Identification of Aflatoxin Knowledge Areas for Developing Natural Language Processing Chatbot Datasets

Natural Language Processing chatbots are gaining popularity in today’s world. This is because of the enabled intelligent human-to-computer conversation in the form of text and speech. These abilities have been contributed by the presence of well-developed and accurate datasets built using well-structured procedures. Unfortunately, domain-specific datasets are limited. This brought the need for this study as it focused on identifying aflatoxin knowledge areas to be used to develop aflatoxin knowledge datasets. These datasets will then be used to develop an aflatoxin knowledge-sharing question-and-answer chatbot. The chatbot will enable aflatoxin knowledge-sharing for awareness creation. The questions and answers grouped as knowledge areas for building the dataset were collected through a data collection process. This was a user-centered approach where focus group discussions were conducted to capture aflatoxin-related questions from expected users. These were the maize and groundnut stakeholders including farmers, traders, consumers, and extension officers. On the other hand, responses to the collected questions were provided by researchers and experts through conducted interview sessions. In the data collection process, 7 focus group discussions and 9 interview sessions were conducted in seven regions in Tanzania named; Arusha, Dodoma, Iringa, Morogoro, Tabora, Mtwara, and Dar-es-salaam. The results of the study presented 6 themes (knowledge areas) with respective sub-themes. These themes are: 1) Meaning of Aflatoxin 2) Effects of Aflatoxin 3) Management of Contaminated crops 4) Laws and Regulations 5) Prevention and 6) Aflatoxin awareness creation. These will then be used to develop English and Swahili datasets for developing an aflatoxin knowledge-sharing chatbot.