MULTILINGUAL

TEXT
RECOGNITION
SYSTEM BY
USING OCR
AN ADVANCED OCR SOLUTION FOR MULTIPLE
LANGUAGES
• Presented by: Ms. Alfiya Sayyed & Ms. Rutuja
Shivale
• Guided by: Prof. Amruta Navale
• Institution: Dr. D. Y. Patil Arts, Commerce and
Science College, Pimpri
• Academic Year: 2023-24
 Table of Contents
01 Abstract 08 Methodology

02
04 Introduction 09 Software Requirement Analysis

0301 Objective 1006 System workflow diagram

0402 Literature Survey 07

11 Result

0503 Comparison of Existing & Proposed System 12 Conclusion

06 What is OCR? 13 Future Enhancement

07 System Architecture 14 Reference

2
 ABSTRACT
• Overview:
The MLTR system aims to provide accurate
text extraction from images containing multiple
languages. This system leverages OCR techniques to
enhance accessibility and usability across different
linguistic contexts.
• Key Features:
Multilingual support: Handles text in
multiple languages.
Robust image preprocessing: Ensures
high- quality text extraction.
User- friendly interface: Simplifies the
user experience.

3
 INTRODUCTION
Background:
 In today’s digital age, there is a growing need to convert
printed documents into electronic formats.
 Electronic documents enhance data security,
accessibility and ease of sharing and editing
Importance of OCR:
 OCR technology allows for the conversion of different
types of documents such as PDFs, or images into
editable and searchable data
 Traditional OCR system are often limited to single
languages, reducing their effectiveness in a multilingual
world.
Multilingual Text Recognition System:
 Our project aim to address the limitations of traditional
OCR systems by supporting text recognition in multiple
languages.
 This system enhances usability and accessibility for
users who work with documents in various languages.

4
 OBJECTIVES
Developing multilingual text recognition system
Enhance Data Security
Improves Accessibility and Usability
Implement Efficient Preprocessing Techniques
Accurate Language Detection
User- Friendly Interface

5
 LITERATURE SURVEY
 Background: Traditional OCR systems includes Tesseract,
Adobe Acrobat OCR, and ABBYY FineReader but this system
struggles with multilingual texts, varied fonts poor quality
images and handwritten text.

 Current Trends: Modern OCR systems use machine learning

and deep learning techniques to improve accuracy and handle
a variety of languages and fonts.

 Limitations of Existing System: Issues with multilingual

support, accuracy in noisy images, and computational
requirements.

6
COMPARISON OF EXISTING SYSTEM & PROPOSED SYSTEM

EXISTING SYSTEM

Proposed system supports multiple

Existing systems only have the functionalities such as extracting text in
capability to convert and recognize multiple languages. It also adds benefit
only the documents of English or a by providing heterogeneous characters
specific language only. That is, the recognition.
older OCR system is uni-lingual or bi-
lingual.

PROPOSED SYSTEM

7
 WHAT IS OCR?
 Definition: OCR (Optical Character Recognition) technology converts different types of documents, such
as scanned papers documents, PDF files, or images captured by a digital camera, into editable and
searchable data.

 Types of OCR:
1.OCR: General character recognition
2.Optical Word Recognition: Recognizes entire words rather than individual characters.
3.Intelligent Character Recognition: Recognizes hand-printed characters.
4.Intelligent Word Recognition: Recognizes hand- printed words.

8
 SYSTEM ARCHITECTURE
 Overview:
Our OCR system architecture is designed to support
multilingual text recognition, ensuring high accuracy
and efficient processing of various languages scripts.
 Components:
1. User interface: Allows users to upload images
and view extracted text and detected language.
2. Image Preprocessing Modules: Converts
uploaded images to grayscale and enhances them
for better OCR performance.
3. OCR Engine: Utilizes Tessereact OCR to extract
text from preprocessed images.
4. Languages Detection Module: Detects the
languages of the extracted text using langdetect
library
5. Output Module: Displays the extracted text and
detected languages to the user in user friendly
format.
9
 METHODOLOGY
The methodology of our OCR system involves several key
steps: image preprocessing, text extraction, language detection, and text
post-processing.
Each step plays a critical role in ensuring the accuracy and efficiency of
text recognition from images.

 Image Preprocessing
• Image preprocessing is the first step, where the uploaded image
is prepared for text extraction.
• This includes converting the image to grayscale to reduce
complexity and using noise reduction techniques to enhance the
quality of the text in the image.
• These steps help in improving the accuracy of the OCR process.
 Text Extraction
• The core of the OCR system is the text extraction phase, where
Image preprocessing the preprocessed image is processed to identify and extract text.
• We use Tesseract OCR, a powerful open-source library,
configured to recognize multiple languages.
• Tesseract analyzes the text areas in the image and converts them
into machine-encoded text.
9
 Language Detection
• After extracting the text, the system detects the language using the Langdetect library.
• This step is crucial for processing documents that contain text in multiple languages.
• The detected language information helps in subsequent text processing and formatting steps.
 Text Post-Processing
• In the post-processing phase, the extracted text is cleaned and formatted.
• This involves removing unwanted characters and symbols, and structuring the text into readable format.
• The final output is a clean, editable document that retains the original text's integrity.

11
 SOFTWARE REQUIREMENTS ANALYSIS

 Problem Statement:
Extracting accurate text from images containing multiple languages.

 System Requirements:
• Hardware: High- resolution camera or scanner, high performance
processor.
• Software: Python, OpenCV, Tesseract, langdetect.

 Libraries used:
• OpenCV: For image processing.
• Pytesseract: Python wrapper for tesseract OCR.
• Langdetect: Language detection library

9
System Workflow Diagram

9
 RESULTS
The Multilingual Text Recognition System project successfully implemented a web application
that allows users to upload images containing text and extracts the text from them. The system supports
multiple languages, enabling users to extract text in various languages accurately.

 Key Performance Metrics:

1. Accuracy of OCR:
• OCR accuracy rate of 95% across various languages
2. Language detection accuracy:
• Language detection accuracy is 90%
3. Processing time:
• Average processing time of 3 seconds per image

9
GUI INTERFACE DESIGN

9
 Examples of Successful Text Extraction and Language Detection
Example 1:

Input Image:
Extracted text:

Detected Language: Hindi

9
Example 2:

9
 CONCLUSION
 Through our project, we successfully addressed the challenges associated with processing multilingual
documents, providing users with a reliable and efficient tool for extracting text from images in various
languages.

 The system ability to accurately detect and extract text in languages ranging from English and Hindi to
Arabic and beyond has far- reaching implication for data accessibility.

 By overcoming language barriers, our system empowers users to efficiently process multilingual
documents, improving productivity and reducing the risk of errors associated with manual transcription.

18
 FUTURE ENHANCEMENT

 Enhanced language support: Continuously expand the language support to encompass additional
languages and dialects, catering to the diverse linguistic needs of users worldwide.

 Mobile Application Development: Develop a mobile application version of the system to enable users
to perform text extraction tasks on the go, leveraging the capabilities of smartphones and tablets.

 Integration with AI: Explore the integration of artificial intelligence (AI) techniques, such as machine
learning and deep learning, to improve language detection accuracy and optimize text extraction
algorithms.

 Improved Accuracy: Invest in research and development efforts to further enhance the accuracy of text
recognition, especially for complex scripts and low-quality images.

9
 REFERENCE

For the complete reference and understanding of OCR refer jeff heaton's chapter 7 from
www.jeffheaton.com
 The IEEE standard reference paper from which we collected our problem statement is authorized by
Dana Petcu, Silviu Panica, Viorel Negru and Andrei Eckstein of Computer Science Department who
are from West University of Timisoara, Romania.

 The reference paper is also authorized by Doina Banciu from National Institute for Research and
Development in Informatics, Romania.

 A. Revathi and N. A. Modi, "Comparative Analysis of Text Extraction from Color Images using
Tesseract and OpenCV," 2021 8th International Conference on Computing for Sustainable Global
Development (INDIACom), 2021, pp. 931-936, DOI: 10.1109/INDIACom51348.2021.00167.