smainye/mpesa-swahili-qa-pairs-v1

This dataset is a remastered version prepared using Adaption's Adaptive Data platform.

mpesa_swahili_qa_pairs

This dataset contains instruction-following and reading comprehension pairs focused on M-Pesa operations, financial literacy, and agent services in Kenya. The content features natural Swahili-English code-switching typical of urban Kenyan registers alongside formal Kiswahili, targeting users from beginner to intermediate levels. Samples cover diverse scenarios including customer support queries, budgeting advice, merchant registration, and factual questions derived from context passages.

Dataset size

There are 134 data points in this dataset. This is an instruction tuning dataset.

Quality of Remastered Dataset

The final quality is B, with a relative quality improvement of 25.7%.

Domain

• Personal-finance (65%)
• Language (13%)
• Corporate-business (10%)

Tone

• Informative (39%)
• Helpful (26%)
• Explanatory (7%)

Evaluation Results

• Quality Gains:

  

• Grade Improvement:

  

• Percentile Chart: