Close
There is a large amount of unstructured data produced by different kinds of information systems, in a variety of formats. Electronic Health Record (EHR) systems store much information about patients in the form of free text notes. Existing text-mining methods aim to extract detailed structured information from clinical narratives. However, the inability to correctly identify and extract several contextualised aspects of clinical events from text (such as temporal information, negation and associating the correct experiencer to the right events) makes it difficult to understand how the extracted events are organised.
This project aims to
- explore some of the lacks in the major Temporal Annotation Guidelines by exploring alternatives to make the process of both: a) manually annotating temporal information in text easier, and b) automatic TIMEX annotation more flexible.
- describing theoretical insights, annotation schemes and corpora, empirical studies on processing and representing the meaning of negation, and applications that benefit from processing negation in clinical documents and explore the challenges posed by negation and propose solutions for an application identifying the surrounding points of improvement
- identify different forms of relationships between patients and relatives and map the distinct ways those relations are described in clinical records and before developing a NLP approach for Experiencer identification