How to build a knowledge-based customer service system? How to realize the key AI in customer service process? How to achieve multi-channel customer access? How to reduce the operation cost of call center?

With the rapid development and application of the Internet and artificial intelligence and the continuous change of user needs, more and more new generation of young customers communicate and handle business with suppliers through Internet channels. The original voice customer service system of Konka Group has been difficult to meet the needs of current users. Internet channels are increasingly in-depth and mature in the application of customer service center with various media as the communication carrier, voice, text, video and other service methods and means, which makes them have the advantages and challenges of low cost, convenience and intelligence, and is also changing the operation mode of traditional customer service center.

In order to meet the needs of more customers, provide fast and convenient intelligent services to customers, improve the customer service ability and customer service quality of Konka Group in the Internet channel, Konka Group needs to build a strong knowledge database, provide full-time, full channel customer service system, and build online intelligent customer service system.

In response to the construction requirements of Konka Group's knowledge-based customer service, Zoomlion made the following iterative upgrading:

Knowledge base content management

The intelligent robot customer service system supports the management of knowledge content by the knowledge base administrator. The system can assign different system management authority to different artificial seats. After the knowledge base administrator obtains the corresponding authority, he can edit and delete the added knowledge classification.

The intelligent robot customer service system can assign different system management rights to different human customer service personnel. After the system assigns the knowledge list management rights to the knowledge base administrator, the knowledge base administrator logs in the background and can edit and delete the knowledge base list in batch.

Batch import and export of knowledge

The intelligent robot customer service system supports the batch import of knowledge. The system provides excel reference template for knowledge import. The knowledge base administrator only needs to maintain the knowledge according to the template, and can import a large number of knowledge in batch at the same time.

At the same time, the system provides a knowledge base created in the management assistant software, which can automatically expand the standard problems to similar problems in batches, reduce the difficulty and workload of knowledge combing, and improve the quality and efficiency of knowledge combing. The system supports the selection of knowledge content for batch export.

　　

Knowledge audit

The system supports the knowledge audit mechanism and can set whether knowledge needs to be audited. When the system opens the knowledge audit function, the knowledge that passes the audit can be regarded as effective knowledge.

　　

Knowledge version control

You can customize the validity period of knowledge to enable and disable knowledge in real time.

The default knowledge validity period of the system is long-term, and the effective time and failure time of the knowledge can be customized; the system can enable and disable the batch knowledge or specified knowledge in real time, and the intelligent robot system will not answer the disabled knowledge.

　　

Knowledge learning

Intelligent customer service system has the function of autonomous learning. Autonomous learning of robot can be divided into two ways: supervised learning and teaching learning.

Supervised learning: the intelligent robot automatically records the questions that are not matched accurately, and carries out semantic analysis on the questions, and intelligently adds similar questions that may be matched. After the customer service personnel passes the click review and recognition, they can summarize different questions into the corresponding answers. At the same time, customer service personnel can quickly add the problem to new knowledge.

Teaching based learning: artificial customer service is the object of knowledge learning of intelligent robot. Intelligent robot can automatically collect knowledge questions and answers according to the dialogue service content of artificial customer service. After being verified and confirmed manually, the knowledge will come into effect.

Rich media knowledge base

The system supports online editing of various formats such as graphics, pictures, audio and video, web pages, customized applications, etc. the edited materials can be directly used in the answers.

The knowledge of knowledge base can contain rich media information such as language, text, graphics, hyperlink, etc. the system adopts artificial intelligence technology based on deep neural network and syntax dependency algorithm, which can intelligently match visitor problems and knowledge base knowledge, and has functions such as fuzzy search, intelligent association, intelligent recommendation, reference resolution, omission and completion.

A Label conversion

In the process of customer consultation, the intelligent robot customer service system will reply questions according to the knowledge base and rule settings. The replies provided by the intelligent robot customer service system include text, rich media and other forms. In some scenarios, the answer provided by the intelligent customer service robot system is connected. In order to improve the end-user experience, the intelligent robot customer service system uses the a-tag function to realize that users can directly access the answers by clicking the content set as the tag without the traditional complex way.

Multichannel

Through accumulating the existing customer service system of Konka call center, the intelligent robot customer service system can support the access channels of wechat end and web end to complete the communication service at the same time. When users use wechat or web services, the system automatically receives the customer consultation information sent by the online customer service system, and the intelligent robot customer service system automatically pushes the answer to the online customer service system, and the online customer service system displays the reply content in the customer terminal.

Dialogue and communication function

Input prompt function

The system has the function of fuzzy question guidance, which can guide customers to ask business related questions and recommend the most likely relevant answers to users. When there are multiple possible corresponding items in the company's business knowledge base, the system can ask customers questions and finally guide customers to get the desired solutions.

Lexical Association

For the unclear user's query vocabulary, the intelligent robot will give relevant questions to guide users to make further choices.

　　

Pinyin Association

According to the Pinyin intelligent guidance of asking questions, through matching Pinyin and knowledge base questions, prompt relevant questions on the question box. It is convenient for users to input, and it can also make the questions more professional and easy to find the answers.

　　

Scenario

The system adopts the artificial intelligence technology based on the deep neural network algorithm, which can intelligently analyze the content of the context dialogue for multiple rounds of dialogue. When the user cannot focus on specific problems, the intelligent robot can intelligently ask questions based on the user's questions, and conduct intelligent guidance for the user's conversation scene, and lock the user's questions through the intelligent questioning and scene guidance The problem.

Multiple conversations

Based on deep neural network, syntax dependency and other artificial intelligence technologies, intelligent robots can accurately understand user problems, and have intelligent capabilities such as multiple rounds of dialogue, intelligent cross examination, etc. the accuracy rate of semantic understanding is more than 95%.

ØSemantic understanding: different from search technology, robots can accurately understand the semantics of Chinese sentences through intelligent word segmentation, part of speech tagging and other technologies;

ØSemantic extension: the system can automatically expand similar problems according to standard problems;

ØAnaphora resolution: it can accurately identify the anaphora content of "this", "that", "it" and other pronouns;

ØEllipsis recovery: according to the content of context dialogue, intelligent understanding of sentence meaning is lack of sentence backbone;

ØMulti round dialogue: the robot realizes multi round dialogue through intelligent rhetorical questions, and gradually reduces the business scope;

ØSemantic reversal: able to accurately understand the expression meaning of reversed word order sentences.

　　

Problem recommendation

For example, when a user enters a question, the robot needs to recommend the question associated with the question vocabulary to the customer, guide the user to view the relevant questions, and realize independent service.

On the basis of answering the user's questions, the intelligent robot can recommend other questions related to this question according to the user's current questions.

If the question is too simple, it will decompose the question step by step and gradually guide the user to ask questions.

　　

Ignore model

In the process of building knowledge base, intelligent robot customer service system supports to set common knowledge points of the same type of products. No matter which product or model the customer consults, intelligent robot customer service system can reply by judging the type of questions the customer consults. In the process of user communication, the intelligent customer service system intelligently identifies the substantive key problems of users through algorithm optimization, decomposes the contents described in the problems and judges the priority. When users consult a general-purpose problem, the system will automatically ignore its auxiliary information such as type number and provide a general-purpose reply.

Chatting and joking

The system supports the robot communication function, which can make fun of the robot, make fun of the robot, and make the robot human-oriented reply.

The system supports the communication function management of robot, and can customize the communication function range of robot.

　　

Session monitoring

The system can automatically check and analyze the dialogue content between intelligent robot and visitors, and can automatically analyze the accuracy of intelligent robot dialogue. At the same time, the intelligent robot system supports managers to customize monitoring keywords, and the system will automatically monitor the conversation and mark the emergence of monitoring targets.

　　

statistical analysis

The system provides multi-dimensional data statistical analysis, which can make statistical analysis on user data, robot data, human customer service data and other multiple data, and display them through numbers, charts and other ways, and support the export of data statistical reports.

Statistical analysis of hot issues ranking

The system supports statistical analysis of hot issues, and can automatically analyze the most concerned issues of users, and rank hot issues according to the number of visits. Displays the ranking of the questions in the leaderboard, the specific questions and the number of times they were visited.

　　

Statistical analysis of visit habits and access channels

The system automatically analyzes users' access habits and access channels, and can intelligently analyze the proportion of users accessing different channels. At the same time, the system supports user-defined start time and end time of data statistics, and can quickly view the statistical analysis data of today, yesterday, 7 days and 30 days.

　　

Statistical analysis of dialogue records

The system can automatically record and analyze dialogue records, not only recording the detailed content of each conversation, but also automatically analyzing the source of visitors, access time, access time, evaluation level, evaluation content and unmatched problems in the dialogue process.

　　

Robot reply data statistics

The system can automatically count and analyze the number of replies, the average conversation time, the accuracy of replies and the proportion of robot conversations. The data is presented by reports and visual icons.

At the same time, the system supports user-defined start time and end time of data statistics, and can quickly view the statistical analysis data of today, yesterday, 7 days and 30 days.

　　

Proprietary vocabulary management

The intelligent robot customer service system effectively combs the definition, recognition and association of industry knowledge through the definition module of professional terms. The proprietary words can be set by themselves, and the system will automatically identify the proprietary words. Proprietary vocabulary can be set up with industry-specific vocabulary and the company's own vocabulary. For example, roller washing and wave wheel refer to roller washing machine and wave wheel washing machine respectively; opposite door and double door refer to opposite door refrigerator and double door refrigerator respectively; super Kimi actually represents Kimi series.

The service accuracy and quality of intelligent robot customer service system can be improved by setting up special vocabulary.

Robot customer service to manual customer service

1.1.1User independent switching

Users can choose to use intelligent robots or artificial customer service to provide customer service. Or when the robot customer service can not provide the customer satisfaction answer, the system supports the user to choose to transfer to the manual service.

The system provides a convenient switching mode, which can switch from intelligent robot to artificial customer service with one key.

　　

Automatic transfer of robot customer service to manual customer service

The system supports the automatic transfer between robot and human customer service. When robot can't answer questions, it can automatically transfer to human service. The system provides a variety of automatic transfer methods, such as useless user evaluation, threshold of robot unable to respond times, etc., which can set the conditions of robot transfer to manual.

　　

Entry Guide

When the intelligent robot customer service system is providing customer service, when it recognizes that the customer needs to make a declaration, the intelligent customer service robot system can use the declaration process and operation settings in the knowledge base to intelligently guide the customer to make a self-service declaration. The time of this function greatly reduces the workload of manual customer service. At the same time, the system greatly improves the customer service experience with simple, easy to understand and detailed guidance.