Overview
The National Cancer Institute (NCI) needed a way to make their data and other resources more accessible to cancer researchers. ESI designed three highly innovative bot applications to curate information on nanotechnology, enabling researchers to find information quickly and easily, without wading through a lot of extraneous information.
The Challenge
NCI, the leading federal agency responsible for cancer research at the National Institutes of Health, makes a wealth of data, tools, and other resources available to researchers in the field.
However, even the best resource won’t be helpful if people can’t easily find the information they need. To make their portal, caNanoLab, a truly centralized hub for biomedical nanotechnology researchers, NCI needed an efficient way to curate data from protocols, nanomaterial samples, and publications, as well as supporting documents.
The Solution
At ESI, we’re researchers too. We know how frustrating it can be when you can’t easily find and download the necessary data and information. We wanted to give NCI a way to simplify this process.
Our solution was to develop three highly innovative bot applications. These bots are powered by artificial intelligence (AI) and trained on vast amounts of text data (i.e., large-language models). Researchers can “talk” to these chatbots to get the results they need for a variety of tasks.
What sets our bots apart, is that they’re trained on caNanolab-centric data, documents, and data models. We needed to be sure the bots would interact with cancer researchers in a way that would “speak their language” and work across very different types of data. We designed the bots to have specific functions, including:
- Curation Assistant. This chatbot leverages a large language model (i.e., Generative Pre-Trained Transformer model) to automatically scour and extract information from PDF documents. The bot not only gleans information from text, such as that found in methods and materials sections and conclusions and protocols, but it can also examine and extract data from figures, tables, and other graphics.
- caNanolabLibrarian. As with our Curation Assistant, we built this chatbot using terms and language that resonate with people in the field, making it especially intuitive for researchers to use. We built an “automated query” function to make it easier to ask complex questions for faster results. We also ensured that users can find a wide range of data types simultaneously, without requiring multiple queries.
- caNanolabWiki. For this chatbot, we used a blended approach, drawing on models that both accurately interpret questions and then seamlessly convert those queries into well-crafted answers. Using this chatbot, the system can readily understand and respond to researchers with different backgrounds and levels of expertise.
The Results
ESI’s solution gave NCI a collection of highly intuitive, specialized chatbots. Now, researchers don’t have to wonder if they’ve missed important data or used queries that return less-than-expected results. ESI’s chatbots give researchers an innovative and simple way to express their search criteria, returning their results quickly and efficiently. In all, researchers can query more than a thousand curated nanomaterials relevant to cancer research.
In creating these bots, ESI pushed the envelope of what Large Language Models can do, tailoring them to a specific audience, and creating a cutting-edge approach for accessing and analyzing data.
Ready to Make an Impact?
Let ESI build a chatbot tailored to your task
We offer customized chatbot solutions built on the latest informatics and computational tools.
Team Members
Connect and Share
