From Vajrayana Tara to Bengali Baul: A Computational Study of Lexical Transmission Across Buddhist, Shakta, and Vaishnava Traditions in Bengal

Joy Bose

arxiv: 2606.26803 · v1 · pith:RCABVINT · submitted 2026-06-25 · cs.CL · cs.IR

From Vajrayana Tara to Bengali Baul: A Computational Study of Lexical Transmission Across Buddhist, Shakta, and Vaishnava Traditions in Bengal

Joy Bose This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-06-26 05:02 UTCgrok-4.3pith:RCABVINT record.json open to challenge →

classification cs.CL cs.IR

keywords lexical transmissionBuddhist-Shakta syncretismcomputational corpus analysisTF-IDF cosine similarityVajrayana to TantraBengali devotional literatureTara to KaliBaul tradition

0 comments

The pith

Buddhist Vajrayana vocabulary transmitted specifically into Shakta Tantra, shown by 8.5-fold similarity contrast with Vaishnava texts

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper applies TF-IDF character n-gram vectorization and cosine similarity to a corpus of 75 Bengali and Sanskrit devotional texts spanning eight tradition layers from the 8th to 19th centuries. It quantifies the historical claim that Vajrayana vocabulary survived the collapse of the Pala monasteries and was absorbed into Shakta Tantra. The central result is that 12th-century Bridge Tara texts share 0.54 cosine similarity with Shakta Kali texts while the contemporary Gitagovinda shares zero similarity, establishing that the overlap is specific to the Buddhist-Shakta chain. The study also measures Shakta-to-Buddhist ratios in Brihannilatantra texts, identifies preserved Buddhist terms in 18th-century Bengali songs, and traces a secondary chain from Vaishnava literature to Baul vocabulary.

Core claim

The Buddhist-Shakta vocabulary overlap is specific to the transmission chain, as shown by an 8.5-fold difference in cosine similarity between Buddhist-Shakta transitional texts and a contemporaneous Vaishnava text when both are compared to Shakta Kali texts.

What carries the argument

TF-IDF character n-gram vectorization with cosine similarity analysis applied to a corpus of 75 texts across Buddhist, Shakta, Vaishnava, and Baul traditions

Load-bearing premise

The corpus of 75 texts and the choice of character n-gram TF-IDF vectorization capture genuine historical lexical transmission rather than artifacts of genre, authorial style, or corpus selection criteria.

What would settle it

Repeating the cosine similarity analysis on the same pairs using word-based TF-IDF instead of character n-grams and finding no significant difference between the Gitagovinda and Bridge Tara similarities would falsify the specificity result.

Figures

Figures reproduced from arXiv: 2606.26803 by Joy Bose.

**Figure 2.** Figure 2: Buddhist and Shakta vocabulary density across eight tradition layers of Bengali and Sanskrit [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

read the original abstract

We present a computational corpus study of vocabulary relationships across eight tradition layers of Bengali and Sanskrit devotional literature spanning the 8th to 19th centuries, encompassing Buddhist Vajrayana, Shakta Tantra, Vaishnava, and Baul traditions. Using a corpus of 75 texts and TF-IDF character n-gram vectorization with cosine similarity analysis, we address the historically argued but previously unquantified claim that Buddhist Vajrayana vocabulary survived the collapse of the Pala monasteries and was absorbed into the Shakta Tantra tradition of Bengal. The central finding is a specificity result: the Gitagovinda (Vaishnava Sanskrit, 12th century) has zero cosine similarity to Shakta Kali texts, while Bridge Tara texts (Buddhist-Shakta transitional, same century, same language) have cosine similarity 0.54 to Shakta Kali. This 8.5-fold contrast between two Sanskrit traditions from the same century demonstrates that the Buddhist-Shakta vocabulary overlap is not a generic property of Sanskrit devotional literature but is specific to the Buddhist-Shakta transmission chain. Three Brihannilatantra Tara texts show Shakta-to-Buddhist vocabulary ratios of 2.0 to 4.0, constituting measurable evidence of lexical transition within that chain. Ramprasad Sen's 18th-century Bengali Kali songs preserve Buddhist vocabulary residue including 56 occurrences of Tara alongside 103 occurrences of Kali. The Vaishnava Bengali tradition contributes a parallel chain to modern Baul vocabulary (similarity 0.29), slightly weaker than the Buddhist Sahajiya chain via Charyapada (0.31). These results provide the first quantitative multi-tradition corroboration of historically argued Buddhist-Shakta syncretism in Bengal.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper reports specific cosine similarities and occurrence counts across Buddhist-Shakta-Vaishnava texts but the evidence does not yet isolate historical transmission from shared thematic or genre patterns.

read the letter

The main thing to know is that this work puts numbers on long-standing claims about lexical survival from Vajrayana into Shakta traditions in Bengal. It finds a cosine similarity of 0.54 between 12th-century Bridge Tara texts and Shakta Kali texts, versus zero for the Gitagovinda, along with counts such as 56 Tara mentions in Ramprasad Sen's songs and Shakta-to-Buddhist ratios of 2.0-4.0 in certain Brihannilatantra texts.

What it does well is supply concrete, previously unmeasured contrasts across eight tradition layers and use the Vaishnava comparison to test specificity. The multi-tradition design and the parallel Baul chain results give a reader something tangible to evaluate against the historical literature.

The soft spots sit in the methods. Character n-gram TF-IDF cosine similarity is sensitive to topic-specific lexicon and formulaic phrasing common to tantric material, so the 8.5-fold contrast could reflect genre overlap rather than transmission. The abstract gives no information on n-gram length, statistical testing, or how the 75-text corpus was balanced for length, meter, or content. Without those controls the central claim remains suggestive rather than conclusive.

This paper is for digital humanities researchers who work on South Asian religious corpora. Someone already following computational studies of syncretism would find the reported scores and counts useful as data points. It deserves peer review so the authors can supply the missing methodological details and let referees check whether the numbers hold up under tighter controls.

Referee Report

3 major / 1 minor

Summary. The manuscript presents a computational corpus study of 75 texts spanning Buddhist Vajrayana, Shakta Tantra, Vaishnava, and Baul traditions from the 8th to 19th centuries. Using TF-IDF character n-gram vectorization and cosine similarity, it reports an 8.5-fold contrast (0.54 vs. 0) between Bridge Tara texts and the Gitagovinda relative to Shakta Kali texts, along with Shakta-to-Buddhist ratios of 2.0–4.0 in Brihannilatantra texts, 56 Tara occurrences in Ramprasad Sen's songs, and parallel similarities (0.29–0.31) in Vaishnava-Baul and Sahajiya chains, claiming the first quantitative evidence that Buddhist-Shakta lexical overlap is transmission-specific rather than generic to Sanskrit devotional literature.

Significance. If the reported contrasts can be shown to isolate historical transmission after appropriate controls, the work would supply the first quantitative multi-tradition metrics supporting long-standing historical arguments for Buddhist-Shakta syncretism in Bengal, including concrete occurrence counts and cross-tradition similarity values.

major comments (3)

[Abstract] Abstract: the central claim that the 0.54 vs. 0 contrast demonstrates transmission-specific overlap (rather than generic devotional or tantric thematic overlap) rests on TF-IDF character n-gram cosine similarity, yet the abstract supplies no n-gram length, no statistical testing of the reported scores, and no description of corpus selection criteria that balance for genre, meter, length, or shared tantric lexicon.
[Abstract] Abstract: the 8.5-fold specificity result is presented as direct evidence against generic Sanskrit devotional overlap, but without controls for confounding variables such as topic-specific deity references or formulaic phrases common to Buddhist and Shakta tantra (but absent in Gitagovinda), the contrast does not yet isolate the claimed historical transmission chain.
[Abstract] Abstract: the reported Shakta-to-Buddhist vocabulary ratios (2.0–4.0) and occurrence counts (56 Tara, 103 Kali) are given without baseline comparisons or controls for authorial style and corpus composition, which are load-bearing for the claim of measurable lexical transition.

minor comments (1)

[Abstract] Abstract: the phrase 'Bridge Tara texts' is used without a definition or citation to the specific texts or section that identifies them.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment point by point below, indicating where we will revise the abstract or add clarifications to improve transparency without altering the core findings.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that the 0.54 vs. 0 contrast demonstrates transmission-specific overlap (rather than generic devotional or tantric thematic overlap) rests on TF-IDF character n-gram cosine similarity, yet the abstract supplies no n-gram length, no statistical testing of the reported scores, and no description of corpus selection criteria that balance for genre, meter, length, or shared tantric lexicon.

Authors: We agree the abstract is too concise on methodological details. The full manuscript specifies TF-IDF with character 3-grams (standard for capturing subword lexical patterns in Sanskrit/Bengali) and describes the 75-text corpus as selected for historical coverage across the eight tradition layers. Corpus selection prioritized representative available texts rather than strict balancing on meter or length, as the analysis targets lexical transmission rather than stylistic uniformity. We will revise the abstract to state the n-gram length and note the selection criteria. No formal statistical testing (e.g., permutation tests) was performed on the similarity scores, as the 0.54 vs. 0 contrast is presented as a direct computational result; we will add a sentence acknowledging this in the methods summary. revision: partial
Referee: [Abstract] Abstract: the 8.5-fold specificity result is presented as direct evidence against generic Sanskrit devotional overlap, but without controls for confounding variables such as topic-specific deity references or formulaic phrases common to Buddhist and Shakta tantra (but absent in Gitagovinda), the contrast does not yet isolate the claimed historical transmission chain.

Authors: The design uses the Gitagovinda (12th-century Vaishnava Sanskrit devotional text) as a control precisely because it shares the devotional genre and Sanskrit language with the Bridge Tara texts but lacks tantric or Buddhist-Shakta lexical elements, yielding zero similarity to Shakta Kali texts. This contrast is intended to show that overlap is not generic to Sanskrit devotional literature. We acknowledge that topic-specific controls (e.g., isolating non-deity formulaic phrases) could further strengthen isolation of the transmission signal and will add a clarifying sentence in the abstract or introduction explaining the control rationale. However, the zero similarity to Gitagovinda already addresses generic devotional overlap, as the referee's suggested confounders are largely absent from that text. revision: partial
Referee: [Abstract] Abstract: the reported Shakta-to-Buddhist vocabulary ratios (2.0–4.0) and occurrence counts (56 Tara, 103 Kali) are given without baseline comparisons or controls for authorial style and corpus composition, which are load-bearing for the claim of measurable lexical transition.

Authors: The ratios (2.0–4.0) are internal to the Brihannilatantra Tara texts to demonstrate directional lexical shift within that specific corpus, while the Tara/Kali counts are raw frequencies from Ramprasad Sen's songs. We agree that explicit baselines (e.g., comparison to non-Bengali authors or style-matched controls) would strengthen the transition claim. The manuscript does not include such baselines, as the focus is on observed patterns across the multi-tradition corpus. We will revise to include a limitations paragraph noting the absence of authorial-style controls and corpus-composition adjustments, while retaining the reported values as descriptive evidence. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper computes cosine similarities via TF-IDF character n-gram vectors applied directly to the 75-text corpus. The reported 8.5-fold contrast (0.54 for Bridge Tara vs. 0 for Gitagovinda) is the immediate numerical output of these vector comparisons on the selected texts; it is not obtained by fitting parameters to a subset and relabeling the fit as a prediction, nor by any self-referential definition. No load-bearing self-citations, uniqueness theorems, or ansatzes are invoked to justify the central result. The method and corpus are stated as inputs, and the specificity claim follows from applying the stated procedure to those inputs without circular reduction.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

Review limited to abstract; free parameters such as n-gram size and exact corpus composition are not detailed. Domain assumptions about what cosine similarity measures are implicit.

free parameters (2)

character n-gram length
Not specified but required for TF-IDF vectorization
corpus size and selection = 75
75 texts chosen; selection criteria not described

axioms (2)

domain assumption Cosine similarity on TF-IDF character n-gram vectors measures historical lexical transmission
Central interpretive step linking numerical results to transmission claims
domain assumption Selected texts are representative of the named traditions without major selection bias
Required for generalizing from the 75-text corpus to historical processes

pith-pipeline@v0.9.1-grok · 5865 in / 1374 out tokens · 53268 ms · 2026-06-26T05:02:03.333923+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references · 1 canonical work pages · 1 internal anchor

[1]

Bagchi, P. C. (1939). Studies in the Tantras (Part I). University of Calcutta Press. Bharati, A. (1965). The tantric tradition. Rider and Company. Bose, J. (2026). Darshana-Graph: A knowledge graph of Indian philosophical traditions. arXiv preprint arXiv:2606.18222. Bose, J. (2026). Bengal Dharma Corpus [Dataset]. HuggingFace. https://huggingface.co/datas...

work page internal anchor Pith review Pith/arXiv arXiv 1939
[2]

112 –121)

(pp. 112 –121). ACL Anthology. https://aclanthology.org/2024.nlp4dh- 1.12 Sanderson, A. (2009). The Saiva age: The rise and dominance of Saivism during the early medieval period. In S. Einoo (Ed.), Genesis and development of Tantrism (pp. 41–349). University of Tokyo, Institute of Oriental Culture. Sarkar, R., Ghosh, S., & Chakraborti, S. (2015). Authorsh...

2024

[1] [1]

Bagchi, P. C. (1939). Studies in the Tantras (Part I). University of Calcutta Press. Bharati, A. (1965). The tantric tradition. Rider and Company. Bose, J. (2026). Darshana-Graph: A knowledge graph of Indian philosophical traditions. arXiv preprint arXiv:2606.18222. Bose, J. (2026). Bengal Dharma Corpus [Dataset]. HuggingFace. https://huggingface.co/datas...

work page internal anchor Pith review Pith/arXiv arXiv 1939

[2] [2]

112 –121)

(pp. 112 –121). ACL Anthology. https://aclanthology.org/2024.nlp4dh- 1.12 Sanderson, A. (2009). The Saiva age: The rise and dominance of Saivism during the early medieval period. In S. Einoo (Ed.), Genesis and development of Tantrism (pp. 41–349). University of Tokyo, Institute of Oriental Culture. Sarkar, R., Ghosh, S., & Chakraborti, S. (2015). Authorsh...

2024