Towards realistic verbal interactions between people and computers-a probabilistic approach. This project aims to facilitate natural spoken interactions between people and computer systems, addressing obstacles to the acceptance of these systems. We will investigate computational models for relevant aspects of spoken dialogue, which will be implemented in computer systems for diverse tasks (for example, home devices and phone-enabled services).
Explaining the outcomes of complex computational models. This project aims to develop new algorithms that automatically generate explanations for the results produced by complex computational models. In recent times, these models have become increasingly accurate, and hence pervasive. However, the reasoning of Deep Neural Networks and Bayesian Networks, and of complex Regression models and Decision Trees is often unclear, impairing effective decision making by practitioners who use the results o ....Explaining the outcomes of complex computational models. This project aims to develop new algorithms that automatically generate explanations for the results produced by complex computational models. In recent times, these models have become increasingly accurate, and hence pervasive. However, the reasoning of Deep Neural Networks and Bayesian Networks, and of complex Regression models and Decision Trees is often unclear, impairing effective decision making by practitioners who use the results of these models or investigate the decisions made by the systems. Practical benefits of clear decision making reasoning by complex computational models include reduced risk, increased productivity and revenue, appropriate adoption of technologies including improved education for practitioners, and improved outcomes for end users. Significant benefits will be demonstrated through the evaluations with practitioners in the areas of healthcare and energy.Read moreRead less
Language engineering in the field: preserving 100 endangered languages in New Guinea. Efforts to preserve the world's endangered linguistic heritage are labour-intensive, and unable to keep up with the pace of language loss. This project investigates a new approach to language preservation, using techniques from language engineering, and leveraging the labour of mother-tongue speakers.
Information access through web-scale question-answer pair finding, ranking and matching. This project will aim to take web search to a new level of sophistication in accepting queries in the form of complex natural language questions, and returning a ranked list of natural language answers automatically extracted from a broad range of web user forums.
Personalised topic modelling and sentiment analysis for enhanced information discovery over document streams. This project will develop personalised information discovery, navigation and management systems of online content for the creative industries, e.g. to help advertising agencies understand market trends, and enable designers to discover and analyse information relating to new product concepts.
Natural language processing for automated validation of protein databases. The project aims to use natural language processing and information retrieval to reconcile and improve sources of biological information. Biological research has produced vast volumes of information about proteins, captured in structured resources (databases) and unstructured documents. However, the accuracy of much of this information is questionable. The project proposes to develop methods to validate data and reduce th ....Natural language processing for automated validation of protein databases. The project aims to use natural language processing and information retrieval to reconcile and improve sources of biological information. Biological research has produced vast volumes of information about proteins, captured in structured resources (databases) and unstructured documents. However, the accuracy of much of this information is questionable. The project proposes to develop methods to validate data and reduce the dramatic inconsistencies in protein information resources by leveraging observed correlations and complementarity between them, and specifically through targeted fact extraction from the biomedical literature. These methods will be applied at scale across millions of published articles, to infer and validate functional information.Read moreRead less
Learning Deep Semantics for Automatic Translation between Human Languages. This project seeks to integrate deep linguistics and deep learning to improve translation quality. The modern world relies increasingly on automatic translation of human languages to deal with billions of documents. Current translation systems struggle with complex texts and often produce misleading or incoherent outputs. Furthermore, they translate sentences independently and ignore their overall document-wide context. T ....Learning Deep Semantics for Automatic Translation between Human Languages. This project seeks to integrate deep linguistics and deep learning to improve translation quality. The modern world relies increasingly on automatic translation of human languages to deal with billions of documents. Current translation systems struggle with complex texts and often produce misleading or incoherent outputs. Furthermore, they translate sentences independently and ignore their overall document-wide context. This project seeks to address these issues by developing a new approach using semantics – the underlying meaning of the text – to drive translation, both as discrete structures and continuous representations learned via deep learning. This may improve translation quality, thereby improving automatic translation for end-users.Read moreRead less
Adaptive Context-Dependent Machine Translation for Heterogeneous Text. While automatic machine translation technologies are undoubtedly useful to a wide range of users, they often produce incoherent outputs for many types of input, for example, medical, literature, or even conversational text. This project will develop new adaptive machine translation systems to handle many domains and text styles, including heterogeneous mixed-domain inputs. It will develop multi-task machine learning methods f ....Adaptive Context-Dependent Machine Translation for Heterogeneous Text. While automatic machine translation technologies are undoubtedly useful to a wide range of users, they often produce incoherent outputs for many types of input, for example, medical, literature, or even conversational text. This project will develop new adaptive machine translation systems to handle many domains and text styles, including heterogeneous mixed-domain inputs. It will develop multi-task machine learning methods for training collections of domain-specific translation systems while leveraging correlations between domains. This approach will reduce the big data requirements of current translation systems, and improve translation quality across a wide range of different language pairs and application domains.Read moreRead less
Responding to requests and situations in assistive computer systems - a decision-theoretic approach. This project aims to enable computer agents to respond appropriately to people's spoken requests and circumstances (e.g., ask questions or perform actions). This project will investigate computational models for response generation, which will be implemented in assistive computer systems, thus enabling people to interact more easily with these systems.
Spoken conversational search: contextual interactive techniques to support effective information search over a speech-only communication channel. This project will develop new techniques for effective information search using speech only, supporting improved information access for visually impaired people or in situations that require focused visual attention (e.g. driving). The techniques are based on a conversational approach to information search and presentation of results.