The field of medical research is enormous. It includes nearly every country in the world. This means that scientists are constantly contributing their findings in a variety of different languages. While that doesn’t typically affect collaboration within a particular area, it hurts scientific teamwork as a whole.
That’s why companies are turning to technologies like artificial intelligence (AI) and natural language processing (NLP) to automatically translate medical research. AbbVie, a global biopharmaceutical company, is leading the way thanks to its partnership with Intel.
AbbVie is a massive company that serves more than 30 million patients in 175 countries around the globe. It also has a team of more than 47,000 employees that need to collaborate on a variety of projects.
Making the information in those projects accessible in multiple languages is an essential part of AbbVie’s business. Unfortunately, commercially available translation software isn’t up the task of translating the complexities of medical writing.
Researchers who speak English as a second language must overcome that hurdle every day. Without the right tools, research is often left either untranslated or poorly translated, making it harder for scientists and team members to collaborate.
To address this issue, AbbVie worked with Intel to design a tool called Abbelfish Machine Translation. It uses AI and NLP to translate biomedical text between multiple languages. The tool currently supports English, Spanish, Italian, Portuguese, Russian, Chinese, and Japanese. Intel notes that more than one million scientific texts are translated by Abbelfish every year.
Brian Martin, the head of artificial intelligence at AbbVie, says, “We built Abbelfish Machine… to accelerate and scale the work of our researchers, reducing the time it takes to discover and deliver transformative medicines and therapies for patients. We’re looking to leverage Intel technology in a new way to deploy these capabilities at scale across the enterprise.”
Abbelfish is hosted on Intel Xeon-based servers to optimize its AI features. Each chunk of scientific text is broken down into 300-word pieces. From there, a separate CPU core is dedicated to each piece, facilitating low-latency translation. The platform can translate up to 10 texts simultaneously.
Meanwhile, since Abbelfish’s AI models were trained on paragraphs rather than sentences, the tool generates more accurate translations.
Of course, being able to read medical research is only one part of the equation. Researchers also need to be able to find the information they need. That presents a huge challenge since papers and journal studies are often hosted in a variety of different locations and formats.
AbbVie built another high-tech tool to solve this problem. Its AbbVie Search platform operates on a question-and-answer basis. This allows researchers to ask simple questions and get targeted answers much like performing a Google search.
For instance, a query for “What is the most common species of Human Coronavirus among adults,” returns an answer of “HCoV-OC43.”
This saves researchers from having to sift through dozens of papers to find simple answers. AbbVie Search is based on the BioBERT transformer model and relies on open-source semantic indexing and various biomedical research articles.
Considering that the field of medical research will only continue to grow in the years ahead, powerful tools like Abbelfish and AbbVie Search are noteworthy developments.