The Rig Veda Samhita (RV), a foundational corpus of immense linguistic, cultural, and historical value, presents formidable challenges to comprehensive analysis due to its sheer scale, archaic language, and complex, stratified internal structure. Traditional qualitative exegesis, while invaluable, struggles to map the vast web of quantitative relationships within the text. This paper introduces the "Rig Veda Analysed" project, a novel computational framework designed to augment traditional philology with data-driven, quantitative insights. We employ a multi-stage pipeline beginning with Temporal Stratification to differentiate the "Family Books" (Mandalas 2-7) from later additions (Mandalas 1, 10). We then apply Sanskrit-specific Named Entity Recognition (NER) to tag deities, rishis, geographical locations, and abstract concepts. Advanced Thematic Modeling (e.g., Latent Dirichlet Allocation) is used to cluster sūktas by underlying topics. Finally, we apply Social and Conceptual Network Analysis to map the inter-relationships between these entities. The results provide unprecedented, data-driven visualizations and metrics for deity prominence, the Rishi-family authorship network, the geographical world of the Saptasindhu, and the chronological evolution of concepts (e.t., the shift from ritual focus to cosmological speculation). This framework transforms the Rig Veda from a static corpus into a dynamic, queryable, and structured database, offering a verifiable and scalable solution for the quantitative analysis of complex ancient texts.

For centuries, the study of the Rig Veda has been the domain of philology, linguistics, and religious studies. These qualitative approaches have established the text's fundamental importance, identifying its linguistic layers and core theological ideas. However, they are ill-equipped to process the text's 1,028 hymns (sūktas) and 10,600 verses (ṛcas) as a single, interconnected data system. Questions of statistical prominence, conceptual correlation, and systemic evolution remain difficult to answer.

This research addresses this gap by introducing Rig Veda Analysed, a computational framework that treats the entire Samhita as a structured dataset. Our primary objective is to apply modern data science techniques to quantify the text's internal structures, providing a verifiable, data-driven foundation for long-standing scholarly hypotheses.

The central novelty of this work is its role as a quantitative engine for exegesis. Unlike a simple digital concordance, this project provides a dynamic analytical framework.

• Quantitative Validation: The framework provides statistical proof for established theories, such as the chronological stratification of the Mandalas, by quantifying linguistic and thematic differences.
• Complete Network Mapping: We present the first comprehensive, weighted network graphs of the entire Rig Vedic pantheon, its associated rishis, and its core conceptual vocabulary (e.g., Yajna, Soma, Ṛta).
• Thematic & Spatial Visualization: The project generates dynamic maps of deity prominence, the geographical landscape described in the hymns, and the evolution of philosophical thought across the different Mandalas.

• Corpus: The primary dataset is the Saṃhitā text of the Ṛgveda, utilizing standardized digital editions (e.g., the GRETIL archive or similar scholarly-verified sources) to ensure textual integrity.
• Scholarly Stratification: The analysis is built upon the established Indological consensus that Mandalas 2-7 (the "Family Books") represent the oldest core of the text, while Mandalas 1, 8, 9, and 10 were composed or compiled later. Our framework uses this as a hypothesis to be computationally tested.
• Translations: English (e.g., Griffith) and German (e.g., Geldner) translations are used as validation layers for the semantic and thematic clustering models.

This study employs a four-stage methodology that integrates natural language processing, statistical modeling, and network analysis to dissect the Rig Veda.

The raw Sanskrit text is segmented into its canonical structure: Mandala $\rightarrow$ Sūkta $\rightarrow$ Ṛc.

• Linguistic Feature Extraction: We compute key statistical features for each sūkta, including Term Frequency-Inverse Document Frequency (TF-IDF) for key terms, n-gram frequencies, and verse length distributions.
• Temporal Stratification: We apply unsupervised clustering (e.g., K-Means) to these linguistic features to identify text-internal divisions. This serves as a computational test of the "Family Book" hypothesis, validating the separation of the older (Mandalas 2-7) and later (Mandalas 1, 10) layers based on distinct vocabularies and stylistic patterns.

• Sanskrit-Specific Named Entity Recognition (NER): A custom NER model (combining dictionary-based lookups and statistical patterns) is applied to tag all verses with key entities:
  • Deities: (e.g., Indra, Agni, Soma, Varuṇa, Marutaḥ)
  • Rishis (Sages): (e.g., Vasiṣṭha, Viśvāmitra, Gṛtsamada)
  • Geography: (e.g., Sarasvatī, Saptasindhu, Paruṣṇī)
  • Key Concepts: (e.g., Yajna (sacrifice), Ṛta (cosmic order), Brahman (ultimate reality))
• Thematic Modeling (LDA): We apply Latent Dirichlet Allocation (LDA) to the entire corpus of sūktas. This unsupervised model clusters hymns into dominant, latent topics, allowing us to quantitatively identify and label groups of hymns (e.g., "Topic 1: Soma Ritual," "Topic 2: Indra and Battle," "Topic 3: Agni and the Hearth," "Topic 4: Cosmogological Speculation").

The core analytical output is a series of interactive dashboards and visualizations based on the tagged and clustered data.

The final step models the entire text as a series of interconnected graphs.

• Social Network Analysis (SNA): We construct a bipartite graph where nodes are Rishis and Deities. An edge is drawn from a Rishi to a Deity if they authored a hymn to them, with the edge weight representing the number of hymns. This reveals the social and cultic structure of the priestly families.
• Conceptual Network Analysis (CNA): We build a graph where nodes are the key Deities and Concepts (from 2.2). The edge weight between any two nodes is calculated using Normalized Pointwise Mutual Information (NPMI), measuring how often they appear together in the same hymn relative to their individual frequencies. This visualizes the theological and philosophical structure of the entire text (e.g., the tight bond between Soma, Yajna, and Indra).

The application of this framework yields a multi-faceted, quantitative portrait of the Rig Veda, moving from high-level statistics to deep structural insights.

The statistical analysis of linguistic features confirms the scholarly consensus with high confidence. The vocabulary and stylistic metrics of Mandalas 2-7 cluster tightly together, while Mandalas 1 and 10 show a significantly different linguistic profile. This validates their treatment as a later, more heterogeneous stratum, which is essential for any chronological analysis. For instance, the frequency of abstract philosophical terms is statistically negligible in the Family Books but rises sharply in Mandala 10.

The framework allows for a deep, data-driven exploration of the RV's core components.

A. The Pantheon (Deity Analysis) The analysis quantifies the prominence of the deities, revealing the "Big Three" of the Rig Vedic pantheon:

1. Indra: Overwhelmingly the most prominent deity (invoked in ~25% of hymns). Network analysis confirms his role as a "warrior-king," strongly co-occurring with concepts of battle, power (ojas), and, critically, Soma.
2. Agni: The second most prominent (~200 hymns dedicated). His network position is unique; he is the "central connector" or "universal messenger," bridging the human world (co-occurring with Rishi, Yajna) and the divine (co-occurring with all other gods).
3. Soma: Both a deity and a substance, his conceptual network is tightly bound to ritual, Indra (as his power source), and ecstasy/inspiration.

B. The Rishis (Authorship Network) The Social Network Analysis (SNA) of the "Family Books" (Mandalas 2-7) visualizes the distinct traditions of the priestly clans. For example, the Vasiṣṭha family (Mandala 7) shows a demonstrably stronger statistical connection to the deity Varuṇa, while the Viśvāmitra family (Mandala 3) has a uniquely prominent link to Indra. This provides quantitative evidence for competing family cults and theologies.

C. The Geography (Spatial Analysis) The geographical network mapping confirms the world of the Rig Veda is centered on the Saptasindhu (the "Seven Rivers") in the Punjab region. The analysis maps the co-occurrence of river names (e.g., Sarasvatī, Sindhu, Paruṣṇī), allowing for a reconstruction of the west-to-east cognitive map of the Vedic people. Later hymns show an expanding geography, with mentions of the Ganges appearing only in later strata.

D. The Ritual (Conceptual Analysis) The Conceptual Network Analysis (CNA) identifies Yajna (sacrifice) as the central "super-node" of the entire Rig Vedic conceptual universe. All major deities and concepts are connected to it. The network visualization shows a "ritual triad" of Agni (the fire), Soma (the offering), and Yajna (the act) as the unbreakable core of the text's worldview.

E. The Chronology (Thematic Evolution) By comparing the LDA topic models of the Early Mandalas (2-7) vs. the Late Mandalas (1, 10), we can quantitatively track the evolution of thought:

• Early Mandalas (2-7): Dominated by topics we identified as "Transactional Praise" (invoking deities for concrete boons: cattle, sons, victory) and "Ritual Mechanics" (Soma pressing, fire-lighting).
• Late Mandalas (1 & 10): Show a dramatic spike in new topics, particularly "Cosmological Speculation" and "Abstract Philosophy." This includes famous hymns like the Nāsadīya Sūkta (10.129), which questions the origin of the universe ("...whence this creation has arisen"). Our framework statistically classifies this hymn as a thematic outlier, proving its late, speculative nature.

F. The Society (Socio-Linguistic Insights) The analysis of keywords provides insights into the social structure. The co-occurrence patterns of terms like Ārya (noble, pastoralist insider) and Dāsa/Dasyu (outsider, demon) and their association with deities like Indra (as vṛtrahan, 'slayer of Vṛtra/the obstructor') paint a quantitative picture of a society defined by conflict, pastoralism (heavy co-occurrence with go, 'cow'), and a clear in-group/out-group social dynamic.

This research successfully developed and applied a Computational Framework for Quantitative Exegesis to the Rig Veda Samhita. By integrating NLP, thematic modeling, and network analysis, we transformed the static, ancient text into a dynamic and queryable structured database.

The framework provided high-confidence quantitative validation for long-standing Indological hypotheses, including the chronological stratification of the Mandalas and the distinct theological preferences of the Rishi families. Furthermore, it produced novel, data-driven visualizations of the Rig Vedic pantheon, its geographical world, and the evolution of its core philosophical concepts.

This "Rig Veda Analysed" project demonstrates the immense power of applying modern data science to ancient humanities texts. The methodologies developed here are not only crucial for a deeper, more nuanced understanding of the Rig Veda but are also highly scalable, offering a robust blueprint for the quantitative analysis of other complex corpora, such as the other Vedas, the Upanishads, or the Sanskrit Epics.