Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
csv
Sub-tasks:
multi-class-classification
Languages:
English
Size:
< 1K
License:
Add dataset files and documentation
Browse filesAdd train/validation/test CSV splits (400 examples total, 200 per class), dataset card (README.md), and metadata (dataset_info.json).
Binary classification dataset derived from ibm-research/watsonxDocsQA.
Labels: conceptual (0), how-to (1). Split: 70/15/15, stratified.
- README.md +93 -0
- dataset_info.json +66 -0
- test.csv +0 -0
- train.csv +0 -0
- validation.csv +0 -0
README.md
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
annotations_creators:
|
| 3 |
+
- expert-generated
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
license:
|
| 7 |
+
- other
|
| 8 |
+
pretty_name: Watsonx Docs Document Type Classification
|
| 9 |
+
size_categories:
|
| 10 |
+
- n<1K
|
| 11 |
+
source_datasets:
|
| 12 |
+
- ibm-research/watsonxDocsQA
|
| 13 |
+
task_categories:
|
| 14 |
+
- text-classification
|
| 15 |
+
task_ids:
|
| 16 |
+
- document-classification
|
| 17 |
+
---
|
| 18 |
+
|
| 19 |
+
# Watsonx Docs Document Type Classification
|
| 20 |
+
|
| 21 |
+
This dataset is a balanced binary document-level classification subset derived
|
| 22 |
+
from `ibm-research/watsonxDocsQA`.
|
| 23 |
+
|
| 24 |
+
## Task
|
| 25 |
+
|
| 26 |
+
Classify IBM Watsonx documentation pages by their dominant user-facing purpose:
|
| 27 |
+
|
| 28 |
+
- `conceptual`: documents primarily used to understand or look up information.
|
| 29 |
+
- `how-to`: documents primarily used to complete a procedure or fix a problem.
|
| 30 |
+
|
| 31 |
+
## Splits
|
| 32 |
+
|
| 33 |
+
| Split | conceptual | how-to | Total |
|
| 34 |
+
|---|---:|---:|---:|
|
| 35 |
+
| train | 140 | 140 | 280 |
|
| 36 |
+
| validation | 30 | 30 | 60 |
|
| 37 |
+
| test | 30 | 30 | 60 |
|
| 38 |
+
|
| 39 |
+
## Fields
|
| 40 |
+
|
| 41 |
+
- `doc_id`: original document ID from the source dataset.
|
| 42 |
+
- `url`: source documentation URL.
|
| 43 |
+
- `title`: documentation page title.
|
| 44 |
+
- `text`: model input text, constructed as `title + "\n" + first 800 words of document`. The title is preserved in full; the document body is truncated to keep inputs manageable for embedding-based classifiers.
|
| 45 |
+
- `label`: string label, either `conceptual` or `how-to`.
|
| 46 |
+
- `label_id`: numeric label ID, where `conceptual = 0` and `how-to = 1`.
|
| 47 |
+
- `split`: dataset split.
|
| 48 |
+
|
| 49 |
+
## Usage
|
| 50 |
+
|
| 51 |
+
```python
|
| 52 |
+
from datasets import load_dataset
|
| 53 |
+
|
| 54 |
+
data_files = {
|
| 55 |
+
"train": "train.csv",
|
| 56 |
+
"validation": "validation.csv",
|
| 57 |
+
"test": "test.csv",
|
| 58 |
+
}
|
| 59 |
+
|
| 60 |
+
dataset = load_dataset("csv", data_files=data_files)
|
| 61 |
+
```
|
| 62 |
+
|
| 63 |
+
## Curation Notes
|
| 64 |
+
|
| 65 |
+
IBM technical documentation has traditionally been structured around DITA
|
| 66 |
+
(Darwin Information Typing Architecture), which classifies documents into four
|
| 67 |
+
types: `task`, `concept`, `reference`, and `troubleshooting`. This dataset
|
| 68 |
+
adapts that taxonomy into two classes: `conceptual` merges `concept` and
|
| 69 |
+
`reference` (both primarily information-seeking); `how-to` merges `task` and
|
| 70 |
+
`troubleshooting` (both action- or fix-oriented). The binary schema was chosen
|
| 71 |
+
because `troubleshooting` was too rare to form a reliable standalone class, and
|
| 72 |
+
`reference` and `concept` were difficult to separate consistently.
|
| 73 |
+
|
| 74 |
+
Annotation followed a semi-automatic process. Labelling rules were first defined
|
| 75 |
+
based on IBM Writing Style guidelines, then applied by a heuristic script to
|
| 76 |
+
generate candidate labels. Each candidate was assigned a confidence tier:
|
| 77 |
+
`title_high` (strong title signal), `body_medium` (body-text signal only, no
|
| 78 |
+
strong title match), or `default_low` (no strong signal in either title or
|
| 79 |
+
body). All tiers except `body_medium` how-to rows were manually reviewed. The
|
| 80 |
+
`body_medium` how-to subset (333 rows) was left unreviewed because the remaining
|
| 81 |
+
manually checked data was sufficient to construct a balanced 400-example
|
| 82 |
+
dataset; retaining unreviewed borderline rows would have introduced noise
|
| 83 |
+
without benefit.
|
| 84 |
+
|
| 85 |
+
Rows marked `X` during manual review were removed because the source document
|
| 86 |
+
was incomplete or too ambiguous to label reliably. Rows marked `?` were
|
| 87 |
+
interpreted as belonging to the opposite binary class.
|
| 88 |
+
|
| 89 |
+
The final subset contains 400 examples, sampled with random seed `42` after
|
| 90 |
+
manual correction and filtering.
|
| 91 |
+
|
| 92 |
+
License follows the terms of the source dataset `ibm-research/watsonxDocsQA`.
|
| 93 |
+
Please refer to the original dataset for licensing details.
|
dataset_info.json
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset_name": "watsonx-docs-document-type",
|
| 3 |
+
"version": "1.0.0",
|
| 4 |
+
"date_created": "2026-05",
|
| 5 |
+
"source_dataset": "ibm-research/watsonxDocsQA",
|
| 6 |
+
"task": "binary document-level technical documentation type classification",
|
| 7 |
+
"labels": {
|
| 8 |
+
"conceptual": 0,
|
| 9 |
+
"how-to": 1
|
| 10 |
+
},
|
| 11 |
+
"text_field": "text",
|
| 12 |
+
"text_construction": "title + '\\n' + first 800 words of document",
|
| 13 |
+
"label_field": "label",
|
| 14 |
+
"label_id_field": "label_id",
|
| 15 |
+
"random_seed": 42,
|
| 16 |
+
"source_rows": 1144,
|
| 17 |
+
"annotation_process": {
|
| 18 |
+
"step_1": "heuristic pre-annotation using title and body patterns based on IBM Writing Style guidelines",
|
| 19 |
+
"step_2": "human review of all rows except suggested_how_to_body_medium subset",
|
| 20 |
+
"confidence_tiers": {
|
| 21 |
+
"title_high": "strong title signal",
|
| 22 |
+
"body_medium": "body-text signal only, no strong title match",
|
| 23 |
+
"default_low": "no strong signal in either title or body"
|
| 24 |
+
},
|
| 25 |
+
"manual_markers": {
|
| 26 |
+
"empty": "accept heuristic label",
|
| 27 |
+
"?": "flip to opposite binary class",
|
| 28 |
+
"X": "remove document"
|
| 29 |
+
}
|
| 30 |
+
},
|
| 31 |
+
"excluded": {
|
| 32 |
+
"suggested_how_to_body_medium": {
|
| 33 |
+
"count": 333,
|
| 34 |
+
"reason": "not manually reviewed; weak body-only heuristic signal considered unreliable"
|
| 35 |
+
},
|
| 36 |
+
"final_label_X": {
|
| 37 |
+
"count": 17,
|
| 38 |
+
"reason": "document was incomplete, low-quality, or too ambiguous to label reliably"
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"usable_after_exclusion": {
|
| 42 |
+
"how-to": 231,
|
| 43 |
+
"conceptual": 563
|
| 44 |
+
},
|
| 45 |
+
"selected": {
|
| 46 |
+
"conceptual": 200,
|
| 47 |
+
"how-to": 200
|
| 48 |
+
},
|
| 49 |
+
"splits": {
|
| 50 |
+
"train": {
|
| 51 |
+
"conceptual": 140,
|
| 52 |
+
"how-to": 140,
|
| 53 |
+
"total": 280
|
| 54 |
+
},
|
| 55 |
+
"validation": {
|
| 56 |
+
"conceptual": 30,
|
| 57 |
+
"how-to": 30,
|
| 58 |
+
"total": 60
|
| 59 |
+
},
|
| 60 |
+
"test": {
|
| 61 |
+
"conceptual": 30,
|
| 62 |
+
"how-to": 30,
|
| 63 |
+
"total": 60
|
| 64 |
+
}
|
| 65 |
+
}
|
| 66 |
+
}
|
test.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
train.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
validation.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|