Summaries are designed to improve access to electronically-encoded patient records, by offering a quick view of a patient's history. Various types of summaries are being constructed, including a longitudinal summary, summaries of diagnoses, investigations and interventions. Having the same data summarised from various viewpoints helps identifying patterns in the history that may have been missed otherwise.

• The Chronicle
  
  

  Patient records consist of hundreds of documents that record a patient's illness history over a number of years and from various perspective. A lot of the information in these documents is duplicated, some is contradictory due to erros in recording, while some potentially useful information is not explicitely stated at all, but can be inferred.
  Consider that several different clinicians may choose to record the specific diagnosis of a given patient, on different days over a possibly long time period, for example in the introductory paragraphs of several clinic letters. Each clinician may use a slightly different degree of precision in stating the diagnosis. One caricature of current electronic patient records, composed as they are from such documents, is that they serve a primarily medicolegal requirement to faithfully record who said what and when.
  Two fundamentally different views of the story of the patient illness, inspired by clinical rather than medicolegal interests, include what was known and how and what was done and why. To reconstruct such views from the medicolegal record, repetitious statements must be hidden such that diagnoses and procedures are described only once but with maximal precision. Further, clinically important information that was regarded as obvious and so never recorded - such as why treatments were given - must be made explicit. Expressed as a semantic network, this view of the patient story is termed the Chronicle.

• Summary generation from the Chronicle
  
  

  A Chronicle is essentially a semantic network, and thus provides a valuable resource for generation. We make extensive use of template relations in both content selection and discourse structuring.

Please address any problems or questions about this page to