Search

CN-121985194-A - AI live conversation operation dynamic generation and field control collaborative management system and method

CN121985194ACN 121985194 ACN121985194 ACN 121985194ACN-121985194-A

Abstract

The invention discloses a live broadcast session generation and field control collaborative management system and method based on an AI large model, and belongs to the technical field of intelligent live broadcast. The system includes a master control system that integrates a speech generation engine with a field controlled collaboration engine. The conversation generation engine generates a conversation through an AI large model based on commodity information and real-time live broadcast data, and dynamically optimizes through conversation round number management and compliance filtering. The field control cooperative engine captures the multi-platform barrage and automatically executes popup window, interaction and flow control according to preset rules. The key is that the system is provided with a linkage arbitration module for carrying out unified scheduling and conflict arbitration on the output instructions of the two engines, thereby realizing deep collaboration and closed-loop management of speech operation content and operation actions. The invention solves the problems of disjoint content generation and field control execution, low automation degree and complicated operation in the prior art, and realizes high-quality and full-automatic live broadcast operation.

Inventors

  • CHEN RONGBIN
  • GUO YADONG

Assignees

  • 江门职业技术学院
  • 江门初牧网络科技有限公司

Dates

Publication Date
20260505
Application Date
20251229

Claims (10)

  1. 1. The AI live conversation operation dynamic generation and field control collaborative management system is characterized by comprising a main control system, wherein the main control system is integrated with: The conversation generation engine is used for generating a live conversation through an AI large model based on the input commodity structured information and the real-time live broadcasting room data, and carrying out round number row duplication and compliance filtering on the generated conversation; the field control cooperative engine is used for capturing live bullet screens in real time through the multi-platform interface adaptation layer and carrying out intention recognition, and loading and executing preset field control rules to automatically control commodity bullet windows, interactive bullet screen sending and live broadcasting processes; And the linkage arbitration module is respectively connected with the speaking operation generation engine and the field control cooperative engine and is used for receiving and arbitrating instruction events from the two engines, generating a unified execution sequence according to a preset priority and driving the live client to execute corresponding operations.
  2. 2. The AI live speech dynamics generation and field control collaborative management system of claim 1, wherein the speech generation engine comprises: the multi-source information fusion unit is used for accessing the commodity database and the live broadcast platform data stream; The AI large model service unit calls a generated pre-training large model and generates an initial conversation text based on the fused information; the voice operation round number management unit calculates Jack similarity between a new voice operation and a same topic calendar historical narrative operation set by adopting a text fingerprint technology based on the xiao Ha th, and triggers voice operation reconstruction if the similarity exceeds a set threshold; and the compliance filtering unit is internally provided with a dynamically updated forbidden word stock and is used for scanning and replacing the dialogue text.
  3. 3. The system of claim 2, wherein the real-time live broadcasting room data includes current explanation commodity ID, real-time online number of people, bullet screen hot word cloud and user interaction frequency.
  4. 4. The dynamic generation and field control collaborative management system for AI live talk according to claim 1, wherein the field control collaborative engine comprises: the rule driving executor is used for analyzing the field control rule defined by the structural configuration file, and the rule at least comprises a trigger condition, an execution action and an action parameter; the triggering condition is a logic judgment expression for the state variable of the live broadcasting room, and the execution action comprises at least one of displaying a commodity popup window, sending a system popup screen and executing up-and-down broadcasting operation.
  5. 5. The dynamic generation and field control collaborative management system for AI live conversation as set forth in claim 4 wherein the structured profile is in YAML or JSON format and the live room state variables comprise system time, current commodity ID, specific keyword hit status, and online population interval.
  6. 6. The system of claim 1, wherein the preset priority policy of the coordinated arbitration module is that the user real-time consultation response instruction priority is higher than a preset task instruction, and the preset task instruction priority is higher than a conventional conversation push instruction.
  7. 7. The AI live talk dynamic generation and field control collaborative management system according to claim 1, wherein the linkage arbitration module comprises: the dynamic weight calculation unit is used for calculating a dynamic execution weight value W for each instruction event to be processed, and the calculation formula is as follows: wherein P_base is a basic priority value preset according to the event type, F_time is a timeliness factor attenuated along with the waiting time of the event, F_business is a service value factor dynamically estimated based on context data of a live broadcasting room in real time, and alpha, beta and gamma are configurable weight coefficients; The linkage arbitration module is configured to periodically sort from high to low according to the dynamic execution weight value W of each event to generate the unified execution sequence.
  8. 8. An AI live session dynamic generation and field control collaborative management method, which applies the AI live session dynamic generation and field control collaborative management system according to any one of claims 1-7, comprising the following steps: S1, initializing a system, loading commodity information, presetting field control rules and a compliance word stock; S2, continuously acquiring real-time data by a speaking operation generation engine, calling an AI large model to generate speaking operation, performing weight removal and filtering, and submitting the available speaking operation to a linkage arbitration module; S3, capturing and analyzing bullet screen events in real time by the field control cooperative engine, monitoring the state of a live broadcasting room to judge the triggering condition of the field control rule, and submitting the events to be responded to the linkage arbitration module; S4, the linkage arbitration module arbitrates and sorts various received instruction events according to a preset strategy to generate a unified execution instruction sequence at the current moment; And S5, driving an external interface to execute corresponding conversation operation pushing, barrage replying, popup window control or flow operation by the system according to the execution instruction sequence.
  9. 9. The method for dynamically generating and cooperatively managing field control of AI live broadcasting session according to claim 8, wherein in step S2, the ranking is specifically that text fingerprints are calculated for the generated session, a history fingerprint set corresponding to the session subject is queried, if the maximum similarity exceeds a threshold T, the AI big model is required to be reconstructed by modifying the prompting word parameters until the similarity between the new session fingerprint and all fingerprints in the history set is lower than the threshold T.
  10. 10. The method according to claim 8, wherein in step S4, the coordinated arbitration module delays or suspends the low-priority periodic task instruction when the high-priority event is triggered.

Description

AI live conversation operation dynamic generation and field control collaborative management system and method Technical Field The invention relates to the technical field of artificial intelligence, in particular to an AI live phone operation dynamic generation and field control collaborative management system and method. Background With the rapid development of live E-commerce, the improvement of the automation and the intelligence level of live operation has become an industry core appeal. The prior art attempts to solve the problem of low manual operation efficiency from different angles, and mainly forms two technical routes, namely, the automation focusing on a live broadcast mode and an interaction mode and the intelligentization focusing on preparation before live broadcast. The first route is represented by a virtual digital person live broadcasting technology, as shown in Chinese patent document with publication number of CN116248909A, a virtual digital person live broadcasting method based on taking over and hosting cooperation is proposed, and the core of the method is that operators prepare speaking and materials in advance, make video content and configure a plurality of reply videos aiming at preset problems. In the live broadcast process, the host mode is mainly adopted, and when a problem outside the preset is encountered, a real person can intervene to take over or the system tries to search answers through a network. The scheme combines the advantages of stable output of the virtual person and flexible response of the real person through the cooperation of the hosting and taking over. However, the speaking and replying content is essentially a prerecorded still video, and the capability of generating and adjusting the content in real time according to the dynamic state of the live broadcasting room, such as the emotion of the user, the change of the number of online people and the like, and the field control actions, such as popup, interaction and the like, are not in intelligent linkage with the speaking content, so that the degree of automation is limited. The second route focuses on the intelligent simplification of the preparation work before the start of the broadcasting by utilizing data, and as shown in Chinese patent document with publication number of CN120223914A, a live broadcasting processing method is disclosed. The scheme significantly improves the efficiency of choice and speech preparation. However, the technical key point is that the plan generation before live broadcast is focused, once the live broadcast starts, the generated voice operation is a relatively fixed text, and dynamic optimization and adjustment can not be performed according to real-time interaction data, such as bullet screen questioning, user behaviors and the like, in the live broadcast process. In addition, the scheme does not relate to field control management in the live broadcast process, such as automatic popup window, intelligent reply, flow control and the like, and a session of speaking generation and live broadcast execution is still split. It can be seen that the prior art has a common limitation that the generation of live content, i.e. the talk operation, and the execution of live operation, i.e. the field control, are in a separated state, and cannot form an intelligent system of a closed loop. This results in: 1. Content rigidification, namely, a conversation operation cannot be dynamically generated and optimized based on a real-time interaction context, and the user experience is single; 2. the marketing actions such as popup windows, sales promotions and the like cannot be accurately synchronized with the real-time generated speaking contents, and a break point exists in a conversion link; 3. The operation is complicated, and although the single-point efficiency is improved, the anchor or the operator still needs to perform high-strength manual coordination and operation in a plurality of links such as call guiding, interactive reply, flow control and the like, and the full-flow automation degree is low. Therefore, a technical scheme capable of deeply coupling and integrally cooperating real-time intelligent speech operation generation and accurate automatic field control execution in a live broadcast whole process is needed in the art so as to truly realize end-to-end automation from content creation to operation execution. Disclosure of Invention Based on this, it is necessary to provide an integrated system and method capable of deeply coupling intelligent content generation with accurate operation execution for overcoming the technical problems of the prior art that the live phone generation and the field control management are mutually disjointed and the degree of automation closed loop is low. In order to solve the technical problems, the invention provides a direct broadcast session generation and field control collaborative management system based on an AI large model, and the c