Echoes Across Centuries: Phonetic Signatures of Persian Poets
Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar · 2026
Read Echoes Across Centuries: Phonetic Signatures of Persian Poets by Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-14443.txt.
Introduction
Shahnazari, Ayyoubzadeh and Keshtparvar measure sound patterns in Persian poetry across centuries. Their corpus holds over 1.1 million lines from nearly 32,000 poems by 83 poets. They limit the set to five classic meters for fair comparison. Each line is turned into phonemes and scored on six traits. These traits cover hardness, sonority, sibilance, vowel share, entropy and consonant clusters. The authors find that poets keep distinct sound profiles even after meter and form are controlled. Sound shifts over time track genre and performance shifts, not sharp breaks.
Catalog record
| CREATOR | Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95544 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.14443
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Echoes Across Centuries: Phonetic Signatures of Persian Poets Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar Abstract This study examines phonetic texture in Persian poetry as a literary-historical problem rather than as a by-product of meter or an auxiliary feature for classification. The corpus comprises 1,116,306 mesras from 31,988 poems and 83 poets drawn from a larger archive after restricting inference to five canonically named classical meters in the retained analytical subset and to well-supported poet-meter cells. Each mesra is represented as a grapheme-to-phoneme symbol stream and summarized through six measures: hardness, sonority, sibilance, vowel ratio, phoneme entropy, and consonant-cluster ratio. The argument combines corpus description, confound-aware modeling, and literary interpretation. Poet contrasts are estimated under controls for meter, poetic form, and line-length opportunity; century trends are interpreted as historical redistribution rather than rupture; and meter is treated as a prosodic institution that shapes which phonetic differences can stabilize within a line. Meter and form explain substantial variance, yet they do not absorb poet-level differentiation. Persian poetic sound is therefore best understood as conditioned variation within shared formal regimes rather than as either pure authorial essence or mere metrical residue. The literary implications emerge most clearly in the stylistic map and case studies. The phonetic space separates high-sonority lyrical profiles, hardness-led rhetorical or epic profiles, sibilant mystical and courtly contours, and high-entropy experimental textures. Century analysis links early firmness to epic and courtly baselines, later sonority to the prominence of ghazal and mystical lyric, and sixteenth- to nineteenth-century reconfiguration to Indo-Persian complexity, revivalist rhetoric, devotional address, and changing performance contexts. Case studies of Safi Alishah, Shah Nematollah Vali, Bidel Dehlavi, Ashofteh Shirazi, Fakhr Al-Din Asad Gorgani, Ferdowsi, Saadi, Khaqani, Jahan Malek Khatun, Bolandeghbal, Hafez, Shahriar, and Khayyam show that familiar literary labels such as firmness, luminosity, difficulty, and fluency correspond to different phonetic mechanisms in different historical settings. Taken together, the results provide a corpus-scale phonetic framework for Persian poetry. They estimate poet differentiation under explicit formal controls while keeping meter legible as classical aruz , model stylistic clustering and historical redistribution beyond a single hard-soft continuum, and show how computational phonetics can sharpen literary-historical interpretation without displacing it. keywords digital humanities, Persian poetry, computational poetics, historical poetics, corpus stylistics † † equal-contributors: These authors contributed equally to this work. † † equal-contributors: These authors contributed equally to this work. 1 Introduction Sound has long been one of the most durable critical languages of Persian literary culture. Epic verse is heard as firm and declarative; courtly qasida as sharpened and brilliant; ghazal as open, memorable, and singable; mystical poetry as fluid, repetitive, and incantatory; later Indo-Persian writing as dense and internally restless; modern lyric as more conversationally spacious. These descriptors are not incidental ornaments. They are shorthand for how poetic language distributes consonantal pressure, vocalic openness, strident contour, and rhythmic closure. Yet they are difficult to compare beyond a limited set of canonical poems. Close reading can hear the sonic pressure of a hemistich, but not easily decide whether that pressure belongs to one line, one meter, one poet, or a broader historical formation. Persian makes the problem especially demanding. The script ordinarily omits short vowels, so purely graphemic comparison is insufficient for a study of poetic sound. At the same time, meter cannot be treated as a weak background variable. In Persian verse, aruz is a constitutive institution: it shapes diction, line architecture, and the range of sonic expectations available to poets. Here meter means a named bahr : a quantitative pattern of long ( - ) and short ( \cup ) syllables distributed across the mesra and realized through historically accepted sub-patterns and zehafat . Poetic form is a different category, referring instead to poem-level organizations such as ghazal , masnavi , qasida , or rubai . A serious corpus study of sound must therefore do two things at once. It must reconstruct a stable phonetic surface from the written text, and it must distinguish poet-level variation from the variation made available by meter, form, and line length. That double requirement turns phonetic analysis into both a methodological problem and a humanities problem. Recent Persian digital humanities has supplied much of the necessary infrastructure. Large literary corpora, poet-level stylometry, meter recognition, and school detection have made comparative work increasingly feasible. What remains underdeveloped is phonetic texture as an interpretive object in its own right. Sound features often appear inside hybrid authorship or classification models, but they are rarely treated as evidence for literary-historical comparison, and even more rarely under explicit control for the formal structures that organize Persian verse. This gap matters because some of the oldest judgments in Persian criticism are precisely sonic judgments. If computation cannot address them responsibly, it leaves a central dimension of literary form understudied. The central questions are whether poet-level phonetic differences remain measurable after controlling for meter, poetic form, and line-length opportunity; how those differences are organized in a multidimensional stylistic field rather than along a single hard-soft axis; what century-by-century redistributions become visible when phonetic measures are read against Persian literary history; and how such quantitative profiles can be translated into historically credible readings of individual poets without collapsing literary value into sound alone. To address them, the analysis combines corpus restriction for comparability, symbolic phonetic representation at mesra scale, confound-aware statistical models, century interpretation, and case studies that return the measurements to literary history. The active inferential cohort comprises 1,116,306 mesras from 31,988 poems and 83 poets. Ten of the main case-study poets fall inside that balanced cohort; Ferdowsi serves as a high-support epic reference; Shahriar and Khayyam are retained as descriptive extensions because their historical importance exceeds their fit with the strongest inferential frame. This design allows rigorous comparison while keeping evidential boundaries explicit. Four features are central: 1. It develops a corpus-scale phonetic analysis of Persian poetry at mesra resolution, using a grapheme-to-phoneme representation suitable for large literary comparison. 2. It estimates poet differences under explicit controls for meter, poetic form, and line-length opportunity, while translating internal meter codes into historically legible classical aruz names. 3. It models Persian poetic sound as a multidimensional stylistic field organized around openness, firmness, contour, and entropy, and reads that field across centuries and across named metrical environments rather than as a single hierarchy of hardness or softness. 4. It demonstrates, through historically grounded case studies, how phonetic metrics can sharpen interpretations of epic, ghazal, mystical verse, courtly rhetoric, revivalist poetics, and modern lyric without reducing poetry to phonetics alone. 2 Related Work Research on Persian poetry in digital humanities is now rich enough that the main question is no longer whether large-scale comparison is possible, but what kind of literary object each computational design actually produces. The present study belongs at the intersection of corpus infrastructure, stylometry and prosody, phonetic representation, and sound-centered digital humanities, but it addresses a gap that those strands have not yet resolved together. 2.1 Persian Corpora and Literary Modeling Corpus construction has been the enabling condition of recent Persian DH. Early work by Asgari and Chappelier showed that Persian poetic corpora could sustain large-scale analysis of literary tendencies rather than only isolated text-processing tasks ( 2 ) . That infrastructural turn was extended decisively by Raji