An intelligent broadcasting method and system, a computer device and a storage medium

Abstract

The present application belongs to the field of digital nuclear power, and relates to an intelligent broadcasting method, comprising: calling an ASR engine to recognize the received incoming call voice and output a plurality of confidence pinyin sequences; performing similarity calculation on the output pinyin sequences and a pinyin sequence list through a pinyin matching algorithm to screen out the highest-similarity candidate result; generating a confirmation question according to the candidate result and an obtained callback telephone number, and interacting with on-site personnel through voice to confirm; generating a broadcast text according to the confirmed information, synthesizing the broadcast text into voice through a TTS engine, and issuing the voice to a voice broadcasting system for broadcasting; monitoring whether the callback telephone is connected after the broadcasting, and triggering a replay or manual intervention when there is no callback within a timeout; and recording process log data for ASR engine and pinyin matching algorithm optimization. The present application also relates to an intelligent broadcasting system, a computer device and a storage medium. The present application ensures complete and accurate broadcast information through multi-confidence output, pinyin matching and dialogue confirmation.

Description

Technical Field

[0001] This invention relates to the field of digital nuclear power technology, and in particular to an intelligent broadcasting method, system, computer equipment and storage medium based on improved voice recognition and dialogue interaction, which is especially suitable for scenarios in nuclear power plants where there are communication blind spots and the efficiency of broadcast calls needs to be improved. Background Technology

[0002] During operation, existing nuclear power plants have cell phone signal dead zones within the plant area, preventing on-site personnel from communicating in real time via mobile devices. In such cases, it is usually necessary to call specific personnel back through the plant's voice broadcast system. However, the ambient noise level is high, often rendering a single broadcast ineffective and requiring multiple repetitions. Especially during major overhauls, communication demands surge, making the traditional method of relying on manual broadcasts inefficient and prone to causing busy communication lines.

[0003] Existing intelligent broadcasting systems can effectively solve the above problems, enabling automatic extraction, broadcasting, and callback tracking of call information. However, in practical applications, there are significant shortcomings in the speech recognition process. Due to differences in staff accents and background noise interference, the results of automatic speech-to-text (ASR) are often inaccurate, leading to errors in the subsequently extracted department names, personnel names, or callback phone numbers. Ultimately, this renders the broadcast ineffective, prevents the establishment of effective communication, and reduces work efficiency. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of information extraction errors and low broadcast effectiveness caused by accents and noise affecting speech recognition in existing intelligent broadcasting systems. It provides an intelligent broadcasting method, system, computer equipment, and storage medium that ensures the integrity and accuracy of broadcast information through multi-confidence output, pinyin matching, and dialogue confirmation, thereby improving communication efficiency in communication blind spots of nuclear power plants and reducing invalid and repeated calls.

[0005] To achieve the above objectives, in one respect, the present invention provides an intelligent broadcasting method, comprising the following steps: Step 1, Voice Access: On-site staff receive calls via the telephone exchange system; Step 2, Speech Recognition Improvement: Receive incoming call voice messages and call the ASR engine to recognize the voice messages, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; the incoming call voice messages include the company name and the person's name; Step 3, Pinyin Matching: The similarity between the Pinyin sequence output in Step 2 and the Pinyin sequence list in the address book is calculated using a Pinyin matching algorithm. The Pinyin sequence with the highest similarity in the address book Pinyin sequence list is selected as the candidate result. Step 4, Confirmation: Obtain the callback phone number; Based on the candidate results selected in Step 3 and the callback phone number, generate a confirmation question and interact with the on-site personnel via voice to confirm the unit name, personnel name, and callback phone number; Step 5, Broadcast Generation and Distribution: Based on the information confirmed in Step 4, generate broadcast text according to predetermined rules, synthesize it into speech through a TTS engine, and distribute it to the speech broadcasting system for voice broadcasting; Step 6, Callback Tracking and Log Recording: Monitor whether a callback call is received after the voice broadcast, and trigger a replay or manual intervention if no callback is received within the timeout period; record the log data of the entire process, and use it to optimize the ASR engine and Pinyin matching algorithm for noise and accent adaptation.

[0006] In one embodiment, the ASR engine supports confidence scoring to recognize incoming call voice and outputs candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences. The candidate Chinese character sequence is composed of candidate Chinese characters at each position in the incoming call voice. There are one or more candidate Chinese characters at each position in the incoming call voice, therefore, there are one or more candidate Chinese character sequences. The candidate Chinese characters in the incoming call voice are combined with their corresponding pinyin to form a candidate pinyin sequence; for polyphonic characters, the corresponding pinyin is generated based on the pronunciation of the people on site and the pronunciation rules. Each candidate Chinese character sequence and its corresponding pinyin sequence are labeled with a confidence level.

[0007] In one embodiment, in step 3, the pinyin matching algorithm uses similarity calculation based on edit distance and / or initial and final weights, and presets a similarity threshold; if the similarity of all candidate results is lower than the similarity threshold, it is determined that information cannot be automatically extracted, and the process is directly transferred to manual intervention.

[0008] In one embodiment, in step 4, the dialogue confirmation is achieved by broadcasting the candidate results through the TTS engine and receiving key or voice feedback from the on-site personnel; if the feedback is negative, voice recognition and matching are allowed to be performed again, or the user is prompted to re-enter, realizing multiple iterations of confirmation until the information is confirmed or a timeout mechanism is triggered to switch to manual intervention.

[0009] To achieve the above objectives, in a second aspect, the present invention also provides an intelligent broadcasting system deployed in a nuclear power plant's computer room to implement the above-mentioned intelligent broadcasting method, comprising: an intelligent broadcasting server that interfaces with a telephone exchange system and a voice broadcasting system respectively, and a contact list database; The address book database is used to generate the pinyin sequence of all unit names and personnel names in the address book and store them as a pinyin sequence list for the address book. The intelligent broadcast server has multiple functional modules deployed internally, including: The speech recognition improvement module is used to receive incoming call voice messages and call the ASR engine to recognize the incoming call voice messages, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; the incoming call voice messages include company names and personnel names; The Pinyin matching module is used to calculate the similarity between the Pinyin sequence output by the speech recognition improvement module and the Pinyin sequence list in the address book using a Pinyin matching algorithm, and select the Pinyin sequence with the highest similarity in the address book as the candidate result. The dialogue confirmation module is used to obtain the callback phone number; based on the candidate results of the pinyin matching module and the callback phone number, it generates a confirmation question and interacts with the on-site personnel via voice to confirm the unit name, personnel name and callback phone number; The broadcast generation and distribution module is used to generate broadcast text according to predetermined rules based on the confirmed information, synthesize speech through the TTS engine, and distribute it to the voice broadcasting system for voice broadcasting. The callback tracking and logging module is used to monitor whether there are callback calls after a voice broadcast, and to trigger a replay or manual intervention if there is no callback after a timeout; it records the log data of the entire process and is used to optimize the ASR engine and pinyin matching algorithm for noise and accent adaptation.

[0010] In one embodiment, when the address book database generates the pinyin sequence of unit name / person's name, if there are polyphonic characters, it generates multiple pinyin sequences.

[0011] In one embodiment, the log data includes the confidence level of each speech recognition, the pinyin matching result, the dialogue confirmation process and result, the number of broadcasts and the callback duration, which are used for statistical analysis and system optimization.

[0012] In one embodiment, access to the system is managed with permissions to ensure that only authorized personnel can view sensitive records; data transmission is encrypted.

[0013] To achieve the above objectives, in a third aspect, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the above-described intelligent broadcasting method.

[0014] To achieve the above objectives, in a fourth aspect, the present invention also provides a computer-readable storage medium storing computer-readable instructions, which, when executed, implement the steps of the above-described intelligent broadcasting method.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: Improved accuracy of information extraction: By introducing multi-confidence output in the speech recognition stage and combining pinyin matching with the address book for secondary screening, the interference of accents and noise is effectively overcome, making the preliminary candidate results more accurate.

[0016] Enhanced system robustness: The introduction of a dialogue confirmation module allows for secondary confirmation of key information through human-computer interaction, preventing broadcasts from becoming invalid due to a single recognition error and significantly improving system reliability in complex noise environments.

[0017] User experience and communication efficiency have been optimized: Automated information extraction and confirmation processes reduce waiting times and busy signal issues associated with manual broadcasting, thus improving communication efficiency. Log recording and model update mechanisms ensure the system can continuously adapt to the field environment and achieve self-optimization. Attached Figure Description

[0018] Figure 1 A flowchart illustrating an embodiment of the intelligent broadcasting method provided by the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent broadcasting system provided in one embodiment of the present invention. Detailed Implementation

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments.

[0022] This invention provides an intelligent broadcasting system deployed in a nuclear power plant's computer room, comprising: an intelligent broadcasting server that interfaces with both a telephone exchange system and a voice broadcasting system, and a contact database; The address book database is used to generate the pinyin sequence of all unit names and personnel names in the address book and store them as a pinyin sequence list for the address book. The intelligent broadcast server has multiple functional modules deployed internally, including: The speech recognition improvement module is used to receive incoming call voice messages and call the ASR engine to recognize the incoming call voice messages, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; the incoming call voice messages include company names and personnel names; The Pinyin matching module is used to calculate the similarity between the Pinyin sequence output by the speech recognition improvement module and the Pinyin sequence list in the address book using a Pinyin matching algorithm, and select the Pinyin sequence with the highest similarity in the address book as the candidate result. The dialogue confirmation module is used to obtain the callback phone number; based on the candidate results of the pinyin matching module and the callback phone number, it generates a confirmation question and interacts with on-site personnel through the telephone exchange system to confirm the unit name, personnel name and callback phone number; The broadcast generation and distribution module is used to generate broadcast text according to predetermined rules based on the confirmed information, synthesize speech through the TTS engine, and distribute it to the voice broadcasting system for voice broadcasting. The callback tracking and logging module is used to monitor whether there are callback calls after a voice broadcast, and to trigger a replay or manual intervention if there is no callback after a timeout; it records the log data of the entire process and is used to optimize the ASR engine and pinyin matching algorithm for noise and accent adaptation.

[0023] In one embodiment, when the address book database generates the pinyin sequence of unit name / person's name, if there are polyphonic characters, it generates multiple pinyin sequences.

[0024] In one embodiment, the log data includes the confidence level of each speech recognition, the pinyin matching result, the dialogue confirmation process and result, the number of broadcasts and the callback duration, which are used for statistical analysis and system optimization.

[0025] In one embodiment, access to the system is managed with permissions to ensure that only authorized personnel can view sensitive records; data transmission is encrypted.

[0026] like Figure 1 As shown, the present invention provides an intelligent broadcasting method, which uses the above-mentioned intelligent broadcasting system and includes the following steps: Step 1: Voice Access In a building of a nuclear power plant with a communication dead zone, worker A needs to contact the control room, but there is no cell phone signal. Therefore, he uses the nearest telephone exchange system to call the intelligent broadcasting system. Step 2, Improvement of speech recognition The voice recognition improvement module receives incoming voice messages and calls the ASR engine to recognize them, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; the incoming voice message is "Please find Li Jianguo from the Maintenance Department 1". The ASR engine supports confidence scoring, recognizes the incoming call voice, and outputs a candidate Chinese character sequence with multiple confidences and the corresponding pinyin sequence. The candidate Chinese character sequence is formed by combining the candidate Chinese characters at each position in the incoming call voice; there is one or more candidate Chinese characters at each position in the incoming call voice. For example, for the character "维" in "维修一科", "维", "为", "微", etc. may be output. Therefore, there is one or more candidate Chinese character sequences. The pinyin combinations corresponding to the candidate Chinese characters at each position in the incoming call voice form the pinyin sequence; for polyphonic characters, the corresponding pinyin is generated according to the pronunciation of on-site personnel and pronunciation rules. For example, "长" is pronounced as "zhang3" in "科长"; each candidate Chinese character sequence and the corresponding pinyin sequence are marked with confidence. The ASR engine combines the confidence and generates several pinyin sequences. For example, Sequence 1: 1: [wei1, xiu, yi, ke]; Sequence 2: [wei2, xiu, yi, ke]; Sequence 3: [wei4, xiu, yi, ke], etc. Step 3: Pinyin matching The pinyin matching module receives multiple pinyin sequences from Step 2, calculates the similarity between these pinyin sequences and the pre-generated address book pinyin sequence list through the pinyin matching algorithm, and screens out the pinyin sequence with the highest similarity of 0.95 in the address book pinyin sequence list as the candidate result, which is higher than the preset threshold of 0.8. In the address book pinyin sequence list, "维修一科" has been converted to "wei xiu yi ke", "维修二科" has been converted to "wei xiu er ke", "李建国" has been converted to "li jian guo", etc., and multiple pinyin sequences have been generated considering polyphonic characters. The pinyin matching algorithm uses the edit distance algorithm and assigns a higher weight to the initials. Step 4: Dialogue confirmation The dialogue confirmation module obtains the callback phone number; the dialogue confirmation module generates a confirmation question based on the candidate result screened out in Step 3 and the callback phone number, and sends a voice inquiry through the telephone exchange system: "Are you looking for Li Jianguo from Maintenance Section 1? Please press 1 to confirm, and press 2 to re-enter."; After hearing the voice inquiry, Staff A presses the phone button "1". After receiving the confirmation signal, the dialogue confirmation module continues to send a voice inquiry: "Is the callback phone number you left for the other party 1234? Please press 1 to confirm, and press 2 to re-enter."; After hearing the voice inquiry again, Staff A presses the phone button "1" again. So far, all the key information that needs to be broadcast by voice has been confirmed. If staff member A presses "2" in the first step, the dialogue confirmation module will prompt "Please repeat the name of the unit and the name of the person you are looking for" and jump back to step 2; if multiple confirmations fail or there is no response after timeout, the dialogue confirmation module will automatically transfer the call to a live broadcast station. Step 5: Broadcast Generation and Distribution The broadcast generation and distribution module receives the confirmed information: Unit = Maintenance Section 1, Name = Li Jianguo, Phone Number = 1234; it generates the broadcast text "Please have Comrade Li Jianguo of Maintenance Section 1 call 1234 immediately" according to the preset template, and then calls the TTS engine to synthesize this text into a clear voice file and distribute it to the voice broadcast system for factory-wide voice broadcast. Step 6: Callback Tracking and Log Recording After the voice broadcast is completed, the callback tracking module starts timing and monitors for any incoming calls from the number "1234". If staff member A calls back within a preset time, such as 5 minutes, the process ends normally; if no call is made within the time limit, the broadcast generation and distribution module will automatically replay the broadcast once. Throughout the process, the multiple candidate pinyin sequences and their confidence scores output by the ASR engine, the similarity scores of pinyin sequence matching, each key confirmation by the user, broadcast content, and callback results are all recorded in detail in the log data. This log data will be used periodically to analyze and identify error cases, and as training data to fine-tune the ASR engine or optimize the weights of the pinyin matching algorithm, thereby achieving continuous improvement of the system.

[0027] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A smart broadcasting method, characterized in that, Includes the following steps: Step 1, Voice Inquiry: On-site staff receive a call; Step 2, Improved Speech Recognition: Receive incoming call voice messages and call the ASR engine to recognize the voice messages, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; The incoming call message includes the company name and the individual's name; Step 3: Pinyin matching: The similarity between the pinyin sequence output in Step 2 and the pinyin sequence list in the address book is calculated using a pinyin matching algorithm. The pinyin sequence with the highest similarity in the address book pinyin sequence list is selected as the candidate result. Step 4: Confirmation: Obtain the callback phone number; Based on the candidate results selected in step 3 and the callback phone number, a confirmation question is generated and interacted with on-site personnel via voice to confirm the unit name, personnel name, and callback phone number. Step 5: Broadcast generation and delivery: Generate broadcast text based on the information confirmed in Step 4, synthesize it into speech through the TTS engine, and deliver it to the speech broadcasting system for voice broadcasting; Step 6: Callback Tracking and Log Recording: Monitor whether there are callback calls after the voice broadcast, and trigger a replay or manual intervention if there is no callback after a timeout; The log data of the entire process is recorded and used to optimize the ASR engine and pinyin matching algorithm for noise and accent resistance.

2. The intelligent broadcasting method according to claim 1, characterized in that, The ASR engine supports confidence scoring, recognizes incoming call voice, and outputs candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences. The candidate Chinese character sequence is composed of candidate Chinese characters at each position in the incoming call voice. There are one or more candidate Chinese characters at each position in the incoming call voice, therefore, there are one or more candidate Chinese character sequences. The candidate Chinese characters in the incoming call voice are combined with their corresponding pinyin to form a candidate pinyin sequence; for polyphonic characters, the corresponding pinyin is generated based on the pronunciation of the people on site and the pronunciation rules. Each candidate Chinese character sequence and its corresponding pinyin sequence are labeled with a confidence level.

3. The intelligent broadcasting method according to claim 1, characterized in that, In one embodiment, in step 3, the pinyin matching algorithm uses similarity calculation based on edit distance and / or initial and final weights, and presets a similarity threshold; if the similarity of all candidate results is lower than the similarity threshold, it is determined that information cannot be automatically extracted, and the process is directly transferred to manual intervention.

4. The intelligent broadcasting method according to claim 1, characterized in that, In step 4, the dialogue confirmation is achieved by broadcasting the candidate results through the TTS engine and receiving key or voice feedback from the on-site personnel. If the feedback is negative, voice recognition and matching are allowed to be performed again, or the user is prompted to re-enter the information, thus achieving multiple iterations of confirmation until the information is confirmed or a timeout mechanism is triggered to switch to manual intervention.

5. An intelligent broadcasting system for implementing the intelligent broadcasting method according to any one of claims 1-4, characterized in that, include: Intelligent broadcasting servers that interface with the telephone exchange system and the voice broadcasting system respectively, as well as a contact database; The address book database is used to generate the pinyin sequence of all unit names and personnel names in the address book and store them as a pinyin sequence list for the address book. The intelligent broadcast server includes: The speech recognition improvement module is used to receive incoming call voice messages and call the ASR engine to recognize the incoming call voice messages, outputting candidate Chinese character sequences with multiple confidence levels and corresponding pinyin sequences; the incoming call voice messages include company names and personnel names; The Pinyin matching module is used to calculate the similarity between the Pinyin sequence output by the speech recognition improvement module and the Pinyin sequence list in the address book using a Pinyin matching algorithm, and select the Pinyin sequence with the highest similarity in the address book as the candidate result. The dialogue confirmation module is used to obtain the callback phone number; based on the candidate results of the pinyin matching module and the callback phone number, it generates a confirmation question and interacts with the on-site personnel via voice to confirm the unit name, personnel name and callback phone number; The broadcast generation and delivery module is used to generate broadcast text based on the confirmed information, synthesize speech through the TTS engine, and deliver it to the voice broadcasting system for voice broadcasting. The callback tracking and logging module is used to monitor whether there are callback calls after a voice broadcast, and to trigger a replay or manual intervention if there is no callback after a timeout; it records the log data of the entire process and is used to optimize the ASR engine and pinyin matching algorithm for noise and accent adaptation.

6. The intelligent broadcasting system according to claim 5, characterized in that, When generating the pinyin sequence of unit name / person's name in the address book database, if there are polyphonic characters, multiple pinyin sequences will be generated.

7. The intelligent broadcasting system according to claim 5, characterized in that, The log data includes the confidence level of each speech recognition, the pinyin matching result, the dialogue confirmation process and result, the number of broadcasts, and the callback duration.

8. The intelligent broadcasting system according to claim 5, characterized in that, Implement access control for the system; encrypt data transmission.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that, When the processor executes the computer-readable instructions, it implements the steps of the intelligent broadcasting method as described in any one of claims 1-4.

10. A computer-readable storage medium storing computer-readable instructions thereon, characterized in that, When the computer-readable instructions are executed, they implement the steps of the intelligent broadcasting method as described in any one of claims 1-4.

Images