Licenses & Credits

The Ideenatlas uses and combines various open-source technologies, AI research, and open data sources. Below you will find an overview of the components used, their purpose, and their licenses.

Terms of Use

The following terms apply to the use of Ideenatlas:

• Service License & Disclaimer
  Ideenatlas is a non-commercial private research project. Its generated content (analyses, visualizations) is licensed under the terms below. This license does NOT apply to the content of the search results themselves; for these, the licenses of the respective data sources (e.g., arXiv) apply. All results may contain errors and are provided without warranty.
  License: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International

Special Acknowledgements

The Ideenatlas would not have been possible in its current form without external support.

• bwHPC (BwUniCluster3.0)
  Only through the provision of high-performance computing resources within the framework of the BwUniCluster3.0 has the vectorization and clustering of the vast volume of research papers listet in the Ideenatlas become financially and temporally feasible. Special thanks are also due for the trust placed in me as an independent student regarding the implementation of and active research on this project. In this context, the current web interface is a purely non-commercial proof of concept, serving primarily to demonstrate the underlying concept of the Ideenatlas. Future research will focus on improving the hierarchical clustering to yield the best possible results.
  
  The authors acknowledge support by the state of Baden-Württemberg through bwHPC.

Data Sources & Licenses

The Ideenatlas exclusively aggregates metadata from the following scientific open-access sources. The licensing conditions for the full texts of the articles remain unaffected by this and can be viewed on the respective source platforms.

• arXiv.org
  Source for preprints in physics, mathematics, computer science, and more.
  Metadata License: Creative Commons CC0 1.0 Universal Public Domain Dedication. The metadata can be used without restrictions.
• PubMed Central (PMC) Open Access Subset
  Source for biomedical and life sciences journal literature.
  Metadata License: The use of metadata from the "Non-Commercial Use" subset is permitted for non-commercial purposes. The licenses of the associated articles are typically CC BY-NC, CC BY-NC-SA, or CC BY-NC-ND.
  PMC requests the following citation:
  PMC Open Access Subset [Internet]. Bethesda (MD): National Library of Medicine. 2003 - [cited YEAR MONTH DAY]. Available from https://pmc.ncbi.nlm.nih.gov/tools/openftlist/.
• openRxiv (bioRxiv & medRxiv)
  Source for preprints in the life sciences and health sciences.
  Metadata License: The metadata is made available via a public API for free use for research and discovery purposes. There is no separate formal license, but use within the scope of such a project is intended.
• RePEc (Research Papers in Economics)
  Source for research papers in economics.
  Metadata License: The metadata is freely available for non-commercial use. The operators request that RePEc be cited as the source when using the data.
• PhilPapers
  Source for research papers in philosophy.
  Metadata License: In accordance with the principles of the OAI initiative (Open Archives Initiative), the metadata is provided for non-commercial purposes and to promote interoperability between scientific archives.

AI Models & API Services

Semantic analysis and processing are based on:

• Google Gemini
  Generation of cluster names and thematic descriptions, reformulation of user input texts.
  License: API service by Google.
• Qwen3-Embedding-0.6B
  Vectorization of texts (embeddings) for semantic search.
  License: Apache 2.0
• NLLB-200-distilled-1.3B
  Translation of search queries and documents.
  License: Creative Commons Attribution Non Commercial 4.0
• GlotLID
  Automatic language detection of input texts.
  License: Apache 2.0
• Piper Neural TTS
  Local generation of speech output (Text-to-Speech) for the read-aloud function.
  License: MIT License (Engine)
  Voice models sourced from the 'rhasspy/piper-voices' repository (https://huggingface.co/rhasspy/piper-voices). Individual voice licenses vary (mostly CC0, CC-BY, or Public Domain).

Python: Data Science, Visualization & Backend

The following is used to calculate the knowledge space, visualization, and server logic:

• SentenceTransformers
  Framework for executing embedding models.
  License: Apache 2.0
• UMAP Learn
  Dimensionality reduction in several places, including for 2D map positions.
  License: OSI Approved
• HDBSCAN
  Hierarchical clustering algorithm for topic discovery.
  License: BSD License
• Scikit-learn
  Machine learning (e.g., normalization, TF-IDF).
  License: BSD 3.0
• Matplotlib
  Generation of static map images (PNG).
  License: BSD
• Shapely
  Calculation of the geometric outlines of the clusters.
  License: BSD 3.0
• NetworkX
  Construction and management of the hierarchical knowledge graph.
  License: BSD 3.0
• Plotly
  Creation of interactive graph visualizations.
  License: MIT License
• FastText
  Base library for the GlotLID model.
  License: MIT License
• PyCountry
  Processing of ISO country codes and language names.
  License: LGPL v2 (OSI Approved)
• Flask
  Web server for Python microservices.
  License: BSD-3-Clause License
• APScheduler
  Scheduling of background tasks and cleanups.
  License: MIT License

Frontend

The user interface uses the following libraries:

• Font Awesome Free
  Provision of icons and symbols.
  License: Icons (CC BY 4.0), Fonts (SIL OFL 1.1), Code (MIT License)
• KaTeX
  Rendering of mathematical formulas in the browser.
  License: MIT License
• Tippy.js & Popper
  Display of tooltips and popovers.
  License: MIT License

Java Backend & Infrastructure

The application server is based on:

• Spring Boot
  Main framework for the web application.
  License: Apache 2.0
• Qdrant
  Vector database for storing and searching embeddings.
  License: Apache 2.0
• JGraphT
  Analysis of graph structures in the backend.
  License: Dual-licensed LGPL 2.1 / EPL 2.0
• SnuggleTeX
  Conversion of LaTeX to XML/MathML.
  License: SnuggleTeX Software License (BSD License)
• LaTeX2Unicode
  Conversion of LaTeX symbols into Unicode text.
  License: Apache 2.0

All trademarks and product names used are the property of their respective owners. The use of this open-source software and open data does not imply any direct endorsement of the project by the authors.