MythoSemantic: An Interactive System for Discovering Semantic Parallels Across Mythological and Religious Traditions

Nikita Ganziuk; Innokentiy S. Humonen; Ilya Makarov; Pavel Lebedev

cs.CLOct 20, 2026

MythoSemantic: An Interactive System for Discovering Semantic Parallels Across Mythological and Religious Traditions

Authors: Nikita Ganziuk, Innokentiy S. Humonen, Ilya Makarov, Pavel Lebedev

Organizations: HSE University · AXXX · Trusted AI Research Center, RAS · AI Talent Hub, ITMO University ·

Abstract

We present a tool for the comparative study of mythological and religious texts. Traditional manual analysis is highly labor-intensive and difficult to scale. To address this, we use neural language models to embed texts into a shared semantic space, so that users can find semantic parallels across cultural traditions in an interactive interface. Embedded fragments form distinct semantic structures whose coarse geometry aligns with well-documented historical relations between traditions. We quantify how much of that geometry is already explained by lexical overlap and by proper names. The current system is available at https://mythoscope.io, and the installable package is available at https://github.com/haafus/MythoSemantic/blob/main/RUN.md.

Figures & tables

Figure 1: The Similarity view of the MythoSemantic interface. The left panel provides a tradition-grouped browser of the corpus; the central panel shows a 2D UMAP projection of BAAI/bge-m3 embeddings, with a tooltip surfacing the tradition, source text, and passage for the fragment under the cursor; the right panel displays the nearest cross-tradition neighbor of the currently selected fragment alongside a free-text semantic search box.

Figure 2: The Clusterisation view: quality metrics (silhouette, ARI, NMI, V-measure, cluster count, noise ratio), a UMAP cluster visualization, and the tradition-cluster correspondence matrix, shown here for BAAI/bge-m3 with OPTICS.

Model	Vector Dimension
sentence-transformers/ paraphrase-multilingual-MiniLM- L12-v2	384
sentence-transformers/ LaBSE	768
intfloat/e5-large-v2	1024
BAAI/bge-m3	1024
Qwen/Qwen3-Embedding-4B	2560

Table 1: Embedding models and their respective vector dimensions.

Figure 3: Distance heatmap between traditions for BAAI/bge-m3, as rendered in the Similarity view (Method: Distance Heatmap).

	Rank among 231 pairs
Tradition pair	Emb.	Lex.	No names	Names
Greek $\leftrightarrow$ Roman	1	9	17	5
Christianity $\leftrightarrow$ Islam	2	1	5	1
Austr. Abor. $\leftrightarrow$ W. African	3	8	10	131
Anglo-Saxon $\leftrightarrow$ Celtic	4	73	53	90
Buddhism $\leftrightarrow$ Taoism	6	27	31	223

Table 2: The tradition pairs discussed in § 5.1 , ranked by closeness under the BGE-M3 embedding and under three TF-IDF views of the same fragments: full text, text with proper names removed, and proper names only. A low embedding rank next to a high lexical rank marks a proximity that word overlap does not explain.

Algorithm	Sil.	ARI	NMI	V	Clusters	Noise
BIRCH	0.019	0.410	0.690	0.690	22	0.0%
Gaussian Mixture	0.472	0.463	0.770	0.770	22	0.0%
HDBSCAN	0.688	0.904	0.908	0.908	15	3.1%
K-means	0.033	0.320	0.651	0.651	22	0.0%
MeanShift	0.573	0.823	0.783	0.783	7	0.0%
OPTICS	0.693	0.925	0.918	0.918	15	13.0%

Table 3: Clustering metrics for BAAI/bge-m3 . Sil. = silhouette coefficient; ARI = Adjusted Rand Index against tradition labels; NMI = Normalized Mutual Information; V = V-measure; Clusters and Noise = number of clusters discovered and the percentage of fragments left unassigned as noise (applicable to density-based algorithms only).

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Tradition	Source text(s)
Christianity	King James Bible
Islam	Quran
Confucianism	Analects of Confucius
Hinduism	Upanishads; Ramayana of Valmiki;
	Bhagavad Gita; Mahabharata
Buddhism	Dhammapada

Appendix

Table A.1: The 22 traditions in the corpus and their source text(s).

Algorithm	Sil.	ARI	NMI	V	Clusters/Noise
BIRCH	0.033	0.332	0.662	0.662	22 / 0.0%
Gaussian Mixture	0.525	0.497	0.748	0.748	22 / 0.0%
HDBSCAN	0.633	0.764	0.850	0.850	14 / 2.4%
K-means	0.042	0.306	0.648	0.648	22 / 0.0%
MeanShift	0.566	0.797	0.782	0.782	7 / 0.0%
OPTICS	0.731	0.878	0.888	0.888	14 / 33.6%

Appendix

Table C.1: Clustering metrics for Qwen/Qwen3-Embedding-4B . Sil. = silhouette coefficient; ARI = Adjusted Rand Index against tradition labels; NMI = Normalized Mutual Information; V = V-measure; Clusters/Noise = number of clusters discovered and the percentage of fragments left unassigned as noise (applicable to density-based algorithms only).

Algorithm	Sil.	ARI	NMI	V	Clusters/Noise
BIRCH	—	—	—	—	1 / 0.0%
Gaussian Mixture	0.507	0.498	0.807	0.807	22 / 0.0%
HDBSCAN	0.699	0.977	0.960	0.960	16 / 3.4%
K-means	0.031	0.368	0.701	0.701	22 / 0.0%
MeanShift	0.394	0.564	0.666	0.666	6 / 0.0%
OPTICS	0.687	0.973	0.956	0.956	16 / 1.8%

Appendix

Table C.2: Clustering metrics for intfloat/e5-large-v2 . Sil. = silhouette coefficient; ARI = Adjusted Rand Index against tradition labels; NMI = Normalized Mutual Information; V = V-measure; Clusters/Noise = number of clusters discovered and the percentage of fragments left unassigned as noise (applicable to density-based algorithms only).

Algorithm	Sil.	ARI	NMI	V	Clusters/Noise
BIRCH	0.024	0.402	0.647	0.647	22 / 0.0%
Gaussian Mixture	0.436	0.471	0.737	0.737	22 / 0.0%
HDBSCAN	0.616	0.716	0.793	0.793	15 / 4.8%
K-means	0.029	0.287	0.592	0.592	22 / 0.0%
MeanShift	0.615	0.688	0.755	0.755	11 / 0.0%
OPTICS	0.801	0.906	0.906	0.906	13 / 40.0%

Appendix

Table C.3: Clustering metrics for sentence-transformers/LaBSE . Sil. = silhouette coefficient; ARI = Adjusted Rand Index against tradition labels; NMI = Normalized Mutual Information; V = V-measure; Clusters/Noise = number of clusters discovered and the percentage of fragments left unassigned as noise (applicable to density-based algorithms only).

Algorithm	Sil.	ARI	NMI	V	Clusters/Noise
BIRCH	0.006	0.247	0.455	0.455	22 / 0.0%
Gaussian Mixture	0.312	0.334	0.569	0.569	22 / 0.0%
HDBSCAN	0.102	0.019	0.088	0.088	3 / 2.4%
K-means	0.026	0.221	0.451	0.451	22 / 0.0%
MeanShift	0.429	0.533	0.540	0.540	4 / 0.0%
OPTICS	0.099	0.018	0.086	0.086	3 / 1.1%

Appendix

Table C.4: Clustering metrics for MiniLM-L12-v2 . Sil. = silhouette coefficient; ARI = Adjusted Rand Index against tradition labels; NMI = Normalized Mutual Information; V = V-measure; Clusters/Noise = number of clusters discovered and the percentage of fragments left unassigned as noise (applicable to density-based algorithms only).

Figure D.1: Visualizations for all used models and dimension reduction algorithms.

Quantifying the Occult: A Comparative Study of Hindu and Buddhist Deities Using Machine Learning Methods
Sep 22, 2026Ankit Bhattacharjee
Department of Mechanical Engineering Indian Institute of Technology Kharagpur
Cultural Awareness is Represented but Not Decoded: Tracing Mythological Knowledge across 18 Open-Source LLMs
Aug 3, 2026Iaroslav Chelombitko, Ekaterina Chelombitko, Mika HämäläinenMultilingual Language Models LLM Interpretability
1DataSpike · 2Metropolia University of Applied Sciences, Helsinki, Finland · 3Neapolis University Pafos, Paphos, Cyprus
From transcription to semantic corpus analysis: unsupervised learning of sentence representations for ancient languages
Jul 27, 2026Th{é}otime de la SelleSemantic Textual Similarity Sentence Embeddings
ISC, HiSoMA, CNRS · Institut des Sources Chrétiennes, HiSoMA, CNRS, Lyon, France

Sep 22, 2026J/K move · Enter open · S save

Quantifying the Occult: A Comparative Study of Hindu and Buddhist Deities Using Machine Learning Methods

Ankit Bhattacharjee

Department of Mechanical Engineering Indian Institute of Technology Kharagpur

Open paper AskPDF arXiv

This study introduces a dual-matrix computational architecture to mathematically quantify the morphological and theological divergence of 196 Hindu and Vajrayana Buddhist esoteric deities. Physical morphology is evaluated via a discrete Gower distance matrix enhanced by a novel "Cardinality Weighting" algorithm, while theological function is mapped via dense vector embeddings generated from Large Language Model (LLM) semantic expansions, explicitly utilized as a synthetic proxy to mitigate circular reasoning. The multi-modal topological projections provide algorithmic validation of "iconographic camouflage", demonstrating how distinct visual forms structurally obscure shared cross-tradition functions. Furthermore, I computationally model the "Atin Effect" - serving simultaneously as a psychological observation of sequential cognitive bias and a machine learning benchmark - demonstrating how high-cardinality esoteric anchors (e.g., a veena or a severed head) override systemic theological disparities to mathematically cluster orthodox and Tantric entities. Cross-tradition spatial analysis establishes that the highest esoteric manifestations, such as the Hindu Chinnamasta and the Buddhist Chinnamunda, share a near-identical mathematical coordinate across both visual (

D_G = 0.288

) and semantic (

D_C = 0.068

) boundaries, indicating a 1:1 esoteric transfer. By open-sourcing this architecture, I provide a scalable, unsupervised machine learning tool for Digital Humanities scholars and comparative theologians to rigorously map latent structural continuities across qualitative cultural corpora.

Abstract

Explore similar work

Quantifying the Occult: A Comparative Study of Hindu and Buddhist Deities Using Machine Learning Methods

Cultural Awareness is Represented but Not Decoded: Tracing Mythological Knowledge across 18 Open-Source LLMs

From transcription to semantic corpus analysis: unsupervised learning of sentence representations for ancient languages