Handwritten text recognition: testing, testing, 1.2.3.

Find out what we're working on

We’re pilot testing Optical Character Recognition (OCR) and Handwritten Text Recognition (HTR) for handwritten and unpublished materials in Trove, such as personal papers, letters, and diaries. The National Library of Australia is working on this behind the scenes, and now we’re going to add it to Trove to see how it works in the Trove ecosystem. We’re also testing how accurately this innovative technology can convert images of different handwriting styles into transcripts. 

If you like browsing the diaries, letters and archives category you will spot what we’re working on. You should notice: 

  • more search results 

  • transcripts with text correcting functionality next to some items.  

  • the ability to scroll through some handwritten and unpublished material, in the same way you can for books. 

Over the next few months, we will be testing 5,000 – 10,000 digitised pages from the personal papers of former Prime Minister of Australia Sir Robert Menzies, General Sir John Monash, and the first two women elected to the Australian federal parliament: Dame Enid Lyons and Dame Dorothy Tangney; and the Australian Joint Copyright Project (AJCP).  

You may see the functionality we are testing come and go on some items.

Screen-capture of Trove website. A handwritten letter can be seen on the right side of the screen and on the left is a computerised transcript of the letter's text

Improving search results 

Search results are fueled by keywords found within collection items – like words in a newspaper article –and supporting contextual information (metadata), like the title and description of an item. By applying an OCR process to digitised items, we can make the entire digitised content searchable. This greatly enhances your Trove search results and information you can easily find. This is how we’ve been delivering relevant newspaper articles to your search results since the first days of Trove. In 2022, we also opened the books on Australia's history by rolling out OCR functionality for books, magazines, newsletters, and almanacs. 

Our wonderful community of volunteer text correctors plays a significant role in enhancing the accuracy of OCR results by correcting the transcriptions that appear next to digitised content. 

OCR vs HTR? 

Optical Character Recognition (OCR) is a technology that converts images of typed or printed text into an editable and searchable transcription. Handwritten Text Recognition (HTR) is a similar, but more advanced technology, that can convert handwriting. 

What you can do 

While we’re testing, text correcting functionality will appear on some handwritten and unpublished materials. If you spot what we’re working on, you are welcome to experiment with text correcting the transcripts for this material. These text corrections will appear in your text correcting history; however, we haven’t yet released a filter for this category in the text correcting hall of fame

You can also continue text correcting newspapers, gazettes, books, magazines, newsletters, and almanacs to help us improve Trove search results. You can learn more about other ways you can take part in the Trove community

Questions and feedback 

Right now, we’re testing this new functionality before we look to roll it out for all handwritten material on Trove. We understand you may have questions and we always welcome your feedback. We use this to improve our services and your experience using Trove. You can contact us to let us know what you think.  

You can also find answers to frequently asked questions about Trove and view recent system updates