the message format of the question consulted by the user and the message format of the reply is completed by a corresponding message processing intermediate technology component; the Heimdall component of the hypertext transfer protocol client is an integral gateway of messages generated in the process of consulting business problems by a user, provides synchronous and asynchronous conversation modes, can receive user messages in a standard format and messages for training a recognition model through an application program interface REST API (representational State transfer) conforming to the constraint principle of a system architecture and return robot messages in the standard format, and can also send the robot messages in the standard format through a callback registered application program interface API to realize asynchronous conversation.
Specifically, in the embodiment of the present invention, the response is provided by the customer service robot. In order to provide responses for a plurality of users at the same time, the embodiment of the invention can form an intelligent customer service robot set EinBot based on a plurality of parallel intelligent customer service robots Brain, wherein each intelligent customer service robot corresponds to an instance set. Therefore, the method is convenient for providing answers for different questions consulted by the user at the same time, and the specific process is as follows:
when a plurality of users input consultation problems through respective corresponding browsers, the browsers send the problems to a server, a specific server comprises a routing unit, after receiving the problems, the routing unit distributes the problems to intelligent service robots in an intelligent service robot set EinBot associated with the intelligent service robots according to load information of each intelligent service robot Brain stored by the routing unit, namely information such as the number of problems currently solved by each intelligent service robot, and the like, through a load balancing technology, and each intelligent service robot receiving the distributed problems determines a response corresponding to the problems according to the method in embodiment 1 and provides the response to the users consulting the problems.
For example, if a user inputs a first question on a browser, the browser sends the first question to a server, and after receiving the first question, a routing unit in the server may determine, according to load information of each intelligent service robot Brain stored by the routing unit, an intelligent service robot with the smallest load, and send the first question to the intelligent service robot, where the intelligent service robot determines a response corresponding to the first question according to the method described in embodiment 1, and provides the response to the user who consults the question; the routing unit may further perform intent recognition according to the first question, so as to recognize a target question type corresponding to the first question, determine a target intelligent service robot corresponding to the target question type according to a correspondence between the question type stored in the routing unit and the intelligent service robot, and send the first question to the target intelligent service robot, where the target intelligent service robot determines a response corresponding to the first question according to the method described in embodiment 1, and provides the response to the user who consults the question. The routing unit determines a target problem type, and then sends the first problem to the target intelligent customer service robot with the smallest load according to the load of each determined target intelligent customer service robot.
The customer service robot EinBot is used for managing message queues according to the content of the user consultation service for consumption, providing the intelligent customer service robot Brains for providing corresponding service consultation through a corresponding route according to the identity information of the user, and simultaneously providing the reply message of the intelligent customer service robot to the user.
In addition, 1 intelligent customer service robot Brain is composed of 1 intention recognition module RouteBot and a minimum processing unit Botlet of a plurality of user messages, is responsible for receiving user consultation service problems and generating corresponding robot messages, and can generate training results according to sample training sets to continuously optimize the service capability of the user, wherein one Botlet is only responsible for processing one type of information (such as question answering and cold turning) and returns processing results, namely after the RouteBot recognizes the type of the user consultation service problems, the problems are divided into corresponding Botlets through routes to be processed, and in order to deal with the question answering process under different scenes, the processing sequence among the Bratlets can be adjusted according to the described file management information; the computing component actual can realize the functions of statistics and query of data information in the question and answer process.
In addition, not only the intelligent customer service robot can provide the user with the response of the business question consulted by the user, but also some technical component sets in the customer service system framework can provide the user with the response of the business question consulted by the user, that is, the customer service robot comprises the technical component sets, and the specific process is as follows: different types of users can input consultation questions through corresponding access interfaces on browsers, the browsers send the questions to a server, after receiving the questions, a routing unit in the server can perform message queue management according to load information of each technical component set related to service business consultation and identity information of the users, which are stored by the routing unit in the server, aiming at the content of each consultation business question, distribute the questions to the corresponding technical component set related to service business consultation through a load balancing technology, each technical component set determines responses of the questions received by the technical component set according to the method in the embodiment 1 and provides the responses to the questions for users consulting the questions, specifically, the communication process among the technical components in the technical component set follows the communication rules of Docker or Kubernetes, and details are omitted here.
In addition, in order to improve the maintainability of the intelligent customer service robot and reduce the operation and maintenance difficulty and the research and development cost, some technical components related to later-stage operation and maintenance and development can be added to the customer service system architecture, for example, a relational database component Mysql is used for storing corpus data, account data, information of each intelligent customer service robot, data statistical information and part of operation and maintenance data in each intelligent customer service robot, wherein the part of operation and maintenance data mainly refers to real-time event logs, data in a gathering platform Sentry and data in an open-source chart visualization system Grafana; the program anomaly collection system component Sentry is used for monitoring real-time anomaly information, for example, using a pre-stored mailbox number of an operation and maintenance manager to perform anomaly notification; the system monitoring program component Monitor is used for effectively ensuring the high availability of the system; the operation and maintenance tool set component Satools can be used for maintaining certain system faults, so that the intelligent customer service robot can be more suitable for enterprises lacking operation and maintenance personnel for the intelligent customer service robot; the system comprises a console Economist/console user interface Economist-UI component, a service management component and a service management component, wherein the console Economist/console user interface Economist-UI component is used for controlling each intelligent customer service robot in an intelligent customer service robot set, and managing information, data, statistical analysis and management and the like in the intelligent customer service robot set; and the administrator interface Admin-UI component is used for providing super administrator services, managing user rights, numerous user information and the like.
The embodiment of the invention provides a customer service system architecture based on a micro-service architecture, which realizes message centralized management, high availability and on-demand asking, is applied to a customer service system in an intelligent customer service robot, realizes automatic learning and knowledge processing of the intelligent customer service robot, effectively ensures the knowledge learning efficiency and knowledge purity of the intelligent customer service robot, and provides a foundation for accurate customer service question and answer.
The customer service system architecture provided by the embodiment of the invention can be applied to intelligent customer service robots in different industries and can also be applied to system designs of other available micro-service architectures, and the customer service system architecture is a system architecture with high availability and high expansibility.
Example 3:
in order to further improve the accuracy of the intelligent customer service robot in responding to the business questions consulted by the user, the first question and the response to the first question are saved on the basis of the embodiment.
Determining each second sample subset according to each stored first question and the response aiming at the first question according to a set time interval, wherein each second sample subset comprises a response and each first question related to the response;
for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule;
forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set;
and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
In order to further improve the accuracy of the service questions to be answered by the user, the intelligent customer service robot needs to store the questions to be consulted in the service consultation and the corresponding answers to the questions every time the intelligent customer service robot answers the service questions to be consulted by the user, and regularly updates the question-answer database so as to facilitate the subsequent new training of the recognition model and further improve the accuracy of the subsequent service questions to be consulted by the user. Specifically, based on the above embodiment, the smart customer service robot saves the first question and the response to the first question.
In order to persistently and stably utilize the recognition model to accurately respond to the service questions consulted by the user, new training needs to be performed on the recognition model periodically, and the recognition model is continuously improved, taking the first question and the corresponding response in the above embodiment as an example, the specific process is as follows: according to a set time interval, the data of each first question and the response to the first question, which are stored by the research and development personnel, are processed correspondingly, such as normalized integration, and then each second sample subset is determined according to the processed data, each second sample subset comprises a response and each first question related to the response, wherein the time interval can be set according to actual conditions, for example, when the customer service visit volume is large, the set time interval can be slightly longer, and when the customer service visit volume is small, the set time interval can be shorter.
After each second sample subset is determined, the answers in the second sample subsets and the third entity words, the third attribute words and the third semantic information included in each first question may be identified according to a preset pattern identification rule for each second sample subset, and the identification process is similar to the identification process of the first entity words, the first attribute words and the first semantic information in the above embodiment, and is not described herein again.
And similarly, forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first problem set, wherein a specific process for forming the third feature vector is similar to a process for forming the first feature vector, and is not repeated herein. Similarly, the process of training the recognition model according to the third feature vector and the answer corresponding to each second sample subset is similar to the process of training the recognition module according to the above embodiment, and is not described herein again.
In the embodiment of the invention, the recognition model is trained according to the stored first question and the response aiming at the first question and the set time interval, so that the recognition model is perfected regularly, and the accuracy of the intelligent customer service robot for responding to the service question consulted by the user can be further improved.
Example 4:
on the basis of the above embodiments, fig. 2 is a schematic structural diagram of a response providing device based on customer service according to an embodiment of the present invention, where the device includes:
a receiving module 201, configured to receive a first question input by a user;
the customer service module 202 is configured to identify a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, where the first entity word is a first noun included in the first question, and the first attribute word is an adjective word modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
Further, the apparatus further comprises:
the first training module is used for aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a response, and identifying the responses in the first sample subset and each question comprises a second entity word, a second attribute word and second semantic information according to a preset pattern identification rule; forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set; and training the recognition model according to the second feature vector and the answer corresponding to each first sample subset.
Further, the apparatus further comprises:
and the storage module is used for storing the first question and the answer aiming at the first question.
Further, the apparatus further comprises:
the second training module is used for determining each second sample subset according to the stored first question and the answer aiming at the first question at a set time interval, wherein each second sample subset comprises one answer and each first question related to the answer; for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule; forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set; and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
Further, the customer service module includes: a routing unit and at least two customer service units,
the routing unit is used for determining the customer service unit with the minimum load according to the stored load information corresponding to each customer service unit and sending the first problem to the customer service unit with the minimum load;
each customer service unit is used for identifying a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, wherein the first entity word is a first noun included in the first question, and the first attribute word is an adjective word for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
Further, the customer service module includes: a routing unit and at least two customer service units,
the routing unit is used for identifying a target problem type corresponding to the first problem according to the first problem, determining a target customer service unit corresponding to the target problem type according to the stored corresponding relation between the problem type and the customer service unit, and sending the first problem to the target customer service unit;
the target customer service unit is configured to identify a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, where the first entity word is a first noun included in the first question, and the first attribute word is an adjective word modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
When the first question input by the user is received, the routing unit may determine the customer service unit with the minimum load according to the pre-stored load information corresponding to each customer service unit, and send the first question to the customer service unit with the minimum load. The customer service unit may determine a response to the first question and provide the response to the user according to the method described in embodiment 1; the routing unit may further perform intent recognition according to the first question, so as to recognize a target question type corresponding to the first question, determine a target customer service unit corresponding to the target question type according to a correspondence relationship between the question type stored in the routing unit and the customer service unit, and send the first question to the target customer service unit, where the target customer service unit may determine a response to the first question by using the method described in embodiment 1 and provide the response to the user. The routing unit may determine, after determining the target problem type, a plurality of target service units corresponding to the determined target problem type, and then send the first problem to the target service unit with the smallest load according to the determined load of each target service unit.
Based on the foregoing embodiments, fig. 3 is a schematic process diagram of training a recognition model in an intelligent customer service robot according to an embodiment of the present invention, as shown in fig. 3:
research personnel classify the stored data information and divide the data information into structured data, semi-structured data and unstructured data, aiming at the classified data, the intelligent customer service robot analyzes the semi-structured data to extract semi-structured information, identifies entity words and attribute words in the unstructured data, and normalizes and integrates the structured data, the extracted semi-structured information, the identified entity words and attribute words, namely integrates heterogeneous data to obtain integrated data information, and obtains entity words, attribute words and semantic information which accord with standard robot information from the integrated data information, and then trains aiming at the integrated data, wherein the specific process is as follows: all entity words form an entity word set, all attribute words form an attribute word set, all semantic information forms a semantic information set, all questions form a question set, a feature vector containing a weight value and each set is determined according to the weight value corresponding to each set, and the feature vector and responses in classified data are input into an identification model, so that training of the identification model in the intelligent customer service robot is completed.
In the embodiment of the invention, the customer service module forms a first feature vector according to a first entity word set, a first attribute word set, first semantic information and a first question which are identified in a first question input by a user; the first feature vector is input into a pre-trained recognition model, and the response of the first question is determined and provided, so that the accuracy of providing the response by the intelligent customer service robot is improved, and the user experience is further improved.
For the system/apparatus embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and reference may be made to some descriptions of the method embodiments for relevant points.
It is to be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or operation from another entity or operation without necessarily requiring or implying any actual such relationship or order between such entities or operations.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely application embodiment, or an embodiment combining application and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, it is intended that the present invention cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.
Claims (11)
1. A customer service-based response providing method, comprising:
receiving a first question input by a user;
according to a preset pattern recognition rule, recognizing first entity words, first attribute words and first semantic information included in the first question, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun;
forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question;
inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
2. The method of claim 1, wherein the training process of the recognition model is:
aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a reply, and each question about the reply identifies a second entity word, a second attribute word and second semantic information included in the reply and each question in the first sample subset according to a preset pattern identification rule;
forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set;
and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding answer.
3. The method of claim 2, wherein the sample subset includes structured data represented by basic data types, unstructured data represented by pictures and speech, and semi-structured data obtained after parsing with a corresponding processing tool.
4. The method of claim 1, wherein the method further comprises:
saving the first question and the response to the first question.
5. The method of claim 4, wherein the method further comprises:
determining each second sample subset according to each stored first question and the response aiming at the first question according to a set time interval, wherein each second sample subset comprises a response and each first question related to the response;
for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule;
forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set;
and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
6. A customer service based response providing apparatus, comprising:
the receiving module is used for receiving a first question input by a user;
the customer service module is used for identifying first entity words, first attribute words and first semantic information included in the first question according to a preset pattern identification rule, wherein the first entity words are first nouns included in the first question, and the first attribute words are adjectives for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
7. The apparatus of claim 6, wherein the apparatus further comprises:
the first training module is used for aiming at each first sample subset in a training sample set, wherein each first sample subset comprises a response, and identifying the responses in the first sample subset and each question comprises a second entity word, a second attribute word and second semantic information according to a preset pattern identification rule; forming a second feature vector according to the identified second entity word set, the identified second attribute word set, the identified second semantic information and the identified problem set; and training the recognition model according to the second feature vector corresponding to each first sample subset and the corresponding answer.
8. The apparatus of claim 6, wherein the apparatus further comprises:
and the storage module is used for storing the first question and the answer aiming at the first question.
9. The apparatus of claim 8, wherein the apparatus further comprises:
the second training module is used for determining each second sample subset according to the stored first question and the answer aiming at the first question at a set time interval, wherein each second sample subset comprises one answer and each first question related to the answer; for each second sample subset, identifying the answers in the second sample subset and a third entity word, a third attribute word and third semantic information included in each first question according to a preset pattern identification rule; forming a third feature vector according to the identified third entity word set, the identified third attribute word set, the identified third semantic information and the identified first question set; and training the recognition model according to the third feature vector and the answer corresponding to each second sample subset.
10. The apparatus of claim 6, wherein the customer service module comprises: a routing unit and at least two customer service units,
the routing unit is used for determining the customer service unit with the minimum load according to the stored load information corresponding to each customer service unit and sending the first problem to the customer service unit with the minimum load;
each customer service unit is used for identifying a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, wherein the first entity word is a first noun included in the first question, and the first attribute word is an adjective word for modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
11. The apparatus of claim 6 or 10, wherein the customer service module comprises:
a routing unit and at least two customer service units,
the routing unit is used for identifying a target problem type corresponding to the first problem according to the first problem, determining a target customer service unit corresponding to the target problem type according to the stored corresponding relation between the problem type and the customer service unit, and sending the first problem to the target customer service unit;
the target customer service unit is configured to identify a first entity word, a first attribute word and first semantic information included in the first question according to a preset pattern identification rule, where the first entity word is a first noun included in the first question, and the first attribute word is an adjective word modifying each first noun; forming a first feature vector according to the identified first entity word set, the identified first attribute word set, the identified first semantic information set and the identified first question; inputting the first feature vector into a pre-trained recognition model, determining and providing an answer to the first question.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811513766.4A CN111382234A (en) | 2018-12-11 | 2018-12-11 | Reply providing method and device based on customer service |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811513766.4A CN111382234A (en) | 2018-12-11 | 2018-12-11 | Reply providing method and device based on customer service |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111382234A true CN111382234A (en) | 2020-07-07 |
Family
ID=71219174
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811513766.4A Pending CN111382234A (en) | 2018-12-11 | 2018-12-11 | Reply providing method and device based on customer service |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382234A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598445A (en) * | 2013-11-01 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Automatic question-answering system and method |
| US9262406B1 (en) * | 2014-05-07 | 2016-02-16 | Google Inc. | Semantic frame identification with distributed word representations |
| CN107301213A (en) * | 2017-06-09 | 2017-10-27 | 腾讯科技（深圳）有限公司 | Intelligent answer method and device |
| CN108052577A (en) * | 2017-12-08 | 2018-05-18 | 北京百度网讯科技有限公司 | A kind of generic text content mining method, apparatus, server and storage medium |
| CN108632137A (en) * | 2018-03-26 | 2018-10-09 | 平安科技（深圳）有限公司 | Answer model training method, intelligent chat method, device, equipment and medium |
-
2018
- 2018-12-11 CN CN201811513766.4A patent/CN111382234A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598445A (en) * | 2013-11-01 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Automatic question-answering system and method |
| US9262406B1 (en) * | 2014-05-07 | 2016-02-16 | Google Inc. | Semantic frame identification with distributed word representations |
| CN107301213A (en) * | 2017-06-09 | 2017-10-27 | 腾讯科技（深圳）有限公司 | Intelligent answer method and device |
| CN108052577A (en) * | 2017-12-08 | 2018-05-18 | 北京百度网讯科技有限公司 | A kind of generic text content mining method, apparatus, server and storage medium |
| CN108632137A (en) * | 2018-03-26 | 2018-10-09 | 平安科技（深圳）有限公司 | Answer model training method, intelligent chat method, device, equipment and medium |
Non-Patent Citations (1)
| Title |
|---|
| 李德毅等: "《中国科协新一代信息技术系列丛书 人工智能导论》", 31 August 2018, 北京：中国科学技术出版社, pages: 191 - 192 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11004013B2 (en) | Training of chatbots from corpus of human-to-human chats | |
| US20190272269A1 (en) | Method and system of classification in a natural language user interface | |
| US11223723B2 (en) | Call center system having reduced communication latency | |
| CN110232109A (en) | A kind of Internet public opinion analysis method and system | |
| KR20180022762A (en) | Method, system and computer-readable recording medium for providing customer counseling service using real-time response message generation | |
| US11258902B2 (en) | Partial automation of text chat conversations | |
| US8954360B2 (en) | Semantic request normalizer | |
| US11238132B2 (en) | Method and system for using existing models in connection with new model development | |
| CN104008160A (en) | Method and system of indistinct logic chatting robot for realizing parallel topic control | |
| CN108268450B (en) | Method and apparatus for generating information | |
| US20190354557A1 (en) | System and Method For Providing Intelligent Customer Service | |
| CN113076423A (en) | Data processing method and device and data query method and device | |
| CN109766451A (en) | A kind of cloud computing platform and its scheduling, data analysing method | |
| CN111523328A (en) | Intelligent customer service semantic processing method | |
| CN113987151A (en) | Intelligent customer service reply method and device | |
| US20210406973A1 (en) | Intelligent inquiry resolution control system | |
| Aattouri et al. | Modeling of an artificial intelligence based enterprise callbot with natural language processing and machine learning algorithms | |
| CN113505606B (en) | Training information acquisition method and device, electronic equipment and storage medium | |
| CN111382234A (en) | Reply providing method and device based on customer service | |
| Koch et al. | A reference architecture for on-premises chatbots in banks and public institutions | |
| Shayegan et al. | A lexicon weighted sentiment analysis approach on Twitter | |
| CN116561288B (en) | Event query method, device, computer equipment, storage medium and program product | |
| CN114818740B (en) | Man-machine cooperation method and system based on domain knowledge graph | |
| US20220101094A1 (en) | System and method of configuring a smart assistant for domain specific self-service smart faq agents | |
| CN117033540A (en) | Report generation method, report generation device, electronic equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |