Abstract

With 60 million speakers in more than 20 countries, Urdu is a widely spoken language, especially in South Asia. It has a rich tradition of poetry and prose. Traditionally, Urdu has been written in Nastalique script. This script is cursive having a complex and context-sensitive structure. Though it is defined by well-formed rules passed down through generations of calligraphers, these rules have not been quantitatively examined and published in enough detail to enable modeling of character-based Nastalique font for computers. In addition, prevalent font specifications, e.g. True Type font, are not mathematically powerful enough to model such complex fonts. Therefore, there is no practical way to post or exchange information in Urdu through electronic media including internet (except putting scanned images of text, which makes web sites extremely slow to access). Recent advances in font technology now enable complex scripts like Nastalique to be modeled. Open Type Font (OTF) specification is one such example. This project aims to perform a quantitative analysis of Nastalique rules, and model them using OTF. On successful completion of the project, it will become easy to disseminate information in Urdu language through electronic media. As OTF is a standard formalism, no specialized software will be required to read and render this font, as it will easily be operable within the existing software supporting OTF. Output of this project will be a character-based Nastalique font for Urdu. Furthermore, the project will also produce research papers quantifying Nastalique rules to significant detail and analyzing methods for modeling and rendering complex fonts.

Background and Justification

Urdu is the national language of Pakistan and has more than 60 million speakers in more than 20 countries. Even with such extensive readership, very limited information is published on internet in Urdu. A significant limiting factor is absence of a character-based font for Urdu. Urdu is written in Nastalique script which is highly context-sensitive and cannot be realized using earlier font specifications (e.g. true type fonts). Therefore, Urdu websites are made by either using Naskh font (normally used for Arabic, and which is unnatural for Urdu readership) or by putting scanned images of text written in Nastalique (which takes a large amount of memory and makes the websites very slow to access). Therefore, to make Urdu web and other publishing more effective and efficient, a character-based Nastalique font for Urdu needs to be developed, which can be used by existing software.

Project Objectives

The recent Open Type Font (OTF) specification extends the existing font formalism to enable the realization of context-sensitive writing systems like Urdu Nastalique. This project focuses on using this formalism to develop a Nastalique font for Urdu, which in turn requires first, a scientific study of Nastalique orthography, and then its modeling using OTF specification.

Development of this font will enable the users to publish electronically in Urdu and reach out to the extensive readership across the world. As OTF is a standard formalism, specialized software would not be needed, and this font will be enabled within the existing applications and web browsers (which support OTF).

Project Beneficiaries

Definition and free disbursement of Nastalique font for Urdu will accelerate Urdu publishing through electronic media and will benefit the 60 million readers of Urdu across the world. People who do not understand a second language (e.g. English, which is the lingua franca of computers and internet) will also be able to publish and access web pages, email, chat, etc., and a host of other computer applications.

Project Sustainability

Center for Research in Urdu Language Processing (CRULP, www.crulp.nu.edu.pk) is a dedicated effort of National University of Computer and Emerging Sciences (NUCES) showing its long-term commitment to the development of Urdu Computing. CRULP is currently working in the areas of script, speech and language processing of Urdu, with dedicated faculty and labs. In addition, NUCES offers coursework and specialization in its graduate Computer Science program in these areas.

Development of OTF for Urdu Nastalique is one of many projects initiated at CRULP to provide Urdu interface to computing and information technology. The research and development undertaken for this project will be used towards these long-term objectives of CRULP.

Thus, the work done for this project will be sustained through the continuing research and development at CRULP.

Project Methodology

What is Nastalique?
Nastalique font is computationally complex for many reasons. First, letters are written using a flat nib (traditionally using bamboo pens) and both trajectory of the pen and angle of the nib define a letter. Each letter has precise writing rules, relative to the length of the flat nib, as described in Figure 1 below.

Figure 1: Rules for writing mad, alif, bay, tay and jim letters of Urdu in Nastalique

Second, this cursive font is highly context sensitive. Shape of a letter depends on multiple neighboring characters. Current work shows that shape of each target letter Lt may depend on the context of four neighboring characters: L1LtL2L3L4. This analysis also shows that a letter may have as many as sixty shapes depending on context. Third, there is no concept of right or left alignment, and all text written in Nastalique is justified. In addition, there is no concept of space in Nastalique font.

Therefore, unlike roman script that uses space to separate words and justify lines, justification in Nastalique is achieved through two measures: (i) stretching of certain letters within ligatures, (ii) moving and overlapping ligatures. This adds another dimension to existing complex contextual rules to determine the shape of letters. Further, this also adds to the difficulty of computing multiple baselines, which depend on letters within ligatures.

Proposed Methodology to Model Nastalique

Due to the complexities outlined above, there is considerable effort required to segment connected letters (ligatures) and to formulate placement and joining rules. To avoid the complexities, current systems use scanned images of ligatures, without segmentation. These ligature based systems, effective in producing good Nastalique, require excessive storage space because the number of ligatures is far more than the number of letters (at least 25,000 ligatures vs. 40 characters for Urdu).

Current project addresses this issue by creating a character-based font for Nastalique for Urdu. The approach taken is to identify all the different shapes assumed by characters in different contexts, formulate rules to join them, and model them using OTF specification.

The project is divided into three main phases. First phase is based on orthographic analysis of Nastalique for Urdu. Second phase concentrates on modeling these findings using the current font technology - Open Type Font (OTF) specification. The final phase focuses on quality assurance of the designed system.
Orthographic Analysis of Nastalique for Urdu
This phase is divided into the following steps:

1. Context-dependent shape identification - part 1
The first part of the analysis will take this continuously varying script and determine a set of shapes for each letter occurring in different contexts. Initial work done in this area shows as many as 700 shapes for 20 characters. Because calligraphers are not trained to look at shapes from the point of view of font modeling, the research team will do this work.

2. Context-dependent shape identification - part 2
Once the initial analysis of shapes is completed, two calligraphers will be employed for this verification to make it more objective and accurate.

3. Classification of Urdu characters
Various letters change similarly in similar contexts. This stage will identify the characters and contexts for which similar shape changes are observed. Documenting this behaviour will reduce the number of entries/rules in the modeling stages.

4. Formulation of placement rules
As noted above, to form a ligature (letters connected together in one form), various letters are moved in a two dimensional plane to join effectively. This results in multiple baselines for a letter depending on the context. This part of the study will concentrate on determining these placement rules.

5. Formulation of joining rules
In the final stage of analysis of Nastalique, the joining rules for various letters will be determined, i.e. how various shapes of letters join effectively with other letter shapes and how pen angle and trajectory may be kept continuous over these joins.

Nastalique modeling using OTF specification

This phase is divided into the following steps:

1. Calligraphy of ligatures
Once all the shapes of each letter have been determined, ligatures containing each shape have to be carefully written on a fixed scale for scanning purposes. This work will be done by an expert calligrapher using a fixed pen width.

2. Scanning and Segmentation
Ligatures that have been written will be scanned at a high resolution (approx. 4000 dpi) and researchers under the supervision of an expert calligrapher will segment target letter shapes. The segmentation will be done using Adobe Photoshop.

3. Vectorization of Segments for font production
The segmented letters will then be fitted with splines using Corel Trace (shipped within Corel Draw). These will be the basic vectorized shapes used in the OTF shape tables.

4. Modeling within OTF specification
Once the basic shapes have been incorporated, the placement and joining rules researched earlier will be modeled within OTF specification to create Nastalique.

Verification and Validation of Nafees Nastalique

During the development of Nafees Nastalique a test plan will be generated to verify and validate the font generated. Two aspects will be addressed. First, how this font described with OTF specification fits in with existing applications supporting this font standard. Second, to what degree the OTF modeling has faithfully captured the rule-base of Nastalique.

1. Conformance with software supporting OTF specification
Conformance will be tested with two types of applications
a. With common browsers, by creating and publishing a web-site.
b. With common desktop tools.

2. Verification of rendered Nastalique
The rendered Nastalique will be verified by expert calligraphers. They will scrutinize the basic shapes, placement and joining of these shapes vis-� -vis Nastalique. Two calligraphers will be used to maintain objectivity.

Project Timelines

The timelines of each of the phases explained in the Methodology section are tabulated below, showing their duration, expected start and end dates and their dependency on any earlier activities. Each phase is divided into sub-phases such that there will be a report generated at the end of each of sub-phase (as explained in the Monitoring section below).

Project Outputs

The projects will have the following outputs.

1. Self-extracting installation program that will install Nafees Nastalique.

2. Publications on the lexicon and rule base developed for Nastalique and modeling techniques employed for the realization of Nafees Nastalique.

3. Sample web-site creation and publication using the Nafees Nastalique.

4. The above will be made available for free download at the website of CRULP, NUCES ( www.crulp.nu.edu.pk ).

5. The developed font will also be advertised at relevant research and development forums.

Project Monitoring

In order to monitor the project, the following mechanisms will be put in place. First, a detailed project plan (in line with the plan shown above) will be made and updated weekly to track the progress of the project.

Second, reports listed below will be published at various milestones shown in the timelines.

Finally, six-monthly progress reports will be published. These reports will highlight progress along the research and development activities and the financial expenditures incurred against the allocated budget.

All the above will be available for internal and external audits.