1. Error Detection
Tokens are normalized and checked against the active language dictionary and orthographic rules.
Review spelling, orthography and punctuation with dedicated English, Persian and Kurdish Sorani engines.
Detected issues are highlighted by category. Click a highlight to open its suggestion.
Highcharts visualizations summarize accuracy, issue categories, confidence and where problems occur in the document.
How detection, candidate generation, ranking and language-specific normalization work in this project.

Specializing in NLP, low-resource languages, spelling systems and speech recognition.
Tokens are normalized and checked against the active language dictionary and orthographic rules.
Damerau-Levenshtein distance finds plausible corrections including insertions, deletions, substitutions and transpositions.
Candidates are scored using edit similarity, common prefix and dictionary frequency signals.
Persian and Kurdish use separate orthographic normalization so Arabic-form letters and punctuation are standardized correctly.
Accepted domain words can be added to the runtime dictionary so they are no longer flagged.
Health, issue types, confidence, top errors and document segments help evaluate text quality visually.