Visualizing Poetic Meter in South Asian Languages

The explication of poetic meter in the modern languages of South Asia is a source of consternation even for experienced poets. Even if poets can competently employ established meters in their work, they and their readers or listeners often cannot describe poetic form using established classificatory schema. Traditional prosodic systems, which are inherited from classical languages, do not align well with the phonological features of modern South Asian languages. Modern scholars have developed alternatives that address these deficiencies. We augment that work by presenting an interactive web-based software package under development to visualize poetic meter using directed graphs that accommodate multiple languages and scripts in order to make poetic knowledge accessible to readers, scholars, and poets.

Background
In South Asia, there are two competing theories of prosody, one derived from Arabic (
‘uruz) and one from Sanskrit (
chhanda). The languages of Urdu and Hindi, which share a common grammar but have differing vocabularies, draw upon both systems of meter. The Arabic system traces its origins to the revelation of the eight century Arab prosodist al-Khalil Ibn Ahmad of Basra. Its basic unit is orthography, specifically the written Arabic letter. Though quite precise for classical Arabic, when used for South Asian languages the system breaks down, because there are so many possible combinations of metrical units, i.e. it is combinatorially explosive. The system used for Sanskrit is also defined by writing, based on abugida (segmental writing system), or long and short units. Modern languages typically drop short vowel endings, however. Both systems have a preference for complex and precise classification. That propensity towards complexity, combined with the disalignment of the systems with modern languages, have posed challenges for modern literary critics, poets, and readers, alike.

Modern prosodists of Urdu and Hindi poetry, often themselves proponents of elegant systems, have attempted to make these prosodies more accessible by referring to patterns of long and short metrical units (Pybus 1924; Pritchett and Khaliq 1987; Fārūqī 1968; Tabassum 1983; Nagasaki and Kim 2012). These are often represented using macrons, breves, or other symbols familiar to readers of English metrical texts.
Modern prosodists are also challenged by the difference between the durational basis for poetic prosody in South Asian languages and the basis of meter in English and other European languages on stress. Durational meter is based on metrical unit length. This difference also poses a problem for graphical exploration of poems, which are mostly designed for English and confine their “prosodic domains” to stress (Abdul-Rahman et al. 2013; Delmonte 2015; McCurdy et al. 2016). Our visualization software addresses both of these challenges.

Graphical visualization of poetic data
Our visualization software consists of a Python and Javascript module in a free and opensource software package that turns poetic texts into poetic data. It provides long and short units, as well as labeled rules for particular contexts. The current implementation supports Hindi and Urdu meters, along with Hindi (Figure 2), Urdu (Figure 3), and English (Figure 4) translations of its interface. The graph layout is performed by the Graphviz library (Gansner and North 2000).

Representing poetic meter as a directed graph
By representing poetic meter as a particular walk through a directed graph, this model offers a significant advancement over previous metrical representations. We assume for our graphical model start and end nodes, short and long metrical unit nodes, and edges between them. We use start and end nodes with distinct shapes and colors. We suggest a circle as the shape for a “short” metrical unit, and a rectangle as that of a “long.” For uncounted metrical units at the end of lines or before caesuras, we suggest a dashed circle. Units that compose metrical feet are grouped into labeled clusters (Figure 1). This system resolves the issues of metrical flexibility and complexity that, in traditional prosody, led to excessive categorization, while visualizing the patterns of durational sound that produce meaning for poets and their listeners.

Discussion
This software package works across the multiple scripts of South Asian reading and listening publics. It advances earlier methods of visualizing meter by affording new sorts of interaction, particularly in web-based environments.
For scholars, directed graphs allow an elegant means to visualize metrical complexity. All of the possible meters of a particular poet can be compared to those of another. The flow through a network also opens new sorts of metrics for comparative analysis.
Further, as we will demonstrate, a walk through a directed graph can be colored in time with a particular audio or visual recording that has been marked up for phoneme timings as well as metrical units, allowing new sorts of insights that are not easily available in text alone.
For listeners who have various levels of knowledge about meter, interactive versions of directed graphs can have instructive qualities. Listeners can learn the rules of prosodic systems by clicking on nodes that represent poetic data.
For poets, this method offers a visual means for composing verse.
While based in Urdu and Hindi, the methodologies described can be easily adapted and applied to a large number of South Asian and other languages to provide renewed access to poetry, conceived as data, in the digital age.

Graphical representation of a sample Urdu/Hindi metrical pattern. Rectangles represent long units, circles short units, and dashed circles uncounted short units. Clusters represent metrical feet, here in the Perso-Arabic system.

Graphical sample of an Urdu verse in the Hindi language interface. Text is rendered in the Devanagari script. Clusters represent metrical feet, here labeled using the traditional Perso-Arabic nomenclature. Word separation is indicated using an interpunct (⋅).

Graphical representation of a sample Urdu verse in the Urdu language interface. Text is rendered in the Perso-Arabic script using Perso-Arabic diacritical markings to show short vowels, which are normally unmarked. Clusters represent metrical feet, here labeled using traditional Perso-Arabic nomenclature.

Graphical representation of a sample Urdu verse in the English language Interface. Text is rendered in scholarly transliteration. Clusters represent metrical feet in traditional Perso-Arabic prosody.

Acknowledgments
This research was supported by an Andrew W. Mellon New Directions Fellowship (grant number 11600613) and by matching funds from the College of Arts and Letters, Michigan State University.

Bibliography
Abdul-Rahman, A.,J. Lein, K. Coles, E. Maguire, M. Meyer, M. Wynne, C. R. Johnson, A. Trefethen, and M. Chen. (2013). Rule-based Visual Mappings –with a case study on poetry visualization.
Computer Graphics Forum, 32: 381–390.

Delmonte, R. (2015). Visualizing poetry with SPARSAR — Visual maps from poetic content.
Proceedings of the Fourth Workshop on Computational Linguistics for Literature, Denver, CO, June 2015.

Fārūqī, S. (1968).
Lafz o ma’nī. Allahabad: Shab-k̲h̲ūn kitāb ghar.

Gansner, E. R. and North, S. C. (2000). An open graph visualization system and its applications to software engineering.
Software-Practice and Experience, 30(11): 1203–1233.

McCurdy, N., Lein, J., Coles, K. and Meyer, M. (2016). Poemage: Visualizing the sonic topology of a poem. 
IEEE transactions on visualization and computer graphics, 
22 (1): 439–448.

Nagasaki, H and Kim, R. I. (eds.) (2012).
Indian and Persian Prosody and Recitation. Delhi: Saujanya Publications.

Pritchett, F. W. and Khaliq, K. A. (1987).
Urdu Meter: A Practical Handbook. Madison, WI: South Asian Studies, University of Wisconsin.

Pybus. G.D. (1924).
A Text-Book of Urdu Prosody and Rhetoric. Lahore: Rama Krishna.

Tabassum, M. (1983).
Āvāz aur ādmī. Hyderabad, India: Iliyās Ṭreḍars.