Exploring AI chatbot technology as an information support system for water resources management

The advent of Artificial Intelligence (AI) brought with it improved efficiency in the execution of various computational tasks. One area of application of this technology is in the development of conversational applications otherwise known as chatbots. The chatbot technology employs the Natural Language Processing (NLP) and Machine Learning (ML) components of AI to develop models that enable computers to better understand natural human languages. This human-computer interaction has allowed for easier and faster access to information. This research, therefore, employs this chatbot technology to develop a prototype conversational application that extracts flooding information from EPA SWMM and relays this information to citizens in flood-prone areas. SWMM simulations for Do Lo, an area in Hanoi, Vietnam, that is severely impacted by floods annually were used as the study case. A database of intents with different terminologies that users may use in conversation with the chatbot was created. These terminologies then served as training phrases for the NLP and ML algorithms in the Google Dialogflow platform which was used in developing the chatbot. Then the Pyswmm library was used in developing Python functions to extract the data from the SWMM model. For integrating the Dialogflow chatbot with the SWMM model the Flask web framework hosted on the Google App Engine was used. Also, to ensure that computations were accurately done for each user based on their input location, a database was created to store the user’s session ID and location. For the user interface a webpage was designed and integrated into the application using Flask. The final developed chatbot prototype was capable of responding to the user’s questions on the flooding situation at their location. To test the performance of the chatbot, a pilot test was done where a majority of the respondents reported that the application provided accurate results and they also found the application easy to use and worth recommending to others. In conclusion, a prototype chatbot application was developed that could respond to the user’s questions providing, otherwise inaccessible flooding information obtained from SWMM. If deployed, the chatbot would assist in flood awareness creation by allowing the users to make informed decisions in case of flooding. Finally, the study recommended among other issues that further research could incorporate rainfall forecast data in the SWMM simulation to provide citizens with real-time information for timely response in times of flooding