Hugging Face's logo

Datasets:
Bronya-Rand
/
genshin-impact

Modalities:
Text
Formats:
json
Languages:
English
Size:
1K - 10K
Libraries:
Datasets
pandas
Dataset card Viewer Files Community
1
genshin-impact / README.md
Bronya-Rand's picture
Bronya-Rand
Update README.md
bbf58ad verified
preview code
|
raw
history blame contribute delete
6.72 kB
metadata
language:
  - en
pretty_name: Genshin Impact
size_categories:
  - 1K<n<10K

Genshin Impact

A collection of dataset entries from Genshin Impact's official wiki (HoYoLab), containing information I think is important in relation to AI models and it learning the world of Genshin itself.

Data Extends From Genshin Impact 1.0 - 4.4

Dataset Details

Dataset Description

The Genshin Impact dataset is a collection of (what I think are) important concepts, terms, people, dialogue and such regarding the Open-World Anime Gacha game 'Genshin Impact' from miHoYo Co. Ltd. (CN)/HoYoverse (GLB). This dataset incorporates all entries created on the Genshin Impact wiki from miHoYo's HoYoLab site.

Sections included in the dataset are:

  • Affiliations
  • Artifacts
  • Books
  • Character Archive
  • Definitions
  • Enemies and Monsters
  • Nations
  • Teyvat's Resources
  • Wildlife

Curated by: Bronya-Rand

Dataset Structure based on anthracite-org/Stheno-Data-Filtered

Parsed from: HoYoLab

Language(s) (NLP): English

Uses

  • Add additional knowledge to AI models about Genshin Impact and it's characters.
  • Append to existing model knowledge of the Genshin Impact universe.
  • Use a more primary source of Genshin Impact knowledge.

Out-of-Scope Use

This dataset does not contain anything that is "fandom" such as shipping and only comes from the HoYoWiki itself. Training on this dataset should not bring any misuse nor malicious use regarding the Genshin Impact universe, it's characters nor miHoYo Co. Ltd./HoYoverse in 'theory' from generating on AI models. However, any usage of this dataset should at least follow 'common sense' terminology like all other datasets.

Everything in this dataset is SFW

Dataset Structure

  • "page_id" - The page ID from HoYoLab
  • "wiki" - The wiki in question. (Genshin Impact/Honkai: Star Rail)
  • "page_name" - The name of the page (typically the name of a character/enemy/item/etc.)
  • "component_name" - The internal structure section from HoYoLab the data came from
  • "conversations" - A processed (or generic) conversation regarding the contents of the component.

    This section has a mix of knowledge (What is a Vision) and voice-over conversions to a conversation (User: About Amber? Eula: ...)

Dataset Creation

Curation Rationale

As I have made bots for both Honkai: Star Rail and Genshin Impact and made a lorebook for both franchises, I was somewhat inspired to bring HSR and Genshin Impact knowledge more than waiting on Meta, Mistral, etc. to catch up so, in a attempt to make a dataset for a model I was keen to help on, I created a parser and fetcher long ago that grabbed all HoYoLab entries and finally parsed it with HTML cleaning, specific data analysis and such (the original JSONs are a mess for AI to read).

Source Data

The source data is a mix of information about Nations, Character Backstories, Voice-Overs and such in the form of a conversation between a user and a assistant or user with the character.

Data Collection and Processing

Selection Process

  1. Is the wiki section detrimental to knowing how the Genshin Impact universe works?

    Stuff related to player content (Action Cards, Tutorial, etc.) wouldn't apply here.

  2. Is the data important for character knowledge?
  3. Is the data important to explain a specific thing in the game lore (say Enkanomiya).

Data Cleanup Custom Python Script using BeautifulSoup and html_sanitizer cleaning.

Who are the source data producers?

miHoYo Co. Ltd/HoYoverse and HoYoLab Editors

Personal and Sensitive Information

No personal information nor sensitive information is in this dataset.

Bias, Risks, and Limitations

This dataset should not bring bias, but may come with a risk of malicious use regarding taking AI responses as "real" or official to the game or a concern for too much "immersion" if one is focused too much on a conversation between a Genshin Impact character in the Genshin Impact world.

Recommendations

With everything AI, treat everything with a grain of salt and use common sense. Do not try to pass over generations with this dataset as official miHoYo dialogue between you and a Genshin Impact character and whatnot. This is supposed to be a fun dataset for all who wants to have a more knowledgeable AI about Genshin Impact itself.

Glossary

  • Affiliations - A term in the game referring to certain groups or organizations in the game (ATM seems to just be Fontaine areas/groups)
  • Artifacts - A term referring to gear one can obtain in dungeons, chests, etc.
  • Character Archive - Character information for all playable Genshin Impact characters.
  • Teyvat's Resources - All items that can be in a player's inventory in Genshin Impact.

Dataset Card Contact

@bronya_rand - Discord