About BIQE AI HTR Software Handwritten Text Recognition

Difficulties of OCR recognizing handwritten documents

OCR recognition of written text is hard. You cannot place someone’s handwriting into a specific font. Fonts like Times New Roman, Calibri, or Arial do not apply. Every person’s handwriting has unique traits.

Handwritten documents often do not use lined paper. This means that words that go together may be at different heights on the line. The challenge here is to segment the lines correctly.

The OCR software cannot tell which words belong together, making it difficult to recognize a handwritten text document.

Segmentation, a technique for separating and identifying individual words in a handwritten document, presents a significant challenge. The goal is to automate this process as much as possible. This task needs advanced OCR technology and a good understanding of handwriting patterns.

Another issue with OCR for handwritten documents is that not all images have the same format or layout. Sometimes, a page has a picture with some text. Other times, a page has only text or a mix of text and photos.

To achieve the best OCR result, you must rotate the different handwritten pages correctly and legibly. They usually perform this while scanning, but sometimes find it impossible.

BIQE AI HTR Software deals with all these issues.

Artificial Intelligence and Machine Learning

Artificial Intelligence makes computer systems do something that ‘normally’ requires natural or human intelligence. It allows computer systems to take actions on their own, which helps them achieve the goals set by the user or developer.

AI applications are diverse. They include:

• BIQE AI HTR software
• Google Search
• Netflix for movies
• YouTube for music
• Voice assistants like Siri, Google Assistant, and ChatGPT.

Machine Learning is a part of Artificial Intelligence that explores statistical algorithms. For example:

Suppose you have a 500-page 1800s book that you want to make searchable (OCR). Transkribus would say, “We train these Books using Machine Learning.”

How? By typing the text of those 50 images into a particular program. You then use ML to train these 50 images with the 50 pages of typed text.

This training creates a language model. With that trained language model, you can use it on the other 450 pages. These pages will then be automatically OCR’d. So you train with ML, and this ML learns to draw general conclusions from similar unknown data.  

There is a lot of confusion about whether Machine Learning is the same as Data Mining. The main difference is that Data Mining finds rules in large data sets. Machine Learning helps a computerunderstand these rules better. So, data mining is a research method that determines a particular outcome from the data collected.  

BIQE says: with our software you don’t have to train, you only have to upload your digital documents. BIQE AI HTR software uses AI and machine learning to solve problems with reading handwritten text documents. Below, I have listed some features we developed to achieve the best OCR results for all your handwritten documents.

Features BIQE HTR Software

• AUTOMATIC ROTATION

The initial and crucial step in document OCR involves scanning. This means scanning at least 300 dpi. If possible, make colour scans to keep as much pixel data as possible for editing.

Sometimes, others have already scanned the material in black and white at 150 dpi. It may also be skewed, upside down, or rotated by 90 degrees or more. Then, we recommend using our BIQE PROduction or BIQE Archive to enhance the images.

We cannot change black-and-white scanned images to color. However, our 39 image filters can enhance almost everything else.

The main goal of the image filters is to improve the text, which helps to achieve the best recognition rate.

BIQE software determines whether your images need rotation. Our BIQE OCR Server or BIQE HTR automatically corrects incorrectly rotated images. A properly rotated document will significantly improve the overall quality of OCR.

• ADVANCED SEGMENTATION ALGORITHM

With typed letters, you usually don’t have segmentation problems because all the words are neatly straight on a line.

With typed text in the background, a good OCR Engine like Abbyy will segment your document correctly before it is OCR-ed. But this is hugely different and much trickier with handwritten documents (see image above).

In many cases of handwritten texts, you will have to use a segmentation tool such as Escriptorium. You can then manually correct page segmentation by drawing a segmentation line under, through, or at the top of the words of each line. This is a time-consuming task.

While OCR engines can automatically handle segmentation for typed text, they cannot do the same for handwritten text. Relying on their expertise for Handwritten Text Recognition can lead to disappointing segmentation and OCR results.

BIQE HTR uses a special algorithm in a robust system. It can solve the segmentation problem in almost any manuscript.

You cannot control this segmentation technique because it works automatically in the background. With our software, you are in control. You can customize segmentation options. Therefore, we believe we’ve developed the best algorithm for Handwritten Text Recognition!

• LANGUAGES-INDEPENDENT

Most OCR Engines recognise one language on a page and use that language’s dictionary. If a handwritten page contains multiple languages, such as Greek and Latin, the OCR of that page will be more prone to OCR errors. It can recognize more than 200 languages.

BIQE HTR Software is foremost language-independent.

With artificial intelligence (AI), the OCR software can recognize the language in a document. It finds all languages in your document, even when more than two languages are present!

In a multilingual document, like our example with Greek and Latin, BIQE HTR software recognizes the languages. Then, it chooses and uses the right Greek and/or Latin dictionary for that page or document. This also includes selecting the correct OCR language.

• PARALLEL PROCESSING OR MULTI-THREAD SYSTEM

As the name suggests, parallel processing is fast because it works simultaneously on multiple processors or cores. These processors or cores/threads work separately to complete (partial) tasks.

So, as you know, multithreading is not the same as parallel processing. It is not the case that more threads execute tasks faster.

To understand better, let’s look at multithreading in single-core and multi-core processors.

Single-core processors

At first glance, multithreading on a single-core processor may seem absurd. After all, how can one physical processor simultaneously perform multiple tasks?

Developers can simulate multithreading on single-core processors using temporary multithreading or context switching.

How it works:

1. Task Queue: The operating system divides tasks into small chunks called threads. All these threads are in one place in a queue and wait to be processed.
2. Fast switching: The processor core quickly switches between threads, giving each of them short periods. During this time, the thread does its share of work and then gives way to the next one in the queue.
3. Illusion of multitasking: By quickly switching between threads, the processor appears to be processing multiple tasks simultaneously.

Technological benefits:

• Improved responsiveness: Quickly switching between tasks makes your computer more responsive, especially when running light-duty applications.
• Efficient use of resources is essential.
• If one thread uses the core, other threads can still use other resources.
• These resources include cache and memory.

Understanding that context switching between threads requires additional resources, which can slightly slow down the system.

When designing programs, you should know that the number of tasks running at the same time should match the number of processor cores. Otherwise, we will not only not increase the program’s performance but also reduce it because of additional context switches.

Multithreading BIQE

Our BIQE products, like BIQE AI HTR and BIQE PRO, effectively utilize all processor cores. We use modern principles and technologies to build programs that effectively use modern multi-core processors.

Our products can process multiple documents or pages at the same time. It is also possible to export different File Types, like ALTO-XML, JP2, and TXT.

BIQE uses multithreading, which is much faster than completing tasks one after another. This change dramatically improves its performance.

QUICK EXPORT AND SEARCH IN VIEWER

When handwritten documents or old or typed documents are OCR-ed, it is to search them. A regular viewer often does not work well when dealing with enormous data files. BIQE created a fast, elastic search viewer.

You can subdivide our viewer into folders and sub-folders as you see fit. This lets you choose more clearly which chapter of a book or document you want to search or leave out. Our viewer works with the combined file types Alto-xml with jp2, which you can easily import via our CMS.