Lvoxx/nier-automata-wiki-dataset

NieR: Automata Wiki Dataset

A comprehensive, structured knowledge base extracted from NieR: Automata wikis, optimized for machine learning, RAG systems, and AI applications.

📊 Dataset Overview

This dataset contains cleaned, structured knowledge from the NieR: Automata Fandom Wiki, covering:

• Characters: YoRHa units (2B, 9S, A2), Machine Lifeforms, NPCs
• Locations: City Ruins, Desert Zone, Bunker, Forest Kingdom
• Technology: Pod Programs, Weapons, Black Box systems
• Lore: Philosophical themes, story events, world-building
• Gameplay: Quests, items, enemies, mechanics

Stats

• Total Entries: ~[number] pages
• Fields: 15 structured columns
• Coverage: Characters, Locations, Technology, Events, Philosophy, Quests
• Language: English
• Format: CSV (UTF-8)

📁 Dataset Structure

File Location

dataset/
└── data.csv    # Main dataset file

Schema

Column	Type	Description	Example
title	string	Page title	"A120: Repair"
url	string	Source URL	"https://nier.fandom.com/wiki/A120:_Repair"
source	string	Data source	"nier_fandom"
page_type	string	Content category	"AI", "Character", "Location", "Technology"
summary	string	Brief overview (1-7 sentences)	"A120: Repair is a Pod Program in NieR: Automata..."
content	string	Full cleaned text content	Complete article text without HTML
lore_layer	string	Narrative depth	"physical", "psychological", "philosophical", "metaphysical", "meta"
entities	string	Key entities mentioned	Named characters, locations, items
emotion_tags	string	Emotional themes	"loss", "hope", "existentialism"
related_topics	string	Connected concepts	Related pages and topics
activation	string	Activation/usage info	How to use items/programs
quest_log	string	Quest-related data	Quest steps and objectives
rewards	string	Obtainable rewards	Items, weapons, programs
notes	string	Additional notes	Miscellaneous information
tags	string	Search keywords	Pipe-separated tags for filtering

Page Types

• Character: Playable and non-playable characters
• Location: Areas and zones in the game world
• Technology: Weapons, Pod Programs, systems
• AI: Artificial intelligence and androids
• Machine: Machine Lifeforms and networks
• Quest: Missions and side quests
• Event: Story events and endings
• Philosophy: Existential and thematic content
• Lore: World-building and backstory

Lore Layers

Content is classified into narrative depth levels:

1. Physical: Tangible objects, locations, combat mechanics
2. Psychological: Character emotions, motivations, relationships
3. Philosophical: Existential themes, meaning, purpose
4. Metaphysical: Consciousness, soul, existence concepts
5. Meta: Fourth-wall breaks, player interaction, narrative structure

🎯 Use Cases

1. RAG (Retrieval-Augmented Generation)

Build lore-accurate chatbots and question-answering systems:

# Query: "What is the Black Box?"
# Filter: page_type="Technology", entities contains "Black Box"
# Retrieve: Relevant rows with lore_layer context

2. AI Character Roleplay

Create authentic YoRHa/Machine character interactions:

# Query: "How would 2B respond to questions about emotions?"
# Filter: entities="2B", emotion_tags contains "humanity"
# Use: psychological + philosophical lore layers

3. Game Knowledge Base

Build wikis, guides, and reference tools:

# Query: "Show all Pod Programs"
# Filter: page_type="AI", tags contains "Pod Program"
# Display: title, summary, activation, rewards

4. Narrative Analysis

Study game themes and storytelling:

# Query: "Analyze existential themes"
# Filter: lore_layer="philosophical", emotion_tags="existentialism"
# Analyze: Summary and content fields

5. Semantic Search

Vector embeddings for intelligent retrieval:

# Embed 'content' field for similarity search
# Use 'tags' and 'entities' for metadata filtering
# Leverage 'lore_layer' for context-aware ranking

🚀 Quick Start

Load Dataset

Python (pandas)

import pandas as pd

df = pd.read_csv('dataset/data.csv')
print(f"Total entries: {len(df)}")
print(df['page_type'].value_counts())

Python (DuckDB)

import duckdb

con = duckdb.connect()
df = con.execute("SELECT * FROM 'dataset/data.csv'").df()

R

library(readr)
data <- read_csv("dataset/data.csv")
summary(data$page_type)

Example Queries

Find all Pod Programs:

pod_programs = df[df['tags'].str.contains('Pod Program', na=False)]
print(pod_programs[['title', 'summary', 'rewards']])

Get philosophical content:

philosophy = df[df['lore_layer'] == 'philosophical']
print(philosophy[['title', 'summary', 'emotion_tags']])

Search by entity:

yorha_2b = df[df['entities'].str.contains('2B', na=False)]
print(yorha_2b[['title', 'page_type', 'summary']])

Filter by emotion:

existential = df[df['emotion_tags'].str.contains('existentialism', na=False)]
print(existential[['title', 'lore_layer', 'content']])

📚 Sample Data

title,page_type,summary,lore_layer,entities,emotion_tags
"YoRHa No.2 Type B",Character,"2B is a combat android...",psychological,"2B | YoRHa | 9S","duty | loneliness"
"Black Box",Technology,"Cores that power YoRHa units...",metaphysical,"YoRHa | Black Box","existentialism"
"City Ruins",Location,"Post-apocalyptic urban area...",physical,"Resistance | Machines",""

🛠️ Data Processing

Cleaning Applied

• ✅ HTML tags removed
• ✅ Wiki markup cleaned
• ✅ Navigation elements stripped
• ✅ Citations normalized
• ✅ Whitespace standardized
• ✅ Special characters handled

Quality Checks

• All entries have title, url, source
• content field never empty
• page_type classified
• tags extracted for searchability

📖 Citation & Attribution

Data Source

This dataset is derived from:

• NieR: Automata Wiki (Fandom): https://nier.fandom.com/

License

• Dataset License: Apache 2.0
• Original Content: Derived from Fandom Wiki (CC BY-SA 3.0)
• Game Content: ©SQUARE ENIX CO., LTD. All Rights Reserved.

Note: While this dataset is released under Apache 2.0, please be aware that the underlying content is derived from Fandom wikis which use CC BY-SA 3.0. Commercial use should verify compatibility with both licenses.

How to Cite

@dataset{nier_automata_wiki_dataset,
  title={NieR: Automata Wiki Dataset},
  author={lvoxx},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Lvoxx/nier-automata-wiki-dataset}
}

Fair Use Statement

This dataset is created for:

• Educational purposes
• Research and analysis
• Non-commercial AI/ML applications
• Fan projects and community tools

Please respect the original creators and SQUARE ENIX's intellectual property.

🔧 Advanced Usage

Build Vector Database

from sentence_transformers import SentenceTransformer
import chromadb

# Load model
model = SentenceTransformer('all-MiniLM-L6-v2')

# Embed content
df['embedding'] = df['content'].apply(lambda x: model.encode(x))

# Store in ChromaDB
client = chromadb.Client()
collection = client.create_collection("nier_knowledge")

for idx, row in df.iterrows():
    collection.add(
        ids=[str(idx)],
        embeddings=[row['embedding'].tolist()],
        metadatas=[{
            'title': row['title'],
            'page_type': row['page_type'],
            'lore_layer': row['lore_layer']
        }],
        documents=[row['content']]
    )

Build RAG Pipeline

from langchain.vectorstores import Chroma
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.llms import Ollama
from langchain.chains import RetrievalQA

# Load embeddings
embeddings = HuggingFaceEmbeddings()

# Create vector store
vectorstore = Chroma.from_texts(
    texts=df['content'].tolist(),
    metadatas=df[['title', 'page_type', 'lore_layer']].to_dict('records'),
    embedding=embeddings
)

# Create RAG chain
llm = Ollama(model="llama2")
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=vectorstore.as_retriever(),
    return_source_documents=True
)

# Query
response = qa_chain("What is the purpose of YoRHa?")
print(response['result'])

Entity Relationship Graph

import networkx as nx
import matplotlib.pyplot as plt

# Build entity co-occurrence graph
G = nx.Graph()

for idx, row in df.iterrows():
    entities = str(row['entities']).split(' | ')
    for i, e1 in enumerate(entities):
        for e2 in entities[i+1:]:
            if G.has_edge(e1, e2):
                G[e1][e2]['weight'] += 1
            else:
                G.add_edge(e1, e2, weight=1)

# Visualize
pos = nx.spring_layout(G)
nx.draw(G, pos, with_labels=True, node_color='lightblue')
plt.show()

🤝 Contributing

Contributions welcome! Areas for improvement:

• Add missing pages
• Enhance emotion_tags classification
• Improve lore_layer accuracy
• Extract more structured fields (stats, requirements)
• Add multilingual support

📧 Contact

For questions, issues, or collaboration:

• Open an issue on GitHub
• Email: lvoxxartist@gmail.com

🙏 Acknowledgments

• Yoko Taro - Game director and creator
• SQUARE ENIX - Publisher
• PlatinumGames - Developer
• NieR Wiki contributors - Community documentation
• Fandom - Wiki platform

---

Glory to Mankind 🤖✨

This dataset is a fan-made project and is not affiliated with SQUARE ENIX or PlatinumGames.