Transforming Smallholder Farmers Support with an AI-Powered FAQbot: A Comparison of Techniques

Mojolaoluwa Jadesola Adedokun, Saad Salami, David Chinweuba Onyeali, Bolaji Olayemi Toheeb, Damilare Adeyoyin, Kingsley Afuzobugwu

Primary Area: generative models

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Agriculture, FAQBot, LLMs, Natural Language Processing

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: A comparative study on faqbot approaches for supporting small holder farmers

Abstract:

Access to sufficient information on desired agricultural practices, such as planting period, when to apply fertiliser, how to transport grains, etc. is of utmost importance in the agricultural industry as it directly affects farm yields. The responses to these questions are closed domain, therefore leading to the development of a question-answering conversational bot (FAQbot) that can provide the appropriate responses immediately. This study undertakes a comparative analysis of three distinct methodologies for constructing a FAQbot. These approaches encompass the development of a generative-based chatbot employing BERT and GPT-2, the creation of an intent classification model leveraging PyTorch and the Natural Language Toolkit (NLTK) libraries, and the implementation of an information retrieval-based model utilising pre-trained Large Language Models (LLMs) using Langchain. Our methodological framework includes the transformation of a FAQ dataset into formats suitable for chatbot training, specifically CSV and JSON. Notably, the retrieval-based method surpassed the generative-based and intent classification methods by consistently providing precise answers for every question in the database, irrespective of rephrasing or reframing.

Keywords: Agriculture, FAQBot, LLMs, Natural Language Processing

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8425