A Response-by-Retrieval Chatbot for Enhancing Horticulture Extension Services in Tanzania
Bibliographic record
Abstract
Horticulture, which encompasses the cultivation of flowers, fruits, herbs, and vegetables, is a key contributor to Tanzania’s export revenue generation. Smallholder farmers are the primary producers of these crops, and they rely heavily on extension services for critical information that shapes both their economic success and long-term sustainability. However, the delivery of such services from the government and other stakeholders faces challenges, including constraints in human capital, geographic barriers, misaligned information needs, as well as issues with the timeliness of information dissemination. To address these challenges, this study developed a Swahili-language chatbot designed to provide timely, context-specific information tailored to the needs of farmers. To ensure credibility and relevance, key private and public stakeholders were consulted, and comprehensive farming guides were collected to build a custom dataset. This dataset consisted of 307 passages and 2,231 question-answer pairs. Four multilingual models, Multilingual Bidirectional Encoder Representations from Transformers (mBERT), Cross-lingual Language Model Pretraining RoBERTa (XLM-R), Multilingual Decoding-Enhanced BERT with Disentangled Attention (mDeBERTa), and Afro Cross-lingual Language Model Pretraining RoBERTa (AfroXLMR), were finetuned on this dataset for a question-answering task. Among them, the mDeBERTa model achieved the strongest performance, with an Exact Match (EM) score of 62.69% and an F1 score of 75.35%. These results demonstrate the potential of adapting advanced language models for specialized, low-resource language tasks in agriculture. The deployment of mDeBERTa in a prototype chatbot highlights a promising pathway to bridge information gaps and enhance the accessibility of extension services for Tanzania’s smallholder farmers.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.007 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".