The Artificial Intelligence Research and Development group explores new ways of addressing critical business problems by applying leading edge AI techniques, including machine learning, natural language processing, knowledge representation and reasoning.
We are presently focused on three classes of problems:
Inquisitive Virtual Agents
Today, a range of chatbot technologies are generating tremendous interest related to providing customer service. Most of these frameworks essentially enable natural language queries to be matched to documents that address them. Our research addresses two critical business needs:
- Scaling Agent Content
With thousands of products and services in constant flux, businesses frequently lack ready documents to match queries in a chatbot. We are developing new approaches to mine repositories of past customer-agent interactions to accelerate the process of extracting and preparing the content necessary for chatbot customer service interactions.
- Advanced Service Dialogs
Many queries do not match a specific, pre-existing document to be served by a chatbot. Instead, they call for an agent to have a diagnostic conversation. We are developing approaches to extend commercial chatbot technologies to enable the more complex dialogs required by specific use cases.
Intelligent Risk Monitoring
Identifying, assessing and mitigating risks is critical to success. Increasingly these risks leave data traces that can alert us to possible challenges. By developing models of different project categories and their associated risks, we then analyze a combination of structured and unstructured data sources to provide project managers with risk alerts.
Collaborative Content Synthesis
Whether its retailers identifying new product trends, hospitals preparing claims for insurance companies, or procurement professionals developing profiles of potential vendors to countless other examples, we find an important underlying theme: The recurring need to examine vast quantities of content, identify relevant portions, and assemble them into a specific artifact, such as an insurance claim or vendor profile. Today’s largely manual approaches cannot scale to the rapid growth in content.
By combining technical approaches including machine learning and natural language processing, and rethinking the best ways for people and systems to work together, we are developing systems that wade through enormous repositories to identify relevant content, and rely on human expertise to confirm the relevance and tune the models.