CtrlK
BlogDocsLog inGet started
Tessl Logo

tessl/maven-org-bytedeco--tesseract-platform

JavaCPP platform aggregator for Tesseract OCR native libraries providing cross-platform OCR capabilities in Java applications

—

Pending
Overview
Eval results
Files

language-support.mddocs/

Language Support

Multi-language OCR with support for 100+ languages, custom language models, and language detection capabilities. Tesseract provides comprehensive support for different scripts, writing systems, and language-specific recognition optimizations.

Capabilities

Language Initialization

Initialize Tesseract with one or more languages for recognition.

public class TessBaseAPI {
    // Language initialization
    public int Init(String datapath, String language, int oem);
    public int Init(String datapath, String language);
    
    // Language information
    public native @Cast("const char*") BytePointer GetInitLanguagesAsString();
    public void GetLoadedLanguagesAsVector(StringVector langs);
    public void GetAvailableLanguagesAsVector(StringVector langs);
}

Language Code Format:

  • Single language: "eng" (English), "fra" (French), "deu" (German)
  • Multiple languages: "eng+fra+deu" (English + French + German)
  • Script-based: "chi_sim" (Simplified Chinese), "ara" (Arabic)

Usage Example

TessBaseAPI api = new TessBaseAPI();

// Initialize with single language
int result = api.Init(null, "eng");

// Initialize with multiple languages
int result2 = api.Init(null, "eng+fra+deu");

// Initialize with mixed scripts
int result3 = api.Init(null, "eng+ara+chi_sim");

// Check which languages were loaded
BytePointer loadedLangsPtr = api.GetInitLanguagesAsString();
String loadedLangs = loadedLangsPtr.getString();
System.out.println("Loaded languages: " + loadedLangs);
loadedLangsPtr.deallocate();

// Get available languages as vector
StringVector availableLangs = new StringVector();
api.GetAvailableLanguagesAsVector(availableLangs);

System.out.println("Available languages:");
for (int i = 0; i < availableLangs.size(); i++) {
    System.out.println("  " + availableLangs.get(i));
}

Common Language Codes

Latin Script Languages:

// Western European
"eng" // English
"fra" // French  
"deu" // German
"ita" // Italian
"spa" // Spanish
"por" // Portuguese
"nld" // Dutch
"dan" // Danish
"nor" // Norwegian
"swe" // Swedish
"fin" // Finnish

// Eastern European
"pol" // Polish
"ces" // Czech
"slk" // Slovak
"hun" // Hungarian
"ron" // Romanian
"hrv" // Croatian
"slv" // Slovenian
"est" // Estonian
"lav" // Latvian
"lit" // Lithuanian

Non-Latin Script Languages:

// Cyrillic
"rus" // Russian
"ukr" // Ukrainian
"bul" // Bulgarian
"srp" // Serbian
"mkd" // Macedonian
"bel" // Belarusian

// Arabic Script
"ara" // Arabic
"fas" // Persian (Farsi)
"urd" // Urdu
"pus" // Pashto

// Asian Scripts
"chi_sim" // Simplified Chinese
"chi_tra" // Traditional Chinese
"jpn"     // Japanese
"kor"     // Korean
"tha"     // Thai
"vie"     // Vietnamese
"khm"     // Khmer (Cambodian)
"lao"     // Lao

// Indic Scripts
"hin"     // Hindi
"ben"     // Bengali
"guj"     // Gujarati
"pan"     // Punjabi
"tel"     // Telugu
"kan"     // Kannada
"mal"     // Malayalam
"tam"     // Tamil
"ori"     // Odia
"san"     // Sanskrit

// Other Scripts
"heb"     // Hebrew
"ell"     // Greek
"amh"     // Amharic
"geo"     // Georgian
"arm"     // Armenian

Language Selection Example

public class MultiLanguageOCR {
    
    public static String recognizeWithLanguageDetection(PIX image) {
        TessBaseAPI api = new TessBaseAPI();
        
        try {
            // Try common language combinations based on context
            String[] languageSets = {
                "eng",                    // English only
                "eng+fra+deu",           // Western European
                "eng+spa+por",           // English + Iberian
                "eng+rus+ukr",           // English + Slavic Cyrillic
                "eng+ara",               // English + Arabic
                "eng+chi_sim+jpn+kor"    // English + East Asian
            };
            
            String bestResult = "";
            int bestConfidence = 0;
            
            for (String langs : languageSets) {
                api.End(); // Clean up previous initialization
                
                if (api.Init(null, langs) == 0) {
                    api.SetImage(image);
                    String text = api.GetUTF8Text();
                    int confidence = api.MeanTextConf();
                    
                    System.out.println("Languages: " + langs + ", Confidence: " + confidence);
                    
                    if (confidence > bestConfidence) {
                        bestConfidence = confidence;
                        bestResult = text;
                    }
                }
            }
            
            return bestResult;
            
        } finally {
            api.End();
        }
    }
}

Script and Writing Direction Detection

Automatic detection of script types and text direction for proper processing.

public class PageIterator {
    // Orientation and script information
    public void Orientation(int[] orientation, int[] writing_direction,
                           int[] textline_order, float[] deskew_angle);
}

public class ResultIterator {
    // Language detection per word
    public String WordRecognitionLanguage();
    public int WordDirection();
    public boolean ParagraphIsLtr();
}

// Writing direction constants
public static final int WRITING_DIRECTION_LEFT_TO_RIGHT = 0;
public static final int WRITING_DIRECTION_RIGHT_TO_LEFT = 1;
public static final int WRITING_DIRECTION_TOP_TO_BOTTOM = 2;

// Script direction constants
public static final int DIR_NEUTRAL = 0;        // Neutral characters
public static final int DIR_LEFT_TO_RIGHT = 1;  // LTR scripts (Latin, Cyrillic)
public static final int DIR_RIGHT_TO_LEFT = 2;  // RTL scripts (Arabic, Hebrew)
public static final int DIR_MIX = 3;            // Mixed direction text

Usage Example

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "eng+ara+heb");  // Mixed LTR/RTL languages
api.SetImage(image);

ResultIterator resultIt = api.GetIterator();
resultIt.Begin();

// Analyze text direction and language per word
do {
    String word = resultIt.GetUTF8Text(RIL_WORD);
    String wordLang = resultIt.WordRecognitionLanguage();
    int direction = resultIt.WordDirection();
    
    String directionName = switch (direction) {
        case DIR_LEFT_TO_RIGHT -> "LTR";
        case DIR_RIGHT_TO_LEFT -> "RTL";
        case DIR_MIX -> "Mixed";
        default -> "Neutral";
    };
    
    System.out.printf("Word: '%s' Language: %s Direction: %s\n", 
                     word, wordLang, directionName);
    
} while (resultIt.Next(RIL_WORD));

// Check paragraph direction
resultIt.Begin();
if (resultIt.IsAtBeginningOf(RIL_PARA)) {
    boolean isLtr = resultIt.ParagraphIsLtr();
    System.out.println("Paragraph direction: " + 
                      (isLtr ? "Left-to-Right" : "Right-to-Left"));
}

Language-Specific Configuration

Optimize recognition for specific languages and scripts.

Arabic Script Configuration

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "ara");

// Arabic-specific optimizations
api.SetVariable("textord_arabic_text", "1");
api.SetVariable("textord_use_cjk_fp_model", "0");
api.SetVariable("preserve_interword_spaces", "1");

// Enable bidirectional text support
api.SetPageSegMode(PSM_AUTO);

Chinese/Japanese/Korean (CJK) Configuration

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "chi_sim+jpn+kor");

// CJK-specific optimizations
api.SetVariable("textord_use_cjk_fp_model", "1");
api.SetVariable("language_model_penalty_non_dict_word", "0.25");
api.SetVariable("language_model_penalty_non_freq_dict_word", "0.15");

// Vertical text support
api.SetPageSegMode(PSM_AUTO);
api.SetVariable("textord_tabfind_vertical_text", "1");

Indic Script Configuration

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "hin+ben+guj");

// Indic script optimizations
api.SetVariable("textord_use_cjk_fp_model", "0");
api.SetVariable("preserve_interword_spaces", "1");
api.SetVariable("segment_penalty_dict_nonword", "1.25");

Custom Language Models

Work with custom trained language models and specialized vocabularies.

Loading Custom Models

// Custom language models are placed in tessdata directory
// with naming convention: <lang>.traineddata

TessBaseAPI api = new TessBaseAPI();

// Load custom model (place custom_eng.traineddata in tessdata)
int result = api.Init("/path/to/custom/tessdata", "custom_eng");

// Combine custom with standard models
int result2 = api.Init("/path/to/tessdata", "eng+custom_domain");

// Use specialized models for specific domains
int result3 = api.Init("/path/to/tessdata", "eng_medical");  // Medical terminology
int result4 = api.Init("/path/to/tessdata", "eng_legal");    // Legal documents

Custom Vocabulary Configuration

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "eng");

// Load custom word list (one word per line in tessdata/eng.user-words)
api.SetVariable("load_system_dawg", "1");
api.SetVariable("load_freq_dawg", "1");  
api.SetVariable("load_unambig_dawg", "1");

// Adjust language model penalties for custom vocabulary
api.SetVariable("language_model_penalty_non_dict_word", "0.3");
api.SetVariable("language_model_penalty_non_freq_dict_word", "0.2");

// Enable user patterns (tessdata/eng.user-patterns)
api.SetVariable("user_patterns_suffix", "user-patterns");

Multilingual Document Processing

Handle documents with mixed languages and scripts.

Language Switching Strategy

public class MultilingualProcessor {
    
    public static class LanguageRegion {
        public String language;
        public int left, top, right, bottom;
        public double confidence;
    }
    
    public static String processMultilingualDocument(PIX image) {
        TessBaseAPI api = new TessBaseAPI();
        StringBuilder result = new StringBuilder();
        
        try {
            // Step 1: Detect layout and potential language regions
            api.Init(null, "osd");  // Orientation and Script Detection
            api.SetPageSegMode(PSM_OSD_ONLY);
            api.SetImage(image);
            
            // Get orientation info
            PageIterator pageIt = api.AnalyseLayout();
            // ... orientation detection logic ...
            
            // Step 2: Process with multiple language models
            String[] languageTests = {
                "eng", "fra", "deu", "spa", "ita",      // Latin scripts
                "rus", "ukr", "bul",                     // Cyrillic
                "ara", "fas",                            // Arabic
                "chi_sim", "jpn", "kor"                  // CJK
            };
            
            api.End();
            
            // Test each language and find best matches per region
            Map<String, Double> languageConfidences = new HashMap<>();
            
            for (String lang : languageTests) {
                api.Init(null, lang);
                api.SetImage(image);
                
                String text = api.GetUTF8Text();
                int confidence = api.MeanTextConf();
                
                if (confidence > 70 && !text.trim().isEmpty()) {
                    languageConfidences.put(lang, (double) confidence);
                }
                
                api.End();
            }
            
            // Step 3: Use best language combination
            String bestLanguages = determineBestLanguageSet(languageConfidences);
            
            api.Init(null, bestLanguages);
            api.SetPageSegMode(PSM_AUTO);
            api.SetImage(image);
            
            result.append(api.GetUTF8Text());
            
        } finally {
            api.End();
        }
        
        return result.toString();
    }
    
    private static String determineBestLanguageSet(Map<String, Double> confidences) {
        // Logic to combine compatible languages based on confidence scores
        List<String> topLanguages = confidences.entrySet().stream()
            .sorted(Map.Entry.<String, Double>comparingByValue().reversed())
            .limit(3)
            .map(Map.Entry::getKey)
            .collect(Collectors.toList());
            
        return String.join("+", topLanguages);
    }
}

Language Model Information

Access information about loaded language models and their capabilities.

public class TessBaseAPI {
    // Language information
    public String GetInitLanguagesAsString();
    public void GetLoadedLanguagesAsVector(StringVector langs);
    public void GetAvailableLanguagesAsVector(StringVector langs);
}

public class ResultIterator {
    // Per-word language detection
    public String WordRecognitionLanguage();
}

Language Introspection Example

TessBaseAPI api = new TessBaseAPI();
api.Init(null, "eng+fra+deu+ara+chi_sim");

// Get comprehensive language information
System.out.println("Initialized languages: " + api.GetInitLanguagesAsString());

StringVector loaded = new StringVector();
api.GetLoadedLanguagesAsVector(loaded);
System.out.println("Loaded language models:");
for (int i = 0; i < loaded.size(); i++) {
    System.out.println("  " + loaded.get(i));
}

StringVector available = new StringVector();
api.GetAvailableLanguagesAsVector(available);
System.out.println("Available language models:");
for (int i = 0; i < available.size(); i++) {
    System.out.println("  " + available.get(i));
}

// Analyze language detection per word
api.SetImage(multilingualImage);
ResultIterator resultIt = api.GetIterator();
resultIt.Begin();

Map<String, Integer> langCounts = new HashMap<>();
do {
    String wordLang = resultIt.WordRecognitionLanguage();
    langCounts.merge(wordLang, 1, Integer::sum);
} while (resultIt.Next(RIL_WORD));

System.out.println("Language distribution in document:");
langCounts.forEach((lang, count) -> 
    System.out.println("  " + lang + ": " + count + " words"));

Error Handling and Language Fallbacks

Handle missing language models and provide fallback strategies.

public class RobustLanguageOCR {
    
    public static String recognizeWithFallback(PIX image, String preferredLangs) {
        TessBaseAPI api = new TessBaseAPI();
        
        try {
            // Try preferred languages first
            if (api.Init(null, preferredLangs) == 0) {
                api.SetImage(image);
                String result = api.GetUTF8Text();
                int confidence = api.MeanTextConf();
                
                if (confidence > 60) { // Good confidence
                    return result;
                }
            }
            
            // Fallback to English if preferred languages fail
            api.End();
            if (api.Init(null, "eng") == 0) {
                api.SetImage(image);
                String result = api.GetUTF8Text();
                System.out.println("Fell back to English recognition");
                return result;
            }
            
            throw new RuntimeException("No language models could be loaded");
            
        } finally {
            api.End();
        }
    }
    
    public static boolean isLanguageAvailable(String language) {
        TessBaseAPI api = new TessBaseAPI();
        try {
            int result = api.Init(null, language);
            return (result == 0);
        } finally {
            api.End();
        }
    }
    
    public static List<String> getWorkingLanguages(String[] candidates) {
        List<String> working = new ArrayList<>();
        
        for (String lang : candidates) {
            if (isLanguageAvailable(lang)) {
                working.add(lang);
            } else {
                System.out.println("Language model not available: " + lang);
            }
        }
        
        return working;
    }
}

Types

Language Code Examples

// Common language codes (ISO 639-3)
public static final String LANG_ENGLISH = "eng";
public static final String LANG_FRENCH = "fra";
public static final String LANG_GERMAN = "deu";
public static final String LANG_SPANISH = "spa";
public static final String LANG_ITALIAN = "ita";
public static final String LANG_PORTUGUESE = "por";
public static final String LANG_RUSSIAN = "rus";
public static final String LANG_ARABIC = "ara";
public static final String LANG_CHINESE_SIMPLIFIED = "chi_sim";
public static final String LANG_CHINESE_TRADITIONAL = "chi_tra";
public static final String LANG_JAPANESE = "jpn";
public static final String LANG_KOREAN = "kor";
public static final String LANG_HINDI = "hin";
public static final String LANG_HEBREW = "heb";

Text Direction Constants

public static final int WRITING_DIRECTION_LEFT_TO_RIGHT = 0;
public static final int WRITING_DIRECTION_RIGHT_TO_LEFT = 1; 
public static final int WRITING_DIRECTION_TOP_TO_BOTTOM = 2;

public static final int DIR_NEUTRAL = 0;
public static final int DIR_LEFT_TO_RIGHT = 1;
public static final int DIR_RIGHT_TO_LEFT = 2;
public static final int DIR_MIX = 3;

Language Information Structures

// String vector for language lists
public class StringVector {
    public long size();
    public String get(long i);
    // Used by GetLoadedLanguagesAsVector and GetAvailableLanguagesAsVector
}

Install with Tessl CLI

npx tessl i tessl/maven-org-bytedeco--tesseract-platform

docs

configuration.md

core-ocr-engine.md

index.md

language-support.md

layout-analysis.md

output-renderers.md

result-navigation.md

tile.json