Sangeetha-Grantha

Metadata	Value
Status	Active
Version	1.1.0
Last Updated	2026-09-10
Author	Sangeetha Grantha Team
Document Type	Evidence record

Data Quality Audit Results (Interim) - Feb 2026

---

[!NOTE] Historical evidence: results, counts, commands, and observations below belong to the original work described here. The editorial update date is not a new test or corpus verification. For present behavior, use current quality checks.

1. Executive Summary

The initial audit phase has identified significant structural inconsistencies in the Krithi repertoire. While the musicological hypothesis states that Krithi structure should be invariant across languages, the database contains variants with widely diverging section counts.

2. Key Findings

Structural Variance

A significant number of Krithis exhibit “Section Count Drift” across language variants.

Krithi ID	Title	Language	Section Count
87996ef9	aruNAcala nAtham	sa	3 or 5
87996ef9	aruNAcala nAtham	te	2 or 4
bd4cf402	Aryaam Abhayaambaam	sa	3 or 6
91079b86	bAlAmbikayA kaTAkshitOaham	te	1 or 2

Case Study: bAlAmbikayA kaTAkshitOaham (91079b86)

Analysis of the krithi_lyric_sections for this krithi reveals several data quality anti-patterns:

1. Section Inconsistency:
   • Some Malayalam (ml) and Kannada (kn) variants are split into 3 sections (Pallavi, Samashti Charanam 1, Samashti Charanam 2).
   • Other variants for the same languages merge all content into fewer sections.
2. Metadata Pollution: The text field in some lyric sections contains non-lyric metadata such as “Updated on 16 Dec 2016”, “Kannada- Word Division”, or translation notes like “tri-kUTa - refer to lalitA sahasra nAma 89”.
3. Redundant Variants: There are multiple variants for the same language (e.g., 3 English variants, 2 Malayalam) that are nearly identical but have different section mappings.

4. Analysis of Authoritative Sources (Feb 2026)

Analysis of PDFs and specialized composer sites (Guruguha, Swathi Thirunal Festival) confirms that:

• Dikshitar Compositions: Standardized using technical markers like (madhyamakālasāhityam). Failure to detect these results in Madhyama Kala being incorrectly merged into the preceding Charanam/Anupallavi.
• Swathi Thirunal: Highly structured with multiple numbered Charanams.
• Transliteration Precision: Sanskrit and Telugu sources provide the most accurate syllabic mapping. Tamil sources often use subscripts (1-4) which our current parser may not fully normalize.

5. Proposed Sourcing Hierarchy & Logic

To ensure structural integrity, the ImportService should adopt a “Composer-to-Authority” mapping:

Composer	Primary Authority Source
Muthuswami Dikshitar	guruguha.org
Tyagaraja	thyagarajavaibhavam.blogspot.com
Swathi Thirunal	swathithirunalfestival.org
General / Others	karnatik.com, shivkumar.org

Logic Firm-up (Heuristics):

1. Structural Voting: If multiple sources are available, the structure with the highest section count (that matches known composer patterns) is chosen as the template.
2. Technical Labeling: TextBlocker.kt must be extended to recognize madhyama kala, chittaswaram, and muktayi swaram as first-class sections, even when parenthesized.
3. Ghost Section Detection: If a lyric variant has a significantly higher line count than its siblings but fewer sections, it is flagged for manual “Section Split” review.

6. Remediation Roadmap

1. Pilot Fix: Apply normalization logic to high-priority Dikshitar Krithis (e.g., Navavarnas).
2. Metadata Purge: Automated script to remove “Updated on…”, “Meaning:”, and “Word-by-word” blocks from the text field.
3. Re-Extraction: Re-run TextBlocker with the enhanced strategy for variants flagged with low section counts.

---

Section index · Documentation home · Feature status