Datasetpublic

msw-ai-tf/maplestory-worlds-creator-qa

MapleStory Worlds Creator QA Synthetic question-answer dataset built from the official MapleStory Worlds Creator Center documentation. Questions are generated to be self-contained and grounded in the source docs; answers avoid source/meta references so they read like an expert explanation. Some QA pairs are composed from multiple related documents (see combo_sources). Parallel Korean/English. Intended for instruction tuning, QA, and retrieval. Composition… See the full description on the dataset page: https://huggingface.co/datasets/msw-ai-tf/maplestory-worlds-creator-qa.

sourceHugging Facecc-by-4.0updated 3mo agoView on Hugging Face

2likes55downloads

Dataset Card

MapleStory Worlds Creator QA

Synthetic question-answer dataset built from the official MapleStory Worlds Creator Center documentation. Questions are generated to be self-contained and grounded in the source docs; answers avoid source/meta references so they read like an expert explanation. Some QA pairs are composed from multiple related documents (see combo_sources).

Parallel Korean/English. Intended for instruction tuning, QA, and retrieval.

Composition

config	Rows
`ko`	61908
`en`	38194
Total	100102

Multi-document (composite) QA: 69256. Question types: {'how_to': 32407, 'property': 21504, 'constraint': 12865, 'comparison': 11961, 'definition': 11868, 'usage': 9497}.

Usage

python

from datasets import load_dataset

# Korean only
ds_ko = load_dataset("msw-ai-tf/maplestory-worlds-creator-qa", "ko", split="train")

# English only
ds_en = load_dataset("msw-ai-tf/maplestory-worlds-creator-qa", "en", split="train")

# Combine both
from datasets import concatenate_datasets
ds_all = concatenate_datasets([ds_ko, ds_en])

Features

Column	Description
`id`	Unique QA id
`language`	`KO` / `EN`
`question_type`	definition / how_to / property / comparison / constraint / usage
`variant_type`	platform_alias / paraphrase / typo / colloquial (surface form)
`platform_alias`	platform name used in the question (MSW / 메월드 / ...)
`question`	user question
`thinking`	reasoning toward the answer
`answer`	self-contained explanatory answer
`intent`	normalized question intent
`source_type` / `source_id` / `source_url` / `source_title`	seed source doc
`combo_sources`	JSON list of docs when composed from multiple (else empty)
`gen_model`	model used to generate the answer

Generation

Documents → embedding-based related-doc combination → question synthesis (diverse types, platform aliases, paraphrases, typos) → embedding-based near-duplicate removal (Qdrant) → grounded self-contained answers.

Source & License

—License: CC BY 4.0