In order to solve the needs of industrial digital transformation and talent cultivation, the Digital Industry Agency of the Ministry of Digital Development promotes the “AI Industry Practical Application Talent Training Program” (hereinfollowing referred to as the “AIGO Plan”), which takes artificial intelligence as the core and integrates industrial intelligent technology and innovative application talents. Painting”), covering diverse industries such as medical care, industrial manufacturing, commerce, information services, etc., through competitions to match problem-solving companies and problem-solving teams. This project is mainly initiated by the Bureau of Industry, with the dual axes of “industry problem solving x talent problem solving” and “curriculum training”, and various learning and cultivation activities are carried out simultaneously in “learning by doing” and “learning by class”. The Fire Bureau of the Pingtung County Government released the requirements into the plan, and many outstanding technical and technical personnel worked together to solve the current difficulties.
Introduce AI voice system to improve the efficiency of firefighters in disaster relief
When a disaster occurs, every minute, every second, from answering the call to dispatching firefighters to the scene is prime time. Once a disaster occurs, a large number of calls are flooded, and how to quickly handle each rescue call with limited manpower is the key. Take the Fire Department of Pingtung County Government as an example. In 2018, it received nearly 115,000 119 calls, of which regarding 43,000 were disaster relief cases, and nearly 4,000 lines were busy due to abandonment of waiting/connection. Due to the instantaneous influx of a large number of traffic, the system and manpower deployment cannot be overloaded, and many people cannot obtain immediate rescue assistance.
In response to the needs of the Fire Department of the Pingtung County Government, aiteam 45, a problem-solving team composed of three experienced scientific and technological talents, developed a set of “AI Semantic Analysis and Application System for Voice-to-Text Report Calls” to assist on-duty personnel in handling incoming calls.
This system is composed of multiple AI algorithm models, using speech recognition (ASR), named entity recognition (NER), intent recognition (Intent Classification), part-of-speech tagging and other technologies to solve pain points. aiteam45 said, “The use of natural language in various fields of life is mostly composed of common vocabulary. The AI model of general speech recognition and natural language understanding is very suitable for this situation, because it can not only inherit the model trained with a large amount of high-quality information , also has the benefits of easy updating of shared data.”
Overcome mixed language interpretation speed, saving nearly 30% time per pass
In the process, when the public’s report call is connected, the system will use the synthetic voice module (TTS) to greet, and then the content, time and other information of the public’s report will be converted into text and recorded in real time through speech recognition technology (ASR). , and then the post-level expert system judges the event type and location, and feeds back the suggested handling method to the duty personnel to reduce the workload.
In the actual implementation of the AIGO project, it was found during the implementation of the system that Cantonese, Thai or different languages may be used in a phone call, which increases the difficulty of real-time interpretation by speech recognition technology. is the next stage goal. At present, it takes regarding 30-50 seconds to handle a phone call to report a case on average. In the future, following the system integration and improvement are completed, the processing time of each case can be reduced by nearly 30% (5 seconds to 15 seconds).
Make good use of science and technology to continuously improve the disaster relief system
The Fire Department of Pingtung County Government is one of several counties and cities where the 119 dispatch system has just started. It has high expectations for technological disaster relief. Through continuous communication with the problem solvers to optimize their needs, the efficiency of the disaster relief system can be further improved to ensure the safety of the county residents. This is also very helpful to the training of talents and the pain points of government administration. It is hoped that a comprehensive disaster relief system can be built, and the fire department of the Pingtung County Government will spread the results to the fire departments across the country to share the results. This is also the original intention of participating in this plan.
For the problem-solving unit, many phased results have been achieved in the application of the “AI Semantic Analysis Application System for Voice-to-Text Reporting of Report Phones”, and it has also achieved good results in the competition of this project. There is also considerable affirmation that “if the development is completed, not only the police and fire departments will be able to integrate this framework to assist more on-duty personnel to accept calls for reports, but also to adjust for different fields, and the application fields are very wide.”
Taking AIGO as a starting point, we hope to build an intelligent dispatch system steadily
The work of the fire station is very special. It is impossible to predict when the disaster will come. It is very important to continue to do everything more accurately to improve the disaster relief capacity. Therefore, now is just a starting point. From the point of demand, we can see the phased results, and then find the need to continue. The optimized program continues to germinate the vision.
The Pingtung County Government Fire Bureau also gave a positive comment on the experience of cooperating with AIGO this time, “Now the local government only needs to provide pain points, ideas, and the information and administrative resources it has. As long as the pain points and ideas are feasible, Talents and corresponding social resources will be provided to help you achieve this.”
In the future, we will continue to issue questions. From the results, we will continue to optimize the system integration and steadily build a 119 intelligent dispatch system. We will also follow the successful cooperation experience of this time, put forward requirements and match the problem solving team through this plan to strengthen the existing disaster relief. system.