Researchers at Northumbria University in England have developed a new method that uses artificial intelligence (AI) to improve visual search engines. The Deep Discoveries project creates a search platform that identifies and combines digitized images to facilitate the user experience.
- Are artificial intelligence and machine learning the same thing? Understand
- Facts about artificial intelligence you need to know
- Artificial intelligence gains ability to “imagine” the world around it
Instead of typing a keyword into a conventional search box, the new visual search uses a primary query image and integrated computer vision to find similar collections based on pre-programmed criteria such as colors, texture patterns, and geometric shapes.
“We found objective ways to visually demonstrate artificial intelligence reasoning to improve search criteria within the platform. After several attempts, we were able to improve the visual search engine using much more complex image comparison functions”, explains design professor Jo Briggs, co-author of the study.
The new visual research method was implemented in the British National Archives to catalog thousands of digitized collections. The idea is to create an interface capable of making search engines more precise, using specific parts of images to find similar items.
Deep Discoveries is one of eight projects funded by the cultural program that aims to open the British National Archives collection to the world. The five-year initiative will benefit the public and researchers, making heritage collections more accessible to all users.
“The design process brought us to a common understanding that allowed us to create a prototype in real time, accommodating the different points of view between the teams involved in this multidisciplinary project. With that, it was possible to develop a robust and easy-to-use platform”, adds the software engineer of the British National Archives Bernard Ogden.
The Deep Discoveries project uses new machine learning methods to create a computer vision search platform that can identify and match images in digitized collections at scale. The search technology focuses on specific themes, for example, it can recognize a rose in a textile pattern and the same flower in an herbarium or in a ceramic vase.
With this visual search system, researchers want to create a network of collections with specific images from across the UK, eliminating the need for a single type of unifying collection based on searching for keywords such as woodcuts, paintings, frescoes or photographs .
“Creating an integrated and accessible digital collection requires research and implementation of many standards and new technologies. But the initiative also opens the door to new playful and creative methods of search and discovery, such as computer vision research, which can lead to unexpected ways of public involvement with digitized images”, concludes the head of research at the British National Archive Lora Angelova.