Outline of the solution

Through the application of the industry's top speech recognition (ASR), semantic understanding (NLU) and speech synthesis (TTS) technologies, the conference intelligent notification interactive robot can notify participants and communicate with them according to the preset outbound call logic. Through multiple rounds of dialogue and speech guidance, it can achieve the goal of participants to obtain the notice of participation and accurately understand the contents of the conference

 Functional framework

The overall logical architecture of the system is the core of the system. Huilan has fully studied the business scenario of conference notification, and designed the following logical architecture of the solution considering the robustness, scalability, interoperability, stability, portability and security of the architecture.　　

Logical structure

Through the integration of industry artificial intelligence + Voice big data + operators and other resources, through the use of voice recognition (ASR), speech synthesis (TTS), semantic understanding (NLP) and other technologies, through the creation of conversation logic suitable for industry enterprises, it can accurately identify user intention and conversation management, accurately mine user needs, define user profiles, screen user intention, and let the meeting notice interact More natural

Application business

1. Content notification

　① notification content template ② autonomous session configuration

2. Phone selection

　　Based on the notice content, the logic execution rules of business scenario are designed, and the automatic outbound call is carried out according to the department organization address book. At the same time, according to the business volume and the implementation of the standard intelligent adjustment of outbound call operations, without human intervention, using intelligent voice technology to build anthropomorphic interaction, so that customers seem to communicate with real people. The unified intelligent call center solution of China Union of science and technology intelligently recognizes the voice information of users, and intelligently assigns it to intelligent robot seats and artificial phone seats

3. dial

       The system supports automatic dialing. According to the selected address book, you do not need to enter the number manually. You can set the time and frequency according to the dialing scenario. The background automatically records the call content, and can convert it into text in real time and accurately, and can call at any time (the total number of outbound calls of the outbound robot in a day can reach 500-1000, which can ensure that all participants are notified). The unified intelligent call center solution of China Union of science and technology adopts the dual service mode of robot and artificial. The intelligent robot solves the common problems, and the artificial seat solves the personalized problems, so as to reduce the work pressure and operation cost of the artificial customer service system in a large range

4. Replay mechanism

If the participants don't receive the call, they can replay it according to the replay policy settings (for example, call again at the interval of 10 minutes, and the maximum number of repeated calls is 3 times

5.Multiple rounds of dialogue

According to the context information of user conversation, some entity related information may be omitted from the user's current problem. Entity replacement and completion can be realized based on reference resolution and ellipsis recovery technology. It supports multiple sessions in the actual scenario, and the semantic expression function supports more flexible interaction scenarios of customer configuration

6. Dialogue support interrupt Q & A

　　During the robot's speaking process, it supports customers to interrupt the conversation at any time. When the robot interrupts the robot's speaking for the information that customers have questions or want to ask, the robot can respond quickly and give feedback, close to the real conversation between people, further increase the trust of customers, and make the follow-up dialogue more fluent. The unified intelligent call center solution of China Union of science and technology provides intelligent recording, analysis and learning of unknown problems, and knowledge learning is efficient and controllable

7. Send SMS, email notification

　　The system supports the docking of SMS and email systems; after the notification is completed, the participants are notified in the form of SMS and email to ensure the multi-channel access of notification information. The unified intelligent call center solution of China Union of science and technology provides full maintenance data report service, which can carry out multi maintenance data report statistical analysis and data mining for user conversation habits, dialogue data mining, user demand hot issues, etc

8. Outbound call data statistics

　　Multi dimensional data statistical analysis shows the results of outbound calls. The total volume of outbound calls, outbound success volume / rate, missed calls volume / rate, completed volume, call duration and average call duration in a unit time are used for data statistics, which can be directly displayed through histogram, line chart and pie chart. Provide data support for improving outbound call quality. China Science and technology union unified intelligent call center solution provides interface optimization, clustered system deployment, high concurrency, high reliability

System technical support

1. NLU semantic recognition accuracy

The system transcends the keyword retrieval technology, adopts the natural language semantic understanding technology based on deep neural network algorithm and syntax dependency algorithm, which can intelligently expand the questions, automatically generate a variety of similar questions, and accurately identify the semantics of the sentences sent by users

2.Rich voice and emotion of real people

　　The robot adopts the real person voice recording mode to reduce the rigid and emotionless dialogue of the traditional robot, increase the trust between people, and add related colloquialism, warm field language and other words in the dialogue process, so as to ensure the quality of the dialogue and make it more acceptable to customers

3. ASR voice text conversion

Through ASR, the system can accurately identify and understand the user's conversation content and answer the user's questions

4. Knowledge learning system

Based on deep learning technology and language base, the system can automatically expand similar problems, and carry out semantic analysis and matching based on semantic analysis engine intelligence. The accuracy of the robot's answer is more than 90%. Participants' unknown scripts and Q & A can be recorded automatically to enrich the knowledge base quickly. According to the unknown questions, they can create questions and answers to add to the existing knowledge base, or connect to the existing problems. Through continuous independent learning, they can enrich the existing knowledge base, making the machine people smarter and easier to deal with more problems

5. Text To Speech (TTS)

　　Text To Speech (TTS) is based on deep neural network technology, which can meet the needs of speech generation from known text and open the closed-loop of human-computer interaction. A variety of voice color options, support customized volume and speed, provide customized own domain lexicon and personalized voice service for enterprise customers, as if communicating with real customer service