research category image

Natural language & AI

Natural Language Processing and Artificial Intelligence technologies enable users to communicate with the digital technology they use everyday, whether for work or recreation.

research category image

Natural language & AI

Powering next gen interactions
Nuance has one of the largest Natural Language Processing and Artificial Intelligence groups in the world with over 140 researchers, most with advanced degrees. This group is distributed among labs around the world extending from Sunnyvale, CA to Montreal, Canada to several locations in Europe. The goal of the research conducted at these labs is to develop the next generation of intelligent, conversational agents powered by Nuance speech infrastructure. The staff has established research reputations in linguistics, dialog systems, computational linguistics, AI, question answering, natural language understanding, and machine learning, as well as systems building, testing and evaluation. These activities are organized into seven subgroups:

  • Linguistic Technologies
  • AI and Reasoning
  • Conversational Systems
  • Question Answering
  • Applied NLP
  • Application Research
  • Clinical Language Understanding in Healthcare

Linguistic Technologies
The Linguistic technologies group focuses on pragmatics, discourse and dialog processing; anaphora resolution; utterance generation; syntactic parsing; data driven methods; entity recognition; morphology; multilingual NLP; and semantic interpretation.

AI and Reasoning
The AI and Reasoning Group performs research in the areas of knowledge representation and reasoning (large scale knowledge bases, logic-based representation tools, knowledge and data integration methods), collaborative dialog systems, and probabilistic reasoning for intent recognition in dialog and ambiguity management in natural language systems.

Conversational Systems
The charter of the Conversational Systems team is to develop advanced conversational systems using components from the Linguistic Technologies and AI and Reasoning teams, other Nuance NLP teams, as well as the larger academic NLP community. Our ultimate goal is to create a mixed-initiative conversational experience with multi-turn dialogs powered by complex reasoning, using technology that is as independent as possible of any one domain.

Question answering
The ability to answer questions is at the heart of any successful conversational assistant. People often need more information before they can make the choices necessary to accomplish a task, and asking questions is how they get that information. Question answering draws on many NL/AI technologies, including parsing, semantics, knowledge representation, inference, problem solving, and dialog. One of the unique aspects of question answering as a task is that the necessary information is often expressed in natural language as well as formal language, requiring a combination of natural language inference and traditional inference techniques.

Applied NLP
The focus of the Applied NLP group is to create technologies that make it easy for application developers to quickly and intuitively build and tune natural language understanding or conversational systems. The group is also involved in the transfer and support of those technologies in Nuance products as well as in the support of many languages.

Application Research
The NLU Application Research team focuses on building scalable and customizable NLU and dialogue solutions, and developing tools and processes that make it easy to build such solutions.

Clinical Language Understanding in Healthcare
The Clinical Language Understanding group works on developing and applying natural language processing algorithms to extract and interpret medical concepts and their relations from medical records which are a mixture of narrative, free-text as well as structured data. This research goes beyond Named Entity Recognition, extracting concepts which are described rather than named and encoding them to their canonical forms such as concepts in SNOMED CT or the Nuance proprietary Ontology LinKBase. Relation extraction is leveraged to keep related concepts together to create medical facts as well as to identify relationships between different facts, such as procedures meant to treat disorders vs procedures performed to diagnose disorders. The group also conducts research in structure detection in medical reports, detection of uncertainty, negation, temporal information, confidence estimation and rule based systems to interpret extracted information for various healthcare applications.

Explore recent publications by Nuance Natural Language and AI researchers.

Selected articles

DeepAAA: Automated Detection of Abdominal Aortic Aneurysms using Deep Learning

Untreated abdominal aortic aneurysms (AAAs) tend to grow and eventually may rupture with mortality rates exceeding 90%. As most of AAAs are asymptomatic until onset

Read more

DeepAAA: clinically applicable and generalizable detection of abdominal aortic aneurysm using deep learning

We propose a deep learning-based technique for detection and quantification of abdominal aortic aneurysms (AAAs). The condition, which leads to more than 10,000 deaths per

Read more

Fixing Comparative Preferences for SPARQL

Sometimes one does not want all the solutions to a query but instead only those that are most desirable according to user-specified preferences. If a

Read more

Contextualization via Qualifiers

A common method for contextualizing facts in knowledge graph formalisms is by adding property-value pairs, qualifiers, to the facts in the knowledge graph. Qualifiers work

Read more

WordNet for “Easy” Textual Inferences

This paper presents a WordNet-based automatic approach for calculating “easy” inferences. We build a rule-based system which extracts the pairs of the SICK corpus whose

Read more

Textual Inference: getting logic from humans

This paper describes a manual investigation of the SICK corpus, which is the proposed testing set for a new system for natural language inference. The

Read more

CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies

The Conference on Computational Natural Language Learning (CoNLL) features a shared task, in which participants train and test their learning systems on the same data

Read more

Universal Dependencies for Portuguese

This paper describes the process of converting the Portuguese Bosque corpus to the Universal Dependencies scheme version 2. The conversion was done by applying to

Read more

EXISTStential aspects of SPARQL (poster)

The SPARQL 1.1 Query Language permits patterns inside FILTER expressions using the EXISTS construct, specified by using substitution. Substitution destroys some of the aspects of

Read more

Building Predicates

This book provides a comprehensive analysis of the processes involved in word formation and the morphosyntax of predication that will appeal to anyone interested in

Read more

1 2 3 67