Next Article in Journal
TopCycle: A Novel High Performance and Fuel Flexible Gas Turbine Cycle
Next Article in Special Issue
A Suggested Methodology for Assessing the Failure Risk of the Final Slopes of Former Open-Pits in Case of Flooding
Previous Article in Journal
Matching Demolition and Construction Material Flows, an Urban Mining Case Study
 
 
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Chatbot System to Support Mine Safety Procedures during Natural Disasters

1
Department of Civil and Construction Engineering, National Taiwan University of Science and Technology, Taipei City 106335, Taiwan
2
Bureau of Mines, Ministry of Economic Affairs, Executive Yuan, Taipei City 10042, Taiwan
*
Author to whom correspondence should be addressed.
Current address: No.43, Keelung Rd., Sec.4, Da’an Dist., Taipei City 106335, Taiwan.
These authors contributed equally to this work.
Sustainability 2021, 13(2), 654; https://doi.org/10.3390/su13020654
Received: 19 November 2020 / Revised: 30 December 2020 / Accepted: 6 January 2021 / Published: 12 January 2021
(This article belongs to the Special Issue Environmental Hazards: Assessing Risk and Reducing Disaster)

Abstract

:
This study developed a chatbot to improve the efficiency of government activation of mine safety procedures during natural disasters. Taiwan has a comprehensive governmental system dedicated to responding to frequent natural disasters, and the Bureau of Mines has instituted clear procedures to ensure the delivery of disaster alarms and damage reports. However, the labor- and time-consumption procedures are inefficient. In this study, we propose a system framework for disaster-related information retrieval and immediate notifications to support the execution of mine safety procedures. The framework utilizes instant messaging (IM) applications as the user interface to look up information and send messages to announce the occurrence of disaster events. We evaluated the efficiency of the procedures before and after adopting the system and achieved a time-cost reduction of 55.8 min among three types of disaster events. The study has proven the feasibility of adopting novel techniques for decision-making and assures the improvement of the efficiency and effectiveness of the procedure activation.

1. Introduction

Frequent natural disasters in Taiwan include earthquakes, typhoons, and heavy rain caused by severe weather events (e.g., thunderstorms) and synoptic-scale systems [1,2,3]. With the mountainous terrain and complex geology of the Taiwan island, disasters are intensified and cause safety issues. For example, the 7.3-magnitude Chi-Chi Earthquake in 1999 severely damaged Taiwan, causing over 2000 deaths [4,5] and destroying over 50,000 households [5]. In 2009, Typhoon Morakot killed over 600 people, left 60 people missing, caused over 1050 injuries, and destroyed 600 buildings [5]. The August 23 flood, which was caused by a tropical depression in 2018, resulted in seven deaths, and almost 8500 residents were evacuated to temporary shelters [6]. Because floods, typhoons, earthquakes, and landslides caused over 4000 deaths in Taiwan between 1994 and 2019 [5], governmental disaster management has been prioritized. After the Chi-Chi Earthquake in 1999, the Disaster Prevention and Protection Act [7] of Taiwan, which establishes how the governmental system protects the safety of people’s lives and properties, was instituted.
The Bureau of Mines of the Ministry of Economic Affairs (MOEA) in Taiwan has developed procedures for alerting mines of potential risks when a natural disaster occurs. For example, the Procedures for Notification of Earthquakes, Typhoons, and Torrential Rain ensure mines in affected areas are notified and damage information is collected. The Disaster Prevention and Protection Act of Taiwan requires central and local governments to implement measures depending on the functional authorities. Such measures include (a) disaster alarm announcement and delivery, (b) response and alert, (c) people evacuation, (d) rescue and refuge advice, (e) information collection of disaster situations, and (f) survey reports of loss. Hence, the Bureau of Mines has instituted procedures to activate the mechanism of information delivery and collection. When the Central Weather Bureau (CWB) issues warnings based on hazardous meteorological or seismological phenomena, the Mine Safety Division of the Bureau of Mines determines whether to activate the procedures in accordance with the severity. When the procedures are activated, the five Mine Safety Centers of the Mine Safety Division and the Eastern District Office are notified, and the mines in their jurisdictions are alerted via short message service (SMS) or instant messaging (IM) applications. During the disaster, the mines should execute emergency response measures. After the disaster, the Mine Safety Division requires the five Mine Safety Centers and the Eastern District Office to collect information on the mines to understand the situation.
The procedures clarify the activation timings and response. However, the Mine Safety Division encounters two key difficulties during implementation:
  • Labor consumption: Shift work is necessary for the Mine Safety Division to be aware of warnings from websites or applications on smart devices.
  • Decision time cost: It takes time to determine whether the severity of an alarm meets the activating conditions and whether there are mines in the affected area.
Thus, although a comprehensive governmental system is dedicated to responding to frequent natural disasters, and the Bureau of Mines has instituted clear procedures to ensure the delivery of disaster alarms and damage reports, inefficiencies remain that could be mitigated.
This study aims to improve the efficiency of the government activation of mine safety procedures during natural disasters by reducing the labor and time consumption during emergency operations for information delivery. By adopting smart device technology, we develop a novel solution featuring immediate notifications, direct indications, and an intuitive interface for disaster information retrieval to support effective decision-making. Section 2 reviews related works, and Section 3 clarifies the research objective. Section 4 introduces the system framework and design method, and Section 5 describes the system implementation. Section 6 discusses the effectiveness evaluation of the system adoption. Section 7 describes the major advantages achieved in this study, including feasibly utilizing novel techniques for decision support and improving efficiency and effectiveness of procedure activation during disaster events.

2. Related Works

Barriers to information accessibility have declined along with the enhancements made to information and communication technology (ICT). Electronic devices continue to become more powerful [8]. Meanwhile, due to the evolution of wireless protocols and techniques, connections to the Internet are no longer bound to time and location. The fusion of faster networks, efficient applications, improved operating systems, better user interfaces, enhanced display technology, and the extending ecosystem of application markets have made modern smartphones successful [9]. Effortless access, process, and production of information are now possible using smartphones.
Following the widespread use of smartphones, the growth of messaging applications has resulted in conversational agents, also called chatbots in the industry, becoming a trending solution for trivial problems. Conversational agents are software applications that interact with users for different purposes using conversations [10]. Chatbots are widely used in various areas, such as business, medical and healthcarefields, and disaster management. Bavaresco et al. [11] reviewed recent studies and found that chatbots have been most explored in the commerce domain, covering subareas such as e-commerce, sales, shopping, and flight booking. Sun et al. [12] adopted chatbots to personalized recommendation systems for building a virtual sales agent by utilizing deep learning technologies. Jusoh [13] proposed a method to persuade e-customers to buy products through recommendation and negotiation. Majumder et al. [14] built a chatbot to help customers with category-sensitive retrieval techniques and obtained quality improvements. Koetter et al. [15] investigated the potential usage of chatbots in insurance companies and developed a prototype for an exemplary insurance scenario. Cui et al. [16] developed a customer service chatbot taking advantage of product descriptions and user-generated content from e-commerce websites. Laranjo et al. [17] observed that recent studies of chatbots in the medical field are commonly related to mental health and suggested that the utilization of chatbots for health-related issues is an emerging research field. Vaidyam et al. [18] noted that the mental health field could adopt chatbots in psychiatric treatment with proper approaches. Mavropoulos et al. [19] presented a context-aware system framework by combining monitoring and chatbots to benefit ailing patients and assist clinical experts to retrieve information about patients. Griol et al. [20] presented a multimodal conversational coach for physical activity training with sensors to provide meaningful coaching and feedback during sessions. Tsai et al. [21] stated that the advance of smart devices and social networking contributes to a trend of using chatbots for disaster prevention and emergencies.
Several governmental agencies in Taiwan have developed public servicing chatbot products for the general Taiwanese population, and some of the products provide disaster-related information. For example, the National Science and Technology Center for Disaster Reduction (NCDR) developed a chatbot providing over 30 types of public alerts, covering severe weather events, hydrology, traffic, and other issues [22,23]. The Agriculture Department of the New Taipei City Government presented the Landslide Guardian 2.0 to provide instructions on landslides and send notifications to warn users against impending landslides and other severe weather events [24]. In addition, some local governments, including the Taipei City Government, provide local disaster information via their official channels.
In addition to alerts or information dissemination, some institutions and studies have adopted chatbots for collecting damage information and providing advanced services. In Japan, Weathernews Inc., the National Research Institute for Earth Science and Disaster Resilience (NIED), and the National Institute of Information and Communications Technology (NICT) developed a chatbot to collect damage information, summarize the collected information, and provide customized advice to help victims [25,26]. In Taiwan, Tsai et al. [27] proposed a three-module conversation-based system framework, implemented as a chatbot to notify and support school building managers completing damage inspections and report submissions after earthquakes. Kung et al. [28] adopted transfer learning to automatically extract information from the collected damage reports and provide analysis for decision-makers to understand the situation.
Some institutions have adopted chatbot products for decision support in the field of disaster management. For example, Tsai et al. [21] developed a chatbot for water-related disaster decision-makers, enabling users to retrieve requested information for decision-making by directly acquiring the chatbot; the system has been validated through a six-month field test and proven its effectiveness. Furthermore, Chan and Tsai [29] enabled disaster decision-makers to retrieve both static and real-time information directly in natural language with the integration of semantic and temporal term types, improving the capacity of question analysis.
Although several disaster-related chatbot products exist, such products are not feasible for the difficulties that the Bureau of Mines is facing. Most only provide basic severe weather alerts; they seldom provide further instructions for the following procedures. In addition, they are typically designed for the general population rather than experts and decision-makers. Although basic information more or less informs government officers and disaster responders to activate emergency operations, officers often need to look up additional documents, especially for institutions with complicated procedures. Tsai et al. [21] developed the water-related chatbot for another governmental agency with different authorities, and it focused on information retrieval for decision support; thus, they did not discuss the need for a notification to activate the safety procedures.
On the other hand, decision support systems (DSS) are commonly adopted and discussed in various aspects of disaster management. Wallace and De Balogh [30] conceptualized a framework for adopting DSS in the field of disaster management. Fogli and Guida [31] designed a knowledge-centered DSS for emergency managers responding to critical situations, overcoming some limitations of user-centered and activity-centered design in the specific context of DSS. Yoon et al. [32] prototyped a computer-based DSS tool to train emergency responders of governmental transportation agencies for their capacities. Zhou et al. [33] observed that new technologies were contributed to the recent development of DSS, and some studies brought advanced information techniques, geographic information systems (GIS), and agent-based designs to enhance emergency decision making. In Taiwan, many governmental agencies developed computer systems considering respective functional authorities during disaster events to support information gathering and decision-making. For example, the Water Resources Agency and the National Fire Agency utilize different systems to fulfill their tasks [34,35]. However, the Bureau of Mines does not have a specific DSS for disaster management.
Thus, we find that the Bureau of Mines lacks the integration of notifications for operations, which is both labor- and time-consuming, and no existing solutions target this issue.

3. Research Objectives

This study aims to improve the efficiency of the government activation of mine safety procedures during natural disasters by reducing labor and time consumption. We focus on the following four goals:
  • providing immediate announcements of severe disaster events to enhance the efficiency of activating the safety procedures;
  • providing accurate indications to follow the activated procedures in different situations to enhance the effectiveness of the safety procedures;
  • ensuring decision-support information, such as real-time observations, charts, documents, and alerts, are available and sufficient for decision-making; and
  • ensuring the retrieval of decision-support information is intuitive to improve information accessibility.

4. Method

In this study, we propose a system framework for disaster-related information retrieval and immediate notifications to support the execution of mine safety procedures. The framework utilizes IM applications as the user interface to look up information and send messages to announce the occurrence of disaster events. The proposed system has the following features:
  • Question-answering about both real-time information and static documentation
  • User permission management to ensure the security of sensitive government information
  • Immediate announcement of severe disaster events
Figure 1 illustrates the comprehensive system framework. The three major parts are the core system providing essential utilization, the IM application as the user interface, and the external data sources providing immediate disaster alerts, including earthquake reports, torrential rain warnings, and typhoon alerts. The core system consists of the database and three modules: (1) the search module (S-module), (2) the user module (U-module), and (3) the notification module (N-module). All three modules connect to the database, containing user data, interaction logs, and past notification records. The S-module provides the disaster-related information via an information architecture and a series of mapping keywords as the entries to the information. The U-module manages the user’s permission to access sensitive data and provides registration forms for the user to apply for authentication. The N-module retrieves immediate alerts from an external data source, sends announcements to users via IM applications, and saves past notification records.
The S-module consists of a keyword mapping list and an information architecture. The keyword mapping provides a list of entries for the disaster-related information that supports the user’s decision-making. The information architecture structures the decision-support information based on the domain knowledge of disaster management to assist the user in searching for requested information. The question-answering process operates on the IM application in the form of a chatbot. To retrieve the requested information, the user asks the chatbot questions via the IM application by sending text messages. The system processes the user’s input text by trimming, compares the trimmed text with the keyword mapping list, finds the requested information of which the keyword matches the text from the information architecture, and replies to the user. The system also stores the user’s message text in the database.
The U-module applies user authentication and permission management. Since sensitive information may be required for governmental decision-makers, authentication and permission management are necessary to ensure information security. Permission management controls the accessibility to information in the information architecture of the S-module. Users that are not yet authenticated or have low permissions cannot retrieve sensitive data from the system via the IM application. This module also provides the authentication application form to authenticate and grant proper permissions to users. Users may fill in the application form via the in-app web browsers embedded in the IM application.
The N-module notifies users of the occurrence of disaster events and displays past notification records. When the N-module detects updates of alerts from the external data source, it retrieves the original data and determines whether the announced disaster events meet the activation standard given in the procedures. If the standard is met, the system continues to determine the related users as receivers and generate messages using predefined templates based on situations. The system sends the announcing message to the selected users via the IM application and stores the announcing record in the database. Similar to the authentication application of the U-module, users may check the notification records via the in-app web browser embedded in the IM application.

4.1. Search Module (S-Module)

The search module provides disaster-related information based on user demand and the domain knowledge of disaster management. It consists of an information architecture and a keyword mapping list. The information architecture is designed referring to the domain knowledge and the user requirements for decision support. The keyword mapping list contains entries to the content of the information architecture. By asking the chatbot via the IM application, the system processes the user’s input text, compares the text with the keyword mapping list, retrieves the information from the information architecture, and generates an appropriate response to the user.
To clarify the user demand for information retrieval and notification, we interviewed nine personnel of the Bureau of Mines, including the director general, the deputy director general, the chief secretary, the director of the Mine Safety Division, the director of Technical Assistance Division, the staff of the Mine Administration Division, the staff of the Mine Safety Division, and the staff of the Eastern District Office, as shown in Figure 2. Figure 3 shows the organization of the Bureau of Mines and describes the essential roles of some of the personnel/units during natural disasters. The director general, the deputy director general, and the chief secretary are the top decision-makers. The Mine Safety Division performs mine safety inspection, management, and promotion. The Eastern District Office and the Mine Safety Centers deliver information for notifying the mines in the affected areas to take precautions at the start and collecting reported damage from the mines at the end. Hence, the interview focused on the personnel of these roles/units.
We identified 12 demands, summarized into three major topics, as shown in Table 1. For the topic of the information architecture, users pointed out that real-time rainfall, landslide alerts, river water level alerts, road alerts, and websites of other emergency-related agencies, such as the National Fire Agency (NFA) of the Ministry of the Interior, are required. Regarding alerts, users specified the detailed activation conditions of torrential rain events and suggested that direct indication for each Mine Safety Center should be included in alert messages. For the topic of operation records, users suggested that records of operation and feedback should be collected for performance evaluation and further system improvement.

4.2. User Module (U-Module)

The user module manages user authentication and permissions. The process of decision making may require sensitive information. To ensure the security of sensitive information, we establish five types of users: decision-makers, safety staff, other staff, system maintainers, and the public. Definition and examples are designed based on different Bureau of Mines roles and are described in Table 2. Users except for the public have permission to access all sensitive information. By controlling the accessibility to the content of the information architecture, sensitive information is protected.
The system provides an application form for users to submit their profiles to gain proper permissions. Users may fill in the application form via the in-app web browsers embedded in the IM application. The system maintainers routinely collect the applications and send them to the Bureau of Mines. The decision-makers at the Bureau of Mines determine the given permission by reviewing each profile.

4.3. Notification Module (N-Module)

The target receivers of the announcement include the decision-makers and the safety staff, as defined in Table 2, since the procedures assign specific tasks to them. Hence, users who are granted the permission of decision-makers or safety staff receive notifications when the conditions are matched. Timings, conditions, and indications vary for different types of disasters. For a typhoon or a torrential rain event, the personnel are notified at both the start and the end of the warning period; the staff should notify the mines in the affected areas to take precautions at the start and report the damage at the end.
For an earthquake, the personnel are only notified after the earthquake to report the damage. Messages are delivered when any earthquake station in at least one of the jurisdictions of the Mine Safety Centers or the Eastern District Office observes an intensity greater than or equal to 5. The personnel of the Mine Safety Centers and the Eastern District Office with the largest intensity in the jurisdictions meeting the alerting criteria should inspect the mines in the jurisdiction and report the damage to the decision-makers.
For invading typhoons, messages are delivered both at the start and the end of a sea and land warning. When a warning is issued, the personnel of all the Mine Safety Centers and the Eastern District Office should notify the mines in the affected jurisdictions to take precautions. After the warning is lifted, the personnel of all the Mine Safety Centers and the Eastern District Office should inspect the mines in the affected jurisdictions and report the damage to the decision-makers.
For torrential rain events, messages are delivered both when the torrential rain warning is issued and when it is lifted in any town in at least one of the jurisdictions of the Mine Safety Centers or the Eastern District Office. The personnel of all the Mine Safety Centers and the Eastern District Office should notify the mines in the affected jurisdictions to take precautions when a warning is issued. After the warning is lifted, the personnel of all the Mine Safety Centers and the Eastern District Office should inspect the mines in the affected jurisdictions and report the damage to the decision-makers.

5. Implementation

The system is implemented with LINE, the most popular IM application in Taiwan [37]. LINE features the Messaging API to build bots for providing customized services to users. The Messaging API passes requests from the LINE Platform to bots by sending them over HyperText Transfer Protocol Secure (HTTPS) [38]. For implementation, Python 3 is used for programming with PostgreSQL as the database. The operating system (OS) is Ubuntu 18.04 deployed on an Amazon Web Service (AWS) virtual machine. Flask, a Python micro-web-framework library, was adopted to implement the system as a HyperText Transfer Protocol (HTTP) service to control the interaction process between users and the chatbot and to host additional web pages. To enable HTTPS, we retrieved certifications from Let’s Encrypt.

5.1. Keyword-Information Mapping

Based on our research of the topic, including the interviews, we implemented 108 sets of keyword information. Of the sets, 107 are classified as “weather,” “disaster management information,” “laws and documents,” and “other utilities,” and the other set is the entrance menu to access the four major classes. The classification forms the basis of the five-layer information architecture, which is simplified in Figure 4 and summarized in Table 3. Over half of the sets are classified on the topic of weather, which provides various meteorological and seismological information to support decision-makers in understanding the situation. Locations of emergency medical stations, real-time alert statuses, notification records, and the Procedures for Notification of Earthquakes, Typhoons, and Torrential Rain can be found in the topic of disaster management information. Other regulations and documents are in laws and documents. Other utilities include the authentication application form and the feedback form.

5.2. Data Source

Most of the static documentation, including the locations of mines, the contact information of mine safety centers, and the procedures and regulations are provided by the Bureau of Mines. To clarify the accurate definitions of different weather alerts, the descriptions of the definitions are collected from the CWB website. Other information is collected from the Internet, such as from related agencies’ websites. For weather charts (such as observation and forecast charts) and alerts that do not directly activate procedures (such as landslides, river water levels, and roads), we collect real-time information from web pages hosted by at least 10 agencies for the user to access via in-app web browsers.
Severe weather alerts, earthquake reports, and rainfall observations are obtained from the Open Weather Data Platform hosted by the CWB. The Open Weather Data Platform, hosted over HTTPS, provides more than 250 types of data covering various topics, including observations, weather forecasts, earthquakes, climate information, alerts and warnings, numerical weather predictions, and astronomy. Most data types, such as real-time rainfall observation records and earthquake reports, are provided following the CWB Open Data Protocol [39], which was specified by the CWB in 2015. The others are provided following the Common Alerting Protocol (CAP) [40] specified by the Organization for the Advancement of Structured Information Standards (OASIS) in 2010, such as torrential rain warnings and typhoon alerts. Both the CWB Open Data Protocol and the CAP adapt Extensible Markup Language (XML) and specify a unique identifier for every single report or warning message. For system implementation, we developed a timer to check if there are updates from the Open Weather Data Platform. We also developed parsers for earthquake reports, torrential rain warnings, and typhoon alerts to extract essential information. The database records the extracted identifiers of such information to identify whether the data is updated.

5.3. Immediate Notification

Following the timer for checking alert updates every 2 min, we implemented the procedures and the activation conditions via programming. When the information extracted from retrieved data meets the conditions, the system generates appropriate text messages based on the templates given by the Bureau of Mines and sends the messages to the decision-makers and the safety staff defined in Table 2 as a notification. Figure 5 shows a notification example in the chat for a torrential rain event. The text message states the alert type, announcement time, affected areas, the Mine Safety Center with impacted jurisdictions, and the relevant procedures. In addition, a hyperlink to the website of the CWB is attached to the message for the user to retrieve detailed information. The system also records the messages sent to users and displays the messages via web pages, as shown in Figure 6. Even though users may find some notification records in their IM application, recent users cannot retrieve past notifications prior to the time they befriend the chatbot. Thus, an additional panel for looking up past notifications is required.

5.4. User Interface

For the user interface, two interactive modes, “by-text” and “by-click,” are provided. Using the by-text mode, users interact with the chatbot via a process mimicking conversation with a real person. This starts with a user acquiring specific information from the chatbot by sending a text message. The chatbot receives the text message, processes the input, and looks up the keyword-information mapping for the most probable answer. Then, the chatbot responds to the user with the information fetched from the information architecture. Users can also interact with the chatbot by clickable images or buttons in the by-click mode. Clickable images or buttons send text messages on behalf of the user, triggering the chatbot to respond like in the by-text mode or opening the in-app web browser to an external website. Users familiar with the system can directly acquire the information using the by-text mode. Users unfamiliar with the system can retrieve the information using the by-click mode. Both modes provide access to the keyword-information mapping and the information architecture.
Figure 7 shows a series of screenshots of the LINE chat on Android 9. The user clicks one of the buttons in the menu at the bottom of the screen. The chatbot sends a text message on behalf of the user. Next, the chatbot retrieves the requested information from the information architecture, generates an appropriate response, and replies to the user. The information may be a specific description, a link to an external website, or a topic of some information in the architecture. For a specific description, the chatbot replies in the form of a text message. For links, the chatbot replies in the form of buttons. For a topic, the chatbot replies with a clickable image composed of the entries covered in this topic as constructed in the information architecture.

6. Effectiveness Evaluation

We evaluated the effectiveness of the implemented system by the change in the efficiency of the procedures before and after adopting the system. We collected the operation records from the Bureau of Mines before and after the system was adopted in the field. For the three different types of disaster specified in the procedures, 31 notification records were collected, covering 21 events from 9 April 2019 to 1 November 2019.

6.1. Processes

Figure 8 illustrates the processes before and after the adoption of the system. The processes start from the observation or the forecast of a disaster event. The CWB provides disaster alerts in multiple ways, such as e-mails, SMS, updates of the Open Weather Data Platform, and announcements on the official website. In the original process, the staff of the Mine Safety Division received the notification and determined whether the Procedures for Notification of Earthquakes, Typhoons, and Torrential Rain should be activated. If the procedures were activated, the staff notified the personnel of the Mine Safety Centers and the Eastern District Office in the affected area(s). For the adapted process, the implemented system retrieves the alert information from the CWB, automatically determines the activation and the affected areas based on the retrieved information, and directly notifies the personnel of both the Mine Safety Division and the Mine Safety Centers or the Eastern District Office with impacted areas. Finally, in both processes, the personnel of the Mine Safety Centers and the Eastern District Office notifies the mines located in the area(s).

6.2. Cases

Three timings are recorded in a disaster event: the occurrence, the notification sent from the system, and the notification manually sent by the staff of the Bureau of Mines. To be precise, the occurrence of an earthquake is defined as the timing of the observation, while the occurrence of a typhoon event or a torrential rain event is defined as the timing of issuing or lifting the warning. For each announcement, we recorded the three timings and found the time span from the occurrence to the message delivery using the different processes. For the original process, we defined the time span of an announcement by subtracting the timing of the occurrence of the disaster event from the timing of the manual delivery of alerts to the staff of the Bureau of Mines. For the adapted process, we define the time span of an announcement by subtracting the timing of the occurrence of the event from the timing of message delivery to the Bureau of Mines personnel by the system.
To evaluate the effectiveness of adopting the system, we compared the time spans for the two processes and calculated the improvement in time cost before and after the adoption. Furthermore, considering the eventual lag for events at night, we marked the announcements published in the common rest period (i.e., from 00:00 to 07:00).
Ten earthquake events were selected for evaluation, with six occurring in the normal period and four in the rest period. Table 4 lists the three timings of the selected records. The adapted process achieved improved time in nine of the ten events, including all four in the rest period and five of the six in the normal period.
Four typhoon events were selected for evaluation, including four warnings each being issued and lifted. One of the warnings was issued in the normal period and was lifted in the rest period, while the other warnings were all issued and lifted in the normal period. Table 5 lists three timings of the selected records. Some warnings were issued or lifted prior to the occurrence because the Central Weather Bureau sometimes published the announcements online earlier than the actual occurrence of an event, making the time spans negative.
Seven torrential rain events were selected for evaluation, including seven warnings being issued and six being lifted; the lifting of one of the warnings was missing. One warning was issued, one was lifted during the rest period, and the others occurred in the normal period. Table 6 lists the three timings of the selected records.

6.3. Results

The average decrease in time cost after system adoption is evident for both the inclusion and exclusion of the rest period cases. Table 7 shows the average improvement in different types of disaster events. If the cases occurring in the rest period are included, an average reduction of 72.6 min of time cost was achieved among the three different types of disaster events. If the cases in the rest period were excluded, the average reduction was 55.8 min.
Note that the comparison of including/excluding the rest period cases also shows the system’s effectiveness in reducing labor costs. By excluding the rest period cases, the time cost of the original process decreases from 81.7 to 65.2 min, which is expected since the staff cannot immediately respond during the rest period. By contrast, the average time cost of the adapted process is less than 10 min, both including and excluding the rest period cases, indicating that the rest period does not affect the efficiency of the adapted process. If the Bureau of Mines wanted to eliminate inefficiency during the rest period, they would previously require continuous shift work. This is solved by the adoption of the system, thus reducing labor costs.
Among all announcement types, the torrential rain event lift and the typhoon event lift gain the most improvements. In contrast to the other announcement types, excluding the rest period cases does not visibly lower the time cost of the original process, according to Table 7. Thus, the improvements obtained in the torrential rain event lift and the typhoon event lift are not related to the rest period. On the other hand, the earthquake and torrential rain event gained fewer improvements in the adapted process. The difference between the improvements may come from higher awareness of the disaster occurrence or warning issue than the warning lift. The staff of the Bureau of Mines notice the announcement quicker after a disaster event occurs than when it is over. Hence, the adapted system saves relatively less but still plenty of time for the occurrence of disaster events, while the effectiveness of the system is significantly demonstrated for the lift of disaster event alerts.

7. Discussion

The major advantages of this study include the feasible utilization of novel techniques for decision support and the improvement of efficiency and effectiveness of procedure activation during disaster events. The study provides a solution by adopting a chatbot as an information retrieval interface and a channel to alert relevant personnel. Unlike other disaster-related products designed for the general population as we mentioned in Section 2, the chatbot in the present study specifically assists the decision-makers and other stakeholders in efficiently retrieving decision-support information via IM applications. In addition, the system consistently receives immediate announcements from the CWB and provides automatic determination based on the activating conditions, reducing labor and time costs for the safety staff of the Bureau of Mines. By enhanced retrieval and reduced consumption, our system improves the efficiency and effectiveness of decision-making.
However, the labor and time consumption of damage reporting is not completely eliminated in the present study. Since we focus on the activation of information delivery between the different personnel, case closure was not considered in the study. Currently, the system automatically notifies staff to collect and report the damage in each jurisdiction area. However, during the research interviews, we noticed that staff had to manually summarize the collected damages. The fundamental problem is that the current damage collection and report process is still based on phone or SMS; that is, there is no integrated system to collect the damage reports, and thus no chance for the developed chatbot in this study to automatize the summarization. Hence, we are unable to comprehensively reduce the labor and time cost of the procedures. To overcome the limitation, a system for damage collection and report process is required. In addition, techniques for unstructured report processing [28,41,42] may be adopted to enable automatic analysis and further reduce the labor and time cost.

8. Conclusions

We developed a chatbot to enhance the efficiency of government activation of mine safety procedures during natural disasters. Since natural disasters in Taiwan, including earthquakes, typhoons, and heavy rain, are frequent and impactful, a comprehensive government system is dedicated to disaster response. The Bureau of Mines has instituted clear procedures to ensure the delivery of disaster alerts and damage reports. However, there is an opportunity for improvement due to inefficiency associated with labor- and time-consuming procedures. In this study, we proposed a system framework for disaster-related information retrieval and immediate notifications to support the execution of mine safety procedures. The framework utilizes IM applications as the user interface to look up information and send messages to announce the occurrence of disaster events. We implemented the system in the form of a chatbot. The evaluation of the change in efficiency before and after system adoption demonstrated an average reduction of 55.8 min among the three types of disaster events. Although the lack of integrated systems to collect damage reports limits the overall reduction of the labor and time cost, the present study has proven the feasibility of adopting novel techniques for decision-making and assures the improvement of the efficiency and effectiveness of the procedure activation.

Author Contributions

Conceptualization, M.-H.T.; methodology, M.-H.T., H.-Y.C., and Y.-L.C.; software, H.-Y.C. and Y.-L.C.; validation, M.-H.T., H.-Y.C., Y.-L.C., and H.-K.S.; formal analysis, M.-H.T., H.-Y.C., Y.-L.C., and H.-K.S.; investigation, M.-H.T., H.-Y.C., Y.-L.C., and H.-K.S.; resources, H.-K.S. and P.-Y.L.; data curation, M.-H.T., H.-Y.C., and Y.-L.C.; writing–original draft preparation, H.-Y.C. and Y.-L.C.; writing—review and editing, M.-H.T., H.-Y.C., Y.-L.C., H.-K.S., P.-Y.L., and C.-W.H.; visualization, H.-Y.C., Y.-L.C., and H.-K.S.; supervision, M.-H.T.; project administration, M.-H.T.; funding acquisition, M.-H.T. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Taiwan’s Ministry of Science and Technology (MOST) under contract 108-2119-M-011-002 and 109-2124-M-002-005.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Acknowledgments

We are grateful to the Bureau of Mines for the generous participation and feedback in this research.

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Abbreviations

The following abbreviations are used in this manuscript:
AWSAmazon Web Service
CAPCommon Alerting Protocol
CWBCentral Weather Bureau
DSSdecision support systems
GISgeographic information systems
HTTPHyperText Transfer Protocol
HTTPSHyperText Transfer Protocol Secure
ICTinformation and communication technology
IMinstant messaging
MOEAMinistry of Economic Affairs
N-modulenotification module
NCDRNational Science and Technology Center for Disaster Reduction
NFANational Fire Agency
NICTNational Institute of Information and Communications Technology
NIEDNational Research Institute for Earth Science and Disaster Resilience
OASISOrganization for the Advancement of Structured Information Standards
OSoperating system
S-modulesearch module
SMSshort message service
U-moduleuser module
XMLExtensible Markup Language

References

  1. Chang, J.C. Natural hazards in Taiwan. GeoJournal 1996, 38, 251–257. [Google Scholar] [CrossRef]
  2. Central Weather Bureau. FAQ for Earthquake; Central Weather Bureau: Taipei, Taiwan, 2017. (In Chinese)
  3. Central Weather Bureau. FAQ for Typhoon; Central Weather Bureau: Taipei, Taiwan, 2020. (In Chinese)
  4. Ministry of Health and Welfare. Statistics of Causes of Death in 1999 [Data File]. Available online: https://dep.mohw.gov.tw/DOS/lp-1836-113-xCat-9.html (accessed on 21 October 2020). (In Chinese)
  5. National Fire Agency, Ministry of Interior. Statistics of Natural Disaster in Taiwan from 1958 to 2019 [Data File]. Available online: https://www.nfa.gov.tw/cht/index.php?code=list&ids=233 (accessed on 21 October 2020). (In Chinese)
  6. National Fire Agency, Ministry of Interior. Statistics of August 23 Flood of Tropical Depression [Data File]. Available online: https://www.emic.gov.tw/14/index.php?code=list&ids=651&detail=94 (accessed on 21 October 2020). (In Chinese)
  7. Disaster Prevention and Protection Act. Available online: https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=D0120014 (accessed on 20 October 2020). (In Chinese)
  8. Suarez-Tangil, G.; Tapiador, J.E.; Peris-Lopez, P.; Ribagorda, A. Evolution, Detection and Analysis of Malware for Smart Devices. IEEE Commun. Surv. Tutor. 2014, 16, 961–987. [Google Scholar] [CrossRef][Green Version]
  9. Mallinson, K. Smartphone Revolution: Technology patenting and licensing fosters innovation, market entry, and exceptional growth. IEEE Consum. Electron. Mag. 2015, 4, 60–66. [Google Scholar] [CrossRef]
  10. Jurafsky, D.; Martin, J.H. Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition, 2nd ed.; Prentice Hall PTR: Upper Saddle River, NJ, USA, 2007. [Google Scholar]
  11. Bavaresco, R.; Silveira, D.; Reis, E.; Barbosa, J.; Righi, R.; Costa, C.; Antunes, R.; Gomes, M.; Gatti, C.; Vanzin, M.; et al. Conversational agents in business: A systematic literature review and future research directions. Comput. Sci. Rev. 2020, 36, 100239. [Google Scholar] [CrossRef]
  12. Sun, Y.; Zhang, Y.; Chen, Y.; Jin, R. Conversational Recommendation System with Unsupervised Learning. In Proceedings of the 10th ACM Conference on Recommender Systems, Boston, MA, USA, 15–19 September 2016; Association for Computing Machinery: New York, NY, USA, 2016; pp. 397–398. [Google Scholar] [CrossRef][Green Version]
  13. Jusoh, S. Intelligent Conversational Agent for Online Sales. In Proceedings of the 2018 10th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), Iasi, Romania, 28–30 June 2018; pp. 1–4. [Google Scholar] [CrossRef]
  14. Majumder, A.; Pande, A.; Vonteru, K.; Gangwar, A.; Maji, S.; Bhatia, P.; Goyal, P. Automated Assistance in E-commerce: An Approach Based on Category-Sensitive Retrieval. In Advances in Information Retrieval; Springer: Cham, Switwerland, 2018; pp. 604–610. [Google Scholar] [CrossRef]
  15. Koetter, F.; Blohm, M.; Kochanowski, M.; Goetzer, J.; Graziotin, D.; Wagner, S. Motivations, Classification and Model Trial of Conversational Agents for Insurance Companies. In Proceedings of the 11th International Conference on Agents and Artificial Intelligence—Volume 1: ICAART, INSTICC, Prague, Czech Republic, 19–21 February 2019; SciTePress: Prauge, Czech Republic, 2019; pp. 19–30. [Google Scholar] [CrossRef]
  16. Cui, L.; Huang, S.; Wei, F.; Tan, C.; Duan, C.; Zhou, M. SuperAgent: A Customer Service Chatbot for E-commerce Websites. In Proceedings of the ACL 2017, System Demonstrations, Vancouver, BC, Canada, 31 July–2 August 2017; Association for Computational Linguistics: Vancouver, BC, Canada, 2017; pp. 97–102. [Google Scholar]
  17. Laranjo, L.; Dunn, A.G.; Tong, H.L.; Kocaballi, A.B.; Chen, J.; Bashir, R.; Surian, D.; Gallego, B.; Magrabi, F.; Lau, A.Y.S.; et al. Conversational agents in healthcare: A systematic review. J. Am. Med. Inform. Assoc. 2018, 25, 1248–1258. [Google Scholar] [CrossRef] [PubMed][Green Version]
  18. Vaidyam, A.N.; Wisniewski, H.; Halamka, J.D.; Kashavan, M.S.; Torous, J.B. Chatbots and Conversational Agents in Mental Health: A Review of the Psychiatric Landscape. Can. J. Psychiatry 2019, 64, 456–464. [Google Scholar] [CrossRef] [PubMed]
  19. Mavropoulos, T.; Meditskos, G.; Symeonidis, S.; Kamateri, E.; Rousi, M.; Tzimikas, D.; Papageorgiou, L.; Eleftheriadis, C.; Adamopoulos, G.; Vrochidis, S.; et al. A Context-Aware Conversational Agent in the Rehabilitation Domain. Future Internet 2019, 11, 231. [Google Scholar] [CrossRef][Green Version]
  20. Griol, D.; Molina, J.M.; Sanchis, A. A multimodal conversational coach for active ageing based on sentient computing and m-health. Expert Syst. 2020, 37, e12454. [Google Scholar] [CrossRef]
  21. Tsai, M.H.; Chen, J.Y.; Kang, S.C. Ask Diana: A Keyword-Based Chatbot System for Water-Related Disaster Management. Water 2019, 11, 234. [Google Scholar] [CrossRef][Green Version]
  22. TechNews. NCDR LINE Subscribers Exceeded One Million 2019. Available online: https://technews.tw/2019/02/18/ncdr-line-subscribers-exceeded-one-million/ (accessed on 12 August 2020). (In Chinese).
  23. LINE Corporation. LINE Timeline of the National Science and Technology Center for Disaster Reduction. Available online: https://timeline.line.me/user/_dUYw2BhYBym-OpC2gsao6pnoUxK0UmI2obWVFYM (accessed on 21 October 2020). (In Chinese).
  24. Agriculture Department, New Taipei City Government. The Landslide Guardian 2.0 of New Taipei City Alarms Disasters and Provides Evacuation Instructions via a LINE Chatbot 2019. Available online: https://www.ntpc.gov.tw/ch/home.jsp?id=28&dataserno=201901180008 (accessed on 30 December 2019). (In Chinese)
  25. Research Center for National Disaster Resilience, National Research Institute for Earth Science and Disaster Resilience. An Experiment of SOCDA, a Disaster Prevention Chatbot, was Carried out in Itami City Flood Control Map Training 2019. Available online: https://www.bosai.go.jp/nr/info/info_detail_01.html (accessed on 22 October 2020). (In Japanese)
  26. LINE Corporation. Disaster Prevention and Reduction 2020. Available online: https://linecorp.com/ja/csr/activity/disasterprevention (accessed on 22 October 2020). (In Japanese).
  27. Tsai, M.H.; Chan, H.Y.; Liu, L.Y. Conversation-Based School Building Inspection Support System. Appl. Sci. 2020, 10, 3739. [Google Scholar] [CrossRef]
  28. Kung, H.K.; Hsieh, C.M.; Ho, C.Y.; Tsai, Y.C.; Chan, H.Y.; Tsai, M.H. Data-Augmented Hybrid Named Entity Recognition for Disaster Management by Transfer Learning. Appl. Sci. 2020, 10, 4234. [Google Scholar] [CrossRef]
  29. Chan, H.Y.; Tsai, M.H. Question-answering dialogue system for emergency operations. Int. J. Disaster Risk Reduct. 2019, 41, 101313. [Google Scholar] [CrossRef]
  30. Wallace, W.A.; De Balogh, F. Decision Support Systems for Disaster Management. Public Adm. Rev. 1985, 45, 134–146. [Google Scholar] [CrossRef]
  31. Fogli, D.; Guida, G. Knowledge-centered design of decision support systems for emergency management. Decis. Support Syst. 2013, 55, 336–347. [Google Scholar] [CrossRef]
  32. Yoon, S.; Velasquez, J.; Partridge, B.; Nof, S. Transportation security decision support system for emergency response: A training prototype. Decis. Support Syst. 2008, 46, 139–148. [Google Scholar] [CrossRef]
  33. Zhou, L.; Wu, X.; Xu, Z.; Fujita, H. Emergency decision making for natural disasters: An overview. Int. J. Disaster Risk Reduct. 2018, 27, 567–576. [Google Scholar] [CrossRef]
  34. Water Resources Agency Disaster Emergency Response System. Available online: http://fhy.wra.gov.tw/dmchyv2/ (accessed on 15 December 2020). (In Chinese)
  35. Emergency Management Information Cloud (EMIC). Available online: https://portal2.emic.gov.tw/ (accessed on 16 December 2020). (In Chinese)
  36. Bureau of Mines, Ministry of Economic Affairs. Organization History—Bureau of Mines, MOEA. Available online: https://www.mine.gov.tw/English/intro/org.asp (accessed on 10 December 2020).
  37. Taiwan Network Information Center. Wireless Internet Usage in Taiwan, Summary Report of November 2017 Survey; Taiwan Network Information Center: Taipei, Taiwan, 2017; Available online: https://www.twnic.tw/doc/twrp/201711d.pdf (accessed on 10 April 2020).
  38. LINE Corporation. Messaging API Overview | LINE Developers. Available online: https://developers.line.biz/en/docs/messaging-api/overview/ (accessed on 10 December 2020).
  39. Central Weather Bureau. Development Manual of the CWB Open Weather Data Platform. Available online: https://opendata.cwb.gov.tw/devManual/insrtuction (accessed on 20 October 2020). (In Chinese)
  40. Organization for the Advancement of Structured Information Standards. Common Alerting Protocol Version 1.2. Available online: http://docs.oasis-open.org/emergency/cap/v1.2/CAP-v1.2-os.html (accessed on 20 October 2020).
  41. Lin, L.H.; Miles, S.B.; Smith, N.A. Natural Language Processing for Analyzing Disaster Recovery Trends Expressed in Large Text Corpora. In Proceedings of the 2018 IEEE Global Humanitarian Technology Conference (GHTC), San Jose, CA, USA, 18–21 October 2018; IEEE: San Jose, CA, USA, 2018; pp. 1–8. [Google Scholar] [CrossRef]
  42. Kreimeyer, K.; Foster, M.; Pandey, A.; Arya, N.; Halford, G.; Jones, S.F.; Forshee, R.; Walderhaug, M.; Botsis, T. Natural language processing systems for capturing and standardizing unstructured clinical information: A systematic review. J. Biomed. Inform. 2017, 73, 14–29. [Google Scholar] [CrossRef]
Figure 1. The system framework.
Figure 1. The system framework.
Sustainability 13 00654 g001
Figure 2. Pictures of the interviews.
Figure 2. Pictures of the interviews.
Sustainability 13 00654 g002
Figure 3. The organization chart of the Bureau of Mines (adapted from Bureau of Mines, Ministry of Economic Affairs [36]).
Figure 3. The organization chart of the Bureau of Mines (adapted from Bureau of Mines, Ministry of Economic Affairs [36]).
Sustainability 13 00654 g003
Figure 4. Top three layers of the information architecture.
Figure 4. Top three layers of the information architecture.
Sustainability 13 00654 g004
Figure 5. An example of a torrential rain announcement in the chat.
Figure 5. An example of a torrential rain announcement in the chat.
Sustainability 13 00654 g005
Figure 6. The past notification look-up web page viewed with the in-app browser.
Figure 6. The past notification look-up web page viewed with the in-app browser.
Sustainability 13 00654 g006
Figure 7. Screenshots of the user interface of the LINE chat on Android 9.
Figure 7. Screenshots of the user interface of the LINE chat on Android 9.
Sustainability 13 00654 g007
Figure 8. Comparison of the original process and the adapted process.
Figure 8. Comparison of the original process and the adapted process.
Sustainability 13 00654 g008
Table 1. The demands identified in interviews with the Bureau of Mines personnel.
Table 1. The demands identified in interviews with the Bureau of Mines personnel.
TopicDemand
Information architectureReal-time rainfall
Landslide alerts
River water level alerts
Road alerts
Websites of other agencies
Architecture rearrangement based on disaster management phases
Consideration of sensitive information
AlertsSpecification of torrential rain announcement timings
Direct indication for different Mine Safety Centers
Operation recordsOperation timeline for further analysis
User feedback collection for system improvement
User log collection for further analysis
Table 2. The description of user permissions.
Table 2. The description of user permissions.
TypeDescription
Decision-makersCore decision-makers of the Bureau of Mines, e.g., the director general, the deputy director general, the chief secretary, and directors of the divisions
Safety staffPersonnel providing immediate response to the emergency, e.g., the staff of the Mine Safety Division, the staff of the Mine Safety Centers, and the staff of the Eastern District Office
Other staffPersonnel that do not need to immediately respond to emergency
System maintainersMaintainers of the system
PublicUsers not yet authenticated or not working for the Bureau of Mines
Table 3. Summary of information in the four major classes.
Table 3. Summary of information in the four major classes.
TopicNumber of KeywordsExamples
Weather57Real-time rainfall, weather charts from different institutions, radar charts, satellite images, typhoon information, earthquake information, definitions of weather alerts
Disaster management information28Locations of emergency medical stations, the Procedures for Notification of Earthquakes, Typhoons, and Torrential Rain, real-time alert statuses, notification records, links to websites of the other agencies
Laws and documents19The Disaster Prevention and Protection Act, the Operation Plan of Disaster Prevention and Protection for Mining, the Manual of Disaster Prevention and Protection Operation for Mining
Other utilities3Authentication, feedback
Table 4. Time evaluation for earthquake cases.
Table 4. Time evaluation for earthquake cases.
OccurrenceIn Rest PeriodOriginal ProcessAdapted ProcessImprovement (minutes)
Notification TimeTime Span (minutes)Notification TimeTime Span (minutes)
9 April 23:13no9 April 23:39269 April 23:22917
10 April 04:24yes10 April 06:3613210 April 04:3410122
15 April 23:26no15 April 23:532715 April 23:34819
18 April 13:01no18 April 13:111018 April 13:1413−3
17 May 00:26yes17 May 01:124617 May 00:31541
23 May 14:12no23 May 14:221023 May 14:1973
4 June 17:46no4 June 18:05194 June 17:51514
6 August 09:19no6 August 09:35166 August 09:29106
8 August 05:28yes8 August 05:40128 August 05:4214−2
26 August 04:41yes26 August 06:5313226 August 04:498124
Table 5. Time evaluation for typhoon cases.
Table 5. Time evaluation for typhoon cases.
TypeOccurrenceIn Rest PeriodOriginal ProcessAdapted ProcessImprovement (minutes)
Notification TimeTime Span (minutes)Notification TimeTime Span (minutes)
issue17 July 11:30no17 July 11:34417 July 11:24−610
lift17 July 20:30no17 July 21:558517 July 20:24−691
issue8 August 08:30no8 August 08:53238 August 08:47176
lift9 August 20:30no9 August 20:3669 August 20:3333
issue23 August 14:30no23 August 14:31123 August 14:3110
lift25 August 08:30no25 August 08:481825 August 08:25−523
issue29 September 20:30no29 September 20:441429 September 20:43131
lift1 October 05:30yes1 October 06:09391 October 05:26−443
Table 6. Time evaluation for torrential rain cases.
Table 6. Time evaluation for torrential rain cases.
TypeOccurrenceIn Rest PeriodOriginal ProcessAdapted ProcessImprovement (minutes)
Notification TimeTime Span (minutes)Notification TimeTime Span (minutes)
issue27 April 15:50no27 April 16:061627 April 16:1121−5
lift28 April 04:20yes28 April 07:3219228 April 04:3616176
issue17 May 09:40no17 May 09:511117 May 09:5414−3
lift17 May 15:10no17 May 19:0223217 May 15:199223
issue18 May 12:55no18 May 13:485318 May 13:091439
lift18 May 22:25no19 May 06:3849318 May 22:4520473
issue11 June 15:10no11 June 15:322211 June 15:26166
lift11 June 18:50no11 June 19:192911 June 19:11218
issue21 September 01:10yes21 September 08:0741721 September 01:2010407
lift21 September 11:15no21 September 17:3437921 September 11:227372
issue26 September 23:10no26 September 23:302026 September 23:22128
lift27 September 12:40no(missing)(missing)27 September 12:5010(missing)
issue31 October 15:00no31 October 15:191931 October 15:12127
lift1 November 07:30no1 November 08:01311 November 07:401021
Table 7. Average improvements in different types of disaster events.
Table 7. Average improvements in different types of disaster events.
Announce TypeRest Period IncludedRest Period Excluded
Number of CasesOriginal (minutes)Adapted (minutes)Improvement (minutes)Number of CasesOriginal (minutes)Adapted (minutes)Improvement (minutes)
Earthquake1043.08.934.1618.08.79.3
Typhoon event issue410.56.34.3410.56.34.3
Typhoon event lift437.0−3.040.0336.3−2.739.0
Torrential rain event issue779.714.165.6623.514.88.7
Torrential rain event lift6226.013.8212.25232.813.4219.4
Overall average3181.79.272.62565.29.455.8
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Tsai, M.-H.; Chan, H.-Y.; Chan, Y.-L.; Shen, H.-K.; Lin, P.-Y.; Hsu, C.-W. A Chatbot System to Support Mine Safety Procedures during Natural Disasters. Sustainability 2021, 13, 654. https://doi.org/10.3390/su13020654

AMA Style

Tsai M-H, Chan H-Y, Chan Y-L, Shen H-K, Lin P-Y, Hsu C-W. A Chatbot System to Support Mine Safety Procedures during Natural Disasters. Sustainability. 2021; 13(2):654. https://doi.org/10.3390/su13020654

Chicago/Turabian Style

Tsai, Meng-Han, Hao-Yung Chan, Yi-Lin Chan, Heng-Kuang Shen, Pei-Yi Lin, and Ching-Wen Hsu. 2021. "A Chatbot System to Support Mine Safety Procedures during Natural Disasters" Sustainability 13, no. 2: 654. https://doi.org/10.3390/su13020654

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop