In 2020, the nine primary results of this project are as follows: (I) Maintain and improve platform functions 1. Main system The user-friendliness of the system platform was optimized according to users’ feedback, including data readjustment and integration for agencies, batch photo upload, and logical message WEB interface. By using the data value-added services, visualization software was adopted to display anomalous cases to enhance the information presentation. 2. GIS system This project improved the GIS system according to the user’s feedback. After the Factory Management Act coming into effect, the factory information and maps were introduced into the system to inspect and manage temporarily registered factories. The function instantaneously provided inspectors with information on the factory conditions and thus improved the overall inspection efficiency. 3. Subsystem By December this year, the inspection app was updated four times to resolve long-standing problems. Functions of picture uploading or location drawing on the sample layout was incorporated into the system. Three educational training courses were held in October. During the training, questions regarding the app were considerably fewer than before, indicating that the current app met the inspectors’ requirements. (II) Update and interconnect related data The platform has collected data from 16 agencies. A total of 386 data sheets were obtained by automatic interconnection. (III) Improve inspection logic modules Different types of examples were established through advanced data analysis based on the professional knowledge, logical thinking, and experience of senior inspecting officers. These examples provided onsite inspection experiences and improved inspection efficiency. This project continued to collect data for the inspection system and establish an analysis platform for examining unusual data. Analysis modules were created using existing industry permits and reporting data to enhance abnormal data inquiry efficiency. The module could automatically introduce analytical results into the app of the subsystem so that inspecting officers could perform examinations based on the unusual data onsite. (IV) Evaluate environmental impact potential This project reviewed previous inspections to determine the required data for analysis. The project personnel carefully identified the correlation logic among data, established modules, and formulated the score and weight of basic and advanced data categories. A conceptual graph of pollution hotspot development was created to enhance the personnel’s knowledge of related topics. (V) Introduce AI applications into the system The project team collected data from books, master theses, journals, press releases, e-newsletters, online articles, as well as used voice importation to incorporate data into the domestic and foreign intelligent voice assistant of the official website. A voice inquiry function was added to Assistant Inspector Grace (created in 2019) with the capability of answering various questions with topics including basic function operations, industry data query, scheduling function, and pollution regulations query. In 2020, this project used the semantic parsing tool supported by Actions on Google and integrated Dialogflow on the server to establish a prototype of voice service. A user initiated the voice service by asking for a control number or the name of a company, and the system would introduce the user into the service framework. The voice service was established through a circular method with steps including semantic analysis, semantic modeling, and semantic training. (VI) Purchase hardware This project purchased a total of 39 sets of Samsung Galaxy Tab A 10.1 (2019), and the accessories included styluses, matte screen protectors, pouches, bags, phone ring holders, and car chargers. Nine sets of EPSON WF-100 mobile printers were also purchased. In addition, this project applied for a 4G internet access service from Chunghwa Telecom for these devices. (VII) Organize training courses on the newly-added functions The educational training courses focused on the topics of “the improvement of existing system functions and operation” and “newly developed functions.” The objectives of the training courses were to introduce the overall framework of abnormal logic, demonstrate general and abnormal data, and interpret abnormal patterns. Each participant was distributed with a mobile device to practice the system to improve their learning outcome. (VIII) Produce one documentary film The project team shot a micro-film to record the overall inspection process. The film demonstrated how the EPA inspectors used digital inspection tools and their inspection procedures. The film was edited and made into a 3-minute documentary film. (IX) Assist in integrating customer feedback and communications management During project implementation, the project personnel kept the spirit of service and continued to provide users with a quality service channel for comments or advice. The project personnel continued to fulfill administrative tasks such as routine work meetings.

Inspection, environmental protection business, information access, smart logic