Collation Data of Deewan-e-Mah Laqa Chanda across four Sources

This dataset presents a comprehensive computational collation of the Diwan-e-Chanda, the collected poetry of Mah Laqa Bai Chanda (c. 1768–1824). For more than a century, scholarship has relied on the 1906 lithograph Gulzar-e-Mah Laqa as the textus receptus. A substantial body of modern criticism has posited that the author intentionally revised her native Dakhni diction to conform to northern Urdu literary standards. However, this dataset offers empirical, codicological evidence that such linguistic normalization was not authorial but rather the result of posthumous editorial intervention. The collation utilizes a computational Lachmannian methodology applied to the four extant witnesses of the text: 1. BL 1799 (British Library, IO Islamic 2768): This presentation copy was gifted to John Malcolm, the British Resident of Hyderabad. It contains 97 ghazals, correcting the historically reported figure of 118, and includes an unpublished Persian dibācha by Syed Naseeruddin Khan Qudrat. 2. TO 1811 (Telangana Oriental Manuscript Library, MS 228): This circle copy was produced within Hyderabad's literary milieu and also contains the dibācha. Together, BL 1799 and TO 1811 constitute the early recension of the Diwan. 3. SJ 1818 (Salar Jung Museum, Hyderabad, His. 364): This manuscript is appended to the Tārīkh-e-Dil-Afrōz and represents the final document prepared under authorial supervision. It serves as the base text for the accompanying critical edition and contains 125 ghazals, correcting the historically reported figure of 123. 4. GM 1906 (Gulzar-e-Mah Laqa, Nizam ul Matabe, Hyderabad): This is the posthumous lithograph print edition, edited by Ghulam Samdani Khan Gauhar. Analysis of shared indicative errors demonstrates that GM 1906 is a direct, corrupted derivative of SJ 1818. The three manuscript witnesses confirm that the text remained consistently and fully Dakhni throughout nineteen years of documented authorial activity. This finding fundamentally challenges the authority of the 1906 textus receptus and the historiographical tradition based upon it. Designed for computational philology and digital humanities research, this dataset functions as a discrete character matrix suitable for phylogenetic analysis (e.g., via SplitsTree4 using the NeighborNet algorithm). By mapping shared errors and variant distributions across 125 ghazals, it provides a replicable basis for the stemmatic and statistical analyses reported in the companion paper (DOI: [paper DOI once assigned]). The matrix is also structured to facilitate the systematic study of linguistic attrition and dialect normalization. It tracks the deliberate suppression and alteration of native Dakhni grammatical, phonetic, and lexical features by the 1906 editor, enabling quantitative study via metrics such as the Dakhni Retention Rate (DRR) introduced in the companion paper. Emendation and Bracket Policy The base text (SJ 1818) is emended conservatively: alterations are made only in cases of clear mechanical disruption, scribal slips, or physical damage to the manuscript. Where words or syllables are missing due to paper deterioration or omission, they are supplied from the alternative witnesses (BL 1799 or TO 1811). All such editorial interventions are marked with square brackets — e.g., [لفظ] — so that the editorial layer remains fully transparent and separable from the copyist layer. Researchers running computational analyses have complete control over how bracketed content is handled: To treat editorial interventions as explicit character deviations (recommended for strict phylogenetic analysis): run algorithms on the raw dataset as-is. String-matching and distance functions will naturally recognize [لفظ] as distinct from لفظ, flagging the supplied text as a deviation. To analyze only the raw copyist layer (recommended for baseline text metrics), strip the brackets programmatically before processing. A simple regex removing [ and ] while retaining the enclosed text merges the supplied content into the baseline string without treating it as a variant.