1357 AI-Powered Chatbot For Effective Weather Communication

Wednesday, 15 January 2020
Hall B (Boston Convention and Exhibition Center)
Saiadithya Cumbulam Thangaraj, Earth Networks, Germantown, MD; and M. Stock and J. Lapierre

Weather forecasting is a complex and often challenging skill that involves observing and processing vast amounts of data. Even though there are Weather systems built to generate automated Weather warnings, they still suffer from many User communication problems. Effective communication of weather information is important both for experts who disseminate crucial, life-saving information about threats and for the public, who need to receive and interpret in ways that allow them to act.

Among the communication problems are the difficulties to answer weather domain related queries to the public, delivering abstruse weather forecast, report delayed weather news, unavailability of 24x7 customer service during severe weather situations, and not effectively communicating the operational and technical information of the weather system used by the public.

To resolve these problems, we aim to design a chatbot that is specifically tailored for handling Weather queries from the public. The Chatbot that we proposed, is designed to 1. Provide effective Weather warnings 2. Educate and provide information on the current Weather, 3. Serve as a 24x7 customer care 4. Provide personalized weather forecasting to help User better plan their day.

The methodology that we followed to build the Chatbot makes use of Natural Language Processing (NLP) to extract useful information from the public queries and provide a relevant response using Machine Learning-based approach. The Machine learning-based approach allows the chatbot to use a repository of predefined responses and heuristics to pick an appropriate response to the public. The heuristic could vary as simple as a rule-based expression match, or as complex as a Machine Learning classification.

For building the Chatbot’s Weather domain knowledge, we would be using techniques for extracting Structured knowledge from Email messages of a Weather organization. The knowledge extraction will be an on-going process and the extracted weather knowledge will be useful for the public for better understanding the current weather.

The test results of the built chatbot show that it serves to be extremely useful for tool for communicating weather information to the public. Moreover, the Chatbot developed can be used as a guideline for future improvement and study.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner