ISHEERO/data_merge_2

Card of the augmented and modified dataset for the LLM challenge on public health systems in Malawi

Dataset Description

This dataset was created within the scope of the Malawi Public Health Systems LLM Challenge organized by the AI Lab of the Malawi University of Business and Applied Sciences. The challenge aimed to develop an AI assistant capable of extracting information from the Malawi Technical Guidelines for Integrated Disease Surveillance and Response (TGs for IDSR). The dataset was derived from the provided "Train.csv" dataset, with the goal of augmenting the data to include additional relevant information.

---

license: mit

task_categories: - question-answering - text-generation
language: - en
tags: - medical
size_categories: - 1K<n<10K

Uses

• Retrieval-Augmented Generation (RAG) Models: The structured nature of the data, where answers to questions are provided within one of two contexts, is conducive to training models optimized for retrieving relevant information from the correct context.

Dataset Structure

dataset.csv

• Columns: "content", "source", "question", "answer", "new_content"
• Description: Contains questions and answers contextualized in Malawi's Technical Guidelines for Integrated Disease Surveillance and Response (TGs for IDSR) or other relevant sources. Questions vary in type, including "what", "why", "who", "where", and those seeking comparisons between concepts. Each question is accompanied by two contexts and an answer extracted from one of the two contexts. Each context includes reference (source) information.

Note: For contexts within the content column, source information is provided in the source column. For contexts within the new_content column, source information is embedded at the beginning of the text.

Source Data

• Challenge Link: Malawi Public Health Systems LLM Challenge
• Dataset Source: Zindi - Malawi Public Health Systems LLM Challenge Data
• Data Source Producers: AI Lab of the Malawi University of Business and Applied Sciences

Dataset Card Contact

• Email: Arnauld.adjovi@isheero.com