CN-121983045-A - AI broadcasting interaction platform and method

CN121983045ACN 121983045 ACN121983045 ACN 121983045ACN-121983045-A

Abstract

The invention belongs to the technical field of nuclear power artificial intelligence, and particularly relates to an AI broadcasting interaction platform and method. The system comprises a broadcast control system, a broadcast task decomposition module and a semantic understanding module, wherein the broadcast control system supports the docking of different brands of broadcast equipment through an adapter, the broadcast task decomposition module decomposes and distributes tasks to a plurality of broadcast equipment for execution, the semantic understanding module generates broadcast tasks from input contents, and the broadcast data analysis module analyzes abnormal data and execution frequency of the broadcast tasks and proposes workflow optimization suggestions. The platform has the beneficial effects that the platform improves the execution efficiency and reliability of the broadcasting task, supports various access modes, and simultaneously provides an intelligent voice recognition function to improve the automation level of the broadcasting task.

Inventors

LI SI
ZHANG KAIPENG
Kou Hanxue
CHEN ZHIGEN
Xue Jiaen
PAN ZHENGXIANG
XIE BIN
Wu Kangqiang
WANG HENG

Assignees

福建福清核电有限公司

Dates

Publication Date: 20260505
Application Date: 20260112

Claims (9)

1. The AI broadcasting interaction platform is characterized by comprising a broadcasting control system, a broadcasting task decomposition module and a semantic understanding module, wherein the broadcasting control system supports the butting of broadcasting equipment of different brands through an adapter, the broadcasting task decomposition module decomposes and distributes tasks to a plurality of broadcasting equipment for execution, the semantic understanding module generates broadcasting tasks from input contents, and the broadcasting data analysis module analyzes abnormal data and execution frequency of the broadcasting tasks and proposes workflow optimization suggestions.
2. The AI broadcasting interactive platform as set forth in claim 1, wherein the broadcasting control system comprises a broadcasting controller and a plurality of broadcasting adapters, wherein the broadcasting controller is respectively connected with the broadcasting adapters, each broadcasting adapter is connected with a broadcasting device, and the broadcasting controller provides a plurality of broadcasting interfaces to access broadcasting tasks of a telephone system, a short message system, a webpage, an APP and a small program.
3. The AI broadcasting interactive platform as set forth in claim 1 wherein the broadcasting control system comprises an adapter manager and a task scheduling module, the adapter manager automatically detecting currently connected broadcasting devices and assigning corresponding broadcasting adapters thereto, and the task scheduling module is responsible for decomposing and distributing broadcasting tasks to ensure that different broadcasting devices can synchronously execute tasks.
4. The AI broadcasting interactive platform as set forth in claim 1 wherein said semantic understanding module performs intelligent voice dialogue to achieve voice input of the broadcasting task by dialing a phone call, and directly generates the broadcasting task by voice translation and natural language processing.
5. The AI broadcasting interactive platform as set forth in claim 1 wherein said semantic understanding module generates a broadcasting task by extracting key information of the broadcasting task by the semantic understanding module when a user sends a short message.
6. The AI broadcasting interactive platform as set forth in claim 1 wherein said semantic understanding module generates a broadcasting task when a user directly inputs the content and execution conditions of the broadcasting task via a web page, APP, applet.
7. The AI broadcasting interactive platform as set forth in claim 1 wherein the broadcasting data analysis module analyzes the voice content by NLP algorithm to extract key information such as broadcasting time, place, and object to generate executable broadcasting tasks.
8. The AI broadcasting interactive platform as set forth in claim 1, wherein the broadcasting task in the broadcasting task decomposition module is decomposed into a plurality of subtasks, which are independently executed by the different brands of the broadcasting systems in the docking according to their own conditions.
9. An AI broadcasting interaction method is characterized by comprising the following steps: Collecting dialogue data of a user, including voice and text; labeling the collected data, wherein the labeling data comprises intention, entity and relation of the dialogue; training a model by using the labeling data, wherein the training model comprises a voice recognition model, a semantic understanding model and a dialogue generating model; The test model is the accuracy and efficiency of the test model; Deploying the model, namely deploying the trained model into a system to realize an intelligent dialogue function; and analyzing broadcast data, namely recording the execution condition of a broadcast task in real time and automatically generating a broadcast statistical report.

Description

AI broadcasting interaction platform and method Technical Field The invention belongs to the technical field of nuclear power artificial intelligence, and particularly relates to an AI broadcasting interaction platform and method. Background Most of the existing nuclear power plants already build a cable broadcast system. And answering the call of the request broadcast through the manual position of each factory area, and then initiating the broadcast to the appointed area by the manual position. This approach has the following problems: 1. the efficiency is low, when the user answers the call manually, the user can not broadcast, and when the user sends the broadcast, he can not answer the call; 2. the time is limited, the work cannot be carried out for 24 hours without interruption, and the broadcasting requirement of overtime during overhaul is difficult to meet; 3. The place is limited, the broadcasting station is required to initiate the broadcasting, and other places cannot initiate the broadcasting; 4. The broadcast has no record, namely, the broadcast is sent to which people, and the broadcast is not recorded, and in special cases, the people can only trace back by brain; during heavy overhaul, the manual broadcasting mode is difficult to meet heavy broadcasting requirements, and the overall working efficiency is affected. The nuclear power unit normally operates for one day, and millions or even tens of millions of economic benefits are created. The working efficiency of nuclear power overhaul is improved, the cost of the overhaul can be reduced, and the unit can be operated normally early. Disclosure of Invention The invention aims to provide an AI broadcasting interaction platform and method, which can be used for docking broadcasting systems of different brands and supporting task decomposition, adapter hot plug, semantic understanding and data analysis optimization. The technical scheme includes that the AI broadcasting interaction platform comprises a broadcasting control system, a broadcasting task decomposition module and a semantic understanding module, wherein the broadcasting control system supports the butting of broadcasting equipment of different brands through an adapter, the broadcasting task decomposition module decomposes tasks and distributes the tasks to a plurality of broadcasting equipment for execution, the semantic understanding module generates broadcasting tasks from input contents, and the broadcasting data analysis module analyzes abnormal data and execution frequency of the broadcasting tasks and proposes workflow optimization suggestions. The broadcast control system comprises a broadcast controller and a plurality of broadcast adapters, wherein the broadcast controller is respectively connected with the broadcast adapters, each broadcast adapter is connected with a broadcast device, and the broadcast controller provides a plurality of broadcast interfaces and can be accessed to the broadcasting tasks of a telephone system, a short message system, a webpage, an APP and a small program. The broadcasting control system comprises an adapter manager and a task scheduling module, wherein the adapter manager automatically detects currently connected broadcasting equipment and distributes corresponding broadcasting adapters for the currently connected broadcasting equipment, and the task scheduling module is responsible for decomposing and distributing broadcasting tasks to ensure that different broadcasting equipment can synchronously execute the tasks. And the semantic understanding module is used for carrying out intelligent voice dialogue to realize voice input of the broadcasting task when a user dials a call, and directly generating the broadcasting task through voice translation and natural language processing. And the semantic understanding module extracts key information of the broadcasting task through the semantic understanding module when the user sends the short message to generate the broadcasting task. And the semantic understanding module directly inputs the content and the execution condition of the broadcasting task through the webpage, the APP and the applet to generate the broadcasting task. The broadcast data analysis module analyzes the voice content through an NLP algorithm, extracts key information such as broadcast time, place and object, and generates executable broadcast tasks. The broadcasting task in the broadcasting task decomposition module is decomposed into a plurality of subtasks which are independently executed by the butted broadcasting systems with different brands according to the conditions of the broadcasting systems. An AI broadcasting interaction method, comprising the steps of: Collecting dialogue data of a user, including voice and text; labeling the collected data, wherein the labeling data comprises intention, entity and relation of the dialogue; training a model by using the labeling data, wherein the training model comprises a vo