Corpus-Driven Grade 2 Braille Conversion for Gurmukhi Script: Contraction Design, Mapping, and System Evaluation

Abstract

Grade 2 Braille uses contractions to represent common character sequences as single Braille cells. It is the standard medium for fluent Braille reading. But no Grade 2 Braille system had ever been defined for Punjabi text written in Gurmukhi script, until now. This paper presents the first corpus-driven design and implementation of a Grade 2 Braille conversion system for Punjabi. A large Punjabi n-gram corpus was used to identify six space-efficient contraction candidates: the words and bigrams ਦਾ, ਦੇ, ਦੀ, ਹੈ, ਅਤੇ, and ਵਿੱਚ. The Braille codes previously assigned to five infrequent Gurmukhi characters (ਙ, ਞ, ਖ਼, ਜ਼, ਲ਼) are reassigned as contraction codes for these high-frequency sequences instead. This paper formally defines the Grade 2 mapping table, the contraction selection method, and a six-stage conversion algorithm that processes both Unicode and legacy ASCII-encoded Punjabi text. The system was evaluated on eight test inputs ranging from 57 to 9,839 characters. It reached Grade 2 conversion accuracy of 99.4% to 100% and space savings of 4.7% to 6.6% for sentence-level inputs. Grade 1 conversion accuracy was 99.8% to 100% across all input types. A comparative analysis shows that the proposed system outperforms the existing Punjabi-to-Braille converter on every measured dimension. It also introduces capabilities that earlier work simply did not have: non-Unicode processing, automatic font detection, and Grade 2 contraction. The system has been reviewed by domain experts at Braille Press Ludhiana and is proposed for standardization through Punjabi University, Patiala.