Text conversion methodology

1. Identify the kind of text problem first

A font problem and an encoding problem can look similar. Unicode text may display poorly because a font is missing glyphs, while legacy text may become Latin letters or symbols because the original font encoded Sinhala or Tamil shapes on non-Unicode character positions.

LankaTextTools asks for the known source font when a deterministic legacy mapping is required. We avoid claiming reliable automatic font detection for arbitrary copied text.

2. Keep mappings deterministic

Legacy-font converters use ordered character and sequence mappings. Longer conjunct and vowel sequences must be processed before simpler single-character substitutions, otherwise an early replacement can destroy the pattern needed for a later rule.

The initial Sinhala converter engine is based on established open-source conversion work that credits the Language Technology Research Laboratory at the University of Colombo School of Computing. LankaTextTools wraps conversion in a new interface, release system and explanatory layer rather than presenting the underlying mapping as original linguistic research.

3. Treat transliteration as a separate operation

Singlish and Tanglish tools work from phonetic Latin-letter input. They are not translation engines. The user already intends a Sinhala or Tamil word, and the tool changes the writing system into Unicode script. Different phonetic spellings can legitimately produce different results.

4. Normalize Unicode carefully

Unicode normalization can make canonically equivalent sequences consistent, but joiner characters need caution. ZWJ and ZWNJ can influence Indic-script rendering. Our Unicode cleaner therefore keeps joiner removal optional rather than stripping them from every input.

5. Test representative text, not only easy words

A converter is more useful when tests cover independent vowels, consonants, dependent vowel signs, conjuncts, punctuation, numerals, joiner sequences and common real-world words. We also use known-answer examples from the underlying mapping implementations where available.

Users should still review important names and document text because PDFs, mixed-font files and already-corrupted source text can introduce information that no mapping can reconstruct with certainty.

6. Keep unpublished tools out of the index

LankaTextTools uses a central release registry. Before a scheduled release, a future tool returns a not-found response and is excluded from navigation, related-tool components and the XML sitemap. After its release time, those surfaces update from the same publication rule. This avoids thin “coming soon” pages being indexed.