A Dynamic Atlas of Persian Poetic Symbolism: Families, Fields, and the Historical Rewiring of Meaning

Kourosh Shahnazari; Seyed Moein Ayyoubzadeh; Mohammadali Keshtparvar

cs.CLApr 1, 2026

A Dynamic Atlas of Persian Poetic Symbolism: Families, Fields, and the Historical Rewiring of Meaning

Authors: Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar

Organizations: Sharif University of Technology, Tehran, Iran. · Amirkabir University of Technology, Tehran, Iran.

Abstract

Persian poetry is often remembered through recurrent symbols before it is remembered through plot. Wine vessels, gardens, flames, sacred titles, bodily beauty, and courtly names return across centuries, yet computational work still tends to flatten this material into isolated words or broad document semantics. That misses a practical unit of organization in Persian poetics: related forms travel as families and gain force through recurring relations. Using a corpus of 129,451 poems, we consolidate recurrent forms into traceable families, separate imagistic material from sacred and courtly reference, and map their relations in a multi-layer graph. The symbolic core is relatively sparse, the referential component much denser, and the attachment zone between them selective rather than diffuse. Across 11 Hijri-century bins, some families remain widely distributed, especially Shab (Night), Ruz (Day), and Khaak (Earth). Wine vessels, garden space, flame, and lyric sound strengthen later, while prestige-coded and heroic-courtly vocabulary is weighted earlier. Century-specific graphs show change in arrangement as well as membership. Modularity rises, cross-scope linkage declines, courtly bridges weaken, and sacred bridges strengthen. Hub positions shift too: Kherqe (Sufi Robe) gains late prominence, Farkhondeh {Blessed} and Banafsheh (Violet) recede, and Saaghar (Wine Cup) stays central across the chronology. In this corpus, Persian symbolism appears less as a fixed repertory than as a long-lived system whose internal weights and connections change over time.

Figures & tables

Measure	Value
Poems	129,451
Verses	1,446,226
Hemistichs	2,891,826
Poets	216
Century bins	11
Raw symbolic tokens consolidated	1,160

Table 1: Corpus and symbolic-inventory summary.

Panel of Figure 1 — Figure 1: Analytical workflow from corpus alignment to dynamic graph analysis.

Diagnostic	Value	Summary
Raw symbolic tokens $\rightarrow$ canonical families	$1{,}160\rightarrow 738$	Family induction removes $36.4\%$ of analytic scatter while preserving token-level traceability.
Median family size / single-member share	$1$ / $83.7\%$	Consolidation is conservative rather than indiscriminate; most forms are left untouched unless evidence is strong.
Top-250 raw candidates $\rightarrow$ distinct families / clear fragment reduction	$250\rightarrow 183$ / $98.65\%$	The same window loses $67$ duplicate raw positions and absorbs $74$ clear fragment-like forms into stronger heads, producing cleaner rankings.
Seed-family recovery	$18/18$	The manually seeded symbolic anchors are all recovered in the induced family inventory.
Referential leakage into top 40 symbolic-core families	$0$	The induced core stays imagistic rather than being overrun by sacred or courtly name-like material.

Table 2: Compact validation of family-level modeling.

Layer	Nodes in layer	Edges	Avg. degree	Density	Role
Imagistic core	127	223	3.51	0.028	Recurrent symbolic fields built from floral, bodily, ritual, elemental, and spatial imagery
Referential layer	30	81	5.40	0.186	Sacred and courtly names treated as meaningful reference rather than folded into imagistic symbolism
Bridge layer	63	143	4.54	0.073	Attachment zone linking referential materials to symbolic neighborhoods
Note: The bridge row is a participation count rather than an additional disjoint node inventory. The unique atlas total in Table 1 reports the 157 retained families assigned primarily to the core or referential residencies; the bridge count records how many of those same retained families also participate in cross-layer attachment edges.

Table 3: Structural summary of the symbolic graph.

Figure 5

Figure 6

Figure 7

Figure 8

Family	Trajectory	Peak	Peak-mode share	Field	Pattern
Peymaaneh (Wine Vessel)	Robust rising	12	76%	Wine-tavern ritual	Late strengthening of vessel imagery within lyric-tavern clusters
Sahbaa (Wine)	Robust rising	14	78%	Wine-tavern ritual	Sustained late growth of wine nomenclature beyond a single century spike
Golshan (Garden)	Robust rising	12	81%	Floral-vegetal	Expansion of garden space as a late symbolic environment
Khaaqaan (Sovereign)	Robust declining	6	63%	Referential / courtly	Recession of prestige-coded royal diction after early prominence
Farrokh (Auspicious)	Robust declining	5	93%	Mixed field	Weakening of auspicious courtly language inside the broader contact zone
Baad (Wind)	Robust declining	6	67%	Mixed field	Earlier-heavy mobility and force imagery loses weight historically

Table 4: Representative family trajectories after temporal refinement. Peak-mode share reports the proportion of

100

poet-bootstrap resamples that recover the same modal peak century.

Figure 10

Figure 11

Figure 12

Metric	Base change	Uncertainty summary	Variant support	Interpretation
Core modularity (late minus early)	$+0.020$	Comparable-variant span $+0.012$ to $+0.030$ ; permutation upper-tail $0.14$	Positive in $4/5$ variants	Small but fairly consistent separation among fields.
Cross-link ratio (late minus early)	$-0.032$	Comparable-variant span $-0.032$ to $-0.019$ ; permutation lower-tail $0.05$	Negative in $4/5$ variants	Cross-scope links decline across most reconstructions.
Courtly bridge strength (late minus early)	$-0.033$	Variant span $-0.033$ to $-0.007$	Negative in $2/4$ , flat in $2/4$	Courtly weakening is visible, though it depends more on threshold choice.
Sacred bridge strength (late minus early)	$+0.049$	Variant span $+0.042$ to $+0.071$	Positive in $4/4$ variants	Sacred-prophetic attachment strengthens late with the strongest support in this table.
Degree centralization (late minus early)	$-0.001$	Comparable-variant span $-0.002$ to $-0.001$ ; high-confidence subset $+0.037$	Flat in $4/5$ variants	Centralization contributes little to the main historical contrast.

Table 5: Dynamic network changes with robustness and uncertainty summary.

Figure 14

Figure 15

Figure 16

Figure 17

Field	Leading poet	Rate / $10{,}000$ verses	Characteristic family heads
Wine-Tavern Ritual	Saeb (11th c.)	3250.2	Mey (Wine); Baadeh (Wine); Sharaab (Wine)
Ascetic-Mystical	Ashofteh Shirazi (13th c.)	604.7	Meykhaaneh (Wine House); Meykadeh (Wine House); Moghaan (Magi)
Body-Beloved	Saeb (11th c.)	1624.6	Cheshm (Eye); Lab (Lip); La’l (Ruby)
Floral-Vegetal	Jahan Malek Khatun (8th c.)	1770.9	Sarv (Cypress); Gol (Rose); Qad (Stature)
Light-Fire	Ahli Shirazi (10th c.)	692.2	Aatash (Fire); Nur (Light); Zolmat (Darkness)
Water-Sea	Saeb (11th c.)	771.0	Daryaa (Sea); Mowj (Wave); Bahr (Sea)

Table 6: Leading poet-level field affinities in the threshold-based core poet pool. Rates report field occurrences per

10{,}000

verses among mapped poets with at least

10{,}000

verses and

1{,}000

symbolic-family occurrences.

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Stage	Positive evidence	Penalties and gates	Output rule
Family induction	Eligible surface relation; static and occurrence-level embeddings; contextual and cluster overlap; seed and exemplar support; head cleanliness	Fragment residue, function-word contamination, genericity noise, and proper-name inflation lower the merge score; staged thresholds keep weak cases as singletons	Each token maps to one family; accepted heads default to the cleanest root form, with synthetic heads admitted only when several strong variants jointly clear the synthetic threshold
Edge construction	Multiview embedding support; weighted context overlap; cluster overlap; representative-context overlap; verse/document co-association; poet overlap; seed support; local co-activation	Genericity is penalized in every layer; referentiality is negative in the core, near-neutral in the referential layer, and slightly positive in the bridge layer; layer thresholds and top- $k$ sparsification prune generic neighborhoods more aggressively	Each retained family keeps one primary atlas residency, while bridge participation records which retained families also sustain cross-layer attachment edges

Appendix

Table 7: Compact scoring summary for family induction and edge construction.

Target	Point estimate	Uncertainty summary	Summary
Peymaaneh (Wine Vessel)	Slope $+2.51$	$95\%$ bootstrap CI $[+1.37,+3.40]$ ; peak-share $76\%$	Late strengthening remains stable under poet-bootstrap resampling.
Sahbaa (Wine)	Slope $+0.60$	$95\%$ bootstrap CI $[+0.29,+0.92]$ ; peak-share $78\%$	The rise is smaller than Peymaaneh (Wine Vessel) but remains consistent.
Golshan (Garden)	Slope $+5.06$	$95\%$ bootstrap CI $[+3.53,+6.82]$ ; peak-share $81\%$	Garden-space intensification is one of the clearest late family gains.
Khaaqaan (Sovereign)	Slope $-1.74$	$95\%$ bootstrap CI $[-2.92,-0.14]$ ; peak-share $63\%$	Early concentration is consistent, even if the exact peak century is not knife-edged.
Farrokh (Auspicious)	Slope $-2.06$	$95\%$ bootstrap CI $[-3.06,-0.39]$ ; peak-share $93\%$	Prestige-coded decline is unusually stable among the declining families.
Baad (Wind)	Slope $-19.31$	$95\%$ bootstrap CI $[-27.97,-11.27]$ ; peak-share $67\%$	Mobility imagery is earlier-weighted with a strongly negative slope.

Appendix

Table 8: Supplementary uncertainty summary for representative families and dynamic claims.

Family	Atlas role	Temporal profile	Note
\persianfont ساغر / Saaghar (Wine Cup)	Core / Wine-Tavern Ritual	Rising	A vessel image that remains central even as late lyric clustering intensifies.
\persianfont پیمانه / Peymaaneh (Wine Vessel)	Core / Wine-Tavern Ritual	Rising	A measure-and-intoxication family that strengthens with later lyric concentration.
\persianfont گلشن / Golshan (Garden)	Core / Floral-Vegetal	Rising	Garden-space imagery expands while remaining lexically transparent at the family level.
\persianfont خرقه / Kherqe (Sufi Robe)	Core / Ascetic-Mystical	Rising	A late-strengthening ritual family that becomes structurally prominent within the mystical field.
\persianfont خرابات / Kharaabaat (Tavern)	Core / Ascetic-Mystical	Peak Middle	A tavern family marking the field’s recurrent traffic with wine imagery.
\persianfont بنفشه / Banafsheh (Violet)	Core / Body-Beloved	Declining	A beloved-image family weighted more strongly toward earlier and middle centuries.

Appendix

Table 9: Supplementary inventory of representative symbolic families.

Field	Illustrative families	Mean active-member ratio	Summary
Wine-Tavern Ritual	Saaghar (Wine Cup); Baadeh (Wine); Qadah (Goblet)	0.990	The most persistent large field in the atlas, with low split pressure and strong late continuity.
Floral-Vegetal	Gol (Rose); Golshan (Garden); Sarv (Cypress)	0.965	The field remains active throughout the chronology but shows the strongest internal differentiation.
Body-Beloved	Banafsheh (Violet); Sonbol (Hyacinth); Meshkin (Musky)	0.960	Bodily beauty imagery persists strongly while repeatedly leaning toward the floral field.
Ascetic-Mystical	Kherqe (Sufi Robe); Meykadeh (Wine House); Kharaabaat (Tavern)	0.857	The most visibly intermediary field, with recurrent outward pull toward wine-tavern ritual.
Water-Sea	Daryaa (Sea); Bahr (Sea); Qolzum (Sea)	0.980	A highly continuous elemental field with little internal fragmentation.
Light-Fire	Aatash (Fire); Forugh (Glow); Partow (Radiance)	0.985	An unusually stable field whose members retain a common intensity profile across bins.

Appendix

Table 10: Supplementary field inventory and structural tendencies.

Figure 24

Poet	C.	Verses	Sym.	WT	AM	BB	FV	LF	WS	MF
Amir Khosrow Dehlavi	7	21,599	19,843	1714.9	170.8	1387.6	885.7	333.3	153.7	4541.0
Jahan Malek Khatun	8	12,276	12,720	165.4	31.8	1376.7	1770.9	432.6	127.1	6457.3
Khwaju Kermani	8	29,301	19,768	1128.3	158.4	1079.5	771.0	414.3	297.9	2897.2
Salman Savaji	8	13,129	10,573	844.7	207.9	1352.7	1132.6	401.4	320.7	3793.1
Seyf Farghani	8	10,279	8,205	803.6	74.9	1192.7	917.4	583.7	180.0	4230.0
Jami	9	39,753	26,396	1129.7	218.3	999.2	739.3	441.0	203.8	2908.7

Appendix

Table 11: Verse-normalized poet-field profiles for the threshold-based core poet pool. The pool retains poets with at least 10,000 verses and 1,000 symbolic-family occurrences. WT = Wine-Tavern, AM = Ascetic-Mystical, BB = Body-Beloved, FV = Floral-Vegetal, LF = Light-Fire, WS = Water-Sea, MF = Mixed Field. Cluster assignments are reported separately in Table 12 .

Poet	Cluster	Nearest neighbor	Second neighbor	Mean-dist. rank
Amir Khosrow Dehlavi	Lyric-amplifying	Salman Savaji	Jami	31
Jahan Malek Khatun	Lyric-amplifying	Seyf Farghani	Salman Savaji	35
Khwaju Kermani	Lyric-amplifying	Jami	Ibn Yamin	9
Salman Savaji	Lyric-amplifying	Khwaju Kermani	Jami	24
Seyf Farghani	Lyric-amplifying	Feyz Kashani	Ibn Yamin	22
Jami	Lyric-amplifying	Khwaju Kermani	Ibn Yamin	12

Appendix

Table 12: Cluster membership and nearest symbolic neighbors in standardized field space. Smaller mean-distance ranks indicate more central positions, while larger ranks indicate more outlying profiles.

Figure 27

Semantic Change Detection

Semantic change detection identifies how meanings, concepts, or scene contents differ across time in text, images, and other data.

See all 62 papers

Between Century and Poet: Graph-Based Lexical Semantic Change in Persian Poetry
Apr 8, 2026Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali KeshtparvarSemantic Change Detection Computational Literary Studies
Sharif University of Technology · Amirkabir University of Technology
Echoes Across Centuries: Phonetic Signatures of Persian Poets
Mar 15, 2026Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar
Sharif University of Technology, Tehran, Iran. · Amirkabir University of Technology, Tehran, Iran.
Eigenmood Space: Uncertainty-Aware Spectral Graph Analysis of Psychological Patterns in Classical Persian Poetry
Feb 18, 2026Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali KeshtparvarSpectral Graph Embedding Computational Literary Studies
Sharif University of Technology, Tehran, Iran. · Amirkabir University of Technology, Tehran, Iran.

Apr 8, 2026J/K move · Enter open · S save

Between Century and Poet: Graph-Based Lexical Semantic Change in Persian Poetry

Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar

Sharif University of Technology · Amirkabir University of Technology

Semantic Change Detection Computational Literary Studies

Open paper AskPDF arXiv

Meaning in Persian poetry is both historical and relational. Words persist through literary tradition while shifting their force through changing constellations of neighbors, rhetorical frames, and poetic voices. This study examines that process using aligned Word2Vec spaces combined with graph-based neighborhood analysis across centuries and major poets. Rather than modeling semantic change as vector displacement alone, it treats lexical history as the rewiring of local semantic graphs: the gain and loss of neighbors, shifts in bridge roles, and movement across communities. The analysis centers on twenty target words, anchored by five recurrent reference terms: Earth, Night, two wine terms, and Heart. Surrounding them are affective, courtly, elemental, and Sufi concepts such as Love, Sorrow, Dervish, King, Annihilation, and Truth. These words exhibit distinct patterns of change. Night is more time-sensitive, Earth more poet-sensitive, and Heart shows continuity despite graph-role mobility. The two wine terms highlight probe sensitivity: one is broad and semantically diffuse, while the other is narrower and more stable. A lexical audit confirms that the corpus contains historically driven terms, poet-specific usages, and sparsely attested mystical vocabulary requiring caution. Overall, semantic change in Persian poetry is better captured as neighborhood rewiring than as abstract drift. For Digital Humanities, this approach restores local structure to computational analysis and supports interpretations closer to literary practice: persistence, migration, mediation, and selective transformation.

Abstract

Semantic Change Detection

Explore similar work

Between Century and Poet: Graph-Based Lexical Semantic Change in Persian Poetry

Echoes Across Centuries: Phonetic Signatures of Persian Poets

Eigenmood Space: Uncertainty-Aware Spectral Graph Analysis of Psychological Patterns in Classical Persian Poetry