File size: 1,925 Bytes
3f8ed2e
 
dd07466
 
 
 
 
 
 
 
 
3f8ed2e
dd07466
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
---
license: cc
task_categories:
- question-answering
- text-generation
language:
- sw
- en
pretty_name: Dolly_Swahili
size_categories:
- 10K<n<100K
---
# README for Swahili Translated Dataset from Toloka

## Dataset Description

This dataset is a Dolly 15k translated from English to Swahili, filtered and processed using the Toloka platform. It includes various contexts, responses, and instructions from diverse domains, providing a rich resource for natural language processing tasks, particularly for those focusing on the Swahili language.

### Data Fields

- **task_id**: A unique identifier for each task in the dataset.
- **INPUT:context_tr**: The context of the task translated into Swahili.
- **INPUT:context_src**: The original context of the task in English.
- **INPUT:response_tr**: The response or answer to the task, translated into Swahili.
- **INPUT:response_src**: The original response or answer to the task in English.
- **INPUT:instruction_tr**: The instruction for the task, translated into Swahili.
- **INPUT:instruction_src**: The original instruction for the task in English.
- **toloka probabilities**: Probabilities indicating the confidence or quality score associated with each task, as evaluated on the Toloka platform.

### Data Format

The dataset is provided in CSV format.

## Dataset Creation

### Curation Rationale

This dataset was curated to enhance the resources available for Swahili natural language processing. It aims to provide a robust set of data for translation tasks, question-answering systems, and other language modeling tasks in Swahili.

### Source Data

#### Initial Data Collection and Normalization

The initial data collection was taken from Dolly-15K

### Dataset Curators

This dataset was curated by Nazar Beknazarov, leveraging the Toloka platform for translation and filtering.


### Citation Information

Please include citation information here, if available.