Case Study


NLG Software is helping comfort the parents of babies in neo-natal intensive care

Arria NLG Chief Scientist Ehud Reiter has this vision for NLG in the healthcare field: “One day soon, NLG will be a great empowering tool for people in everyday life, giving them understandable access to medical information about themselves in a form and language that is appropriate for them.”

The BabyTalk research project, an ambitious undertaking of the University of Aberdeen, the University of Edinburgh, others, and a team that included Arria NLG Chief Scientists Ehud Reiter and Yaji Sripada (working as faculty members of the University of Aberdeen), is a strong proof point in the healthcare vertical and shows the potential NLG technologies have to deliver value in an environment as mission critical as a Neonatal Intensive Care Unit (NICU) at a major hospital.

The BabyTalk case study is included as a case study because it is valuable in the process of understanding the technology. The intellectual property (IP) associated with the BabyTalk Research Project is the property of the University of Aberdeen. Arria NLG Chief Scientists remain active in the BabyTalk project as academics while Arria NLG staff provide ongoing support services for the BabyTalk modules that are active in the NICU.

See an example of the BabyTalk NLG output below

BabyTalk: In the Neonatal Intensive Care Unit (NICU), babies are connected to sensors and systems that electronically monitor, record, report and display the condition of each baby. Doctors, nurses and other medical staff are trained to interpret and use this data to aid them in the care of the baby. The data is also used by medical staff to generate a variety of reports such as shift changeover reports and reports to the parents of the baby.

For the BabyTalk project an early version of NLG software was built by the University to: a) embody the expertise of both the medical specialists and the nurses who attend to the babies, b) accept input data from the NICU monitoring systems and c) automatically write three kinds of reports, for three different purposes, from the same digital input data. BabyTalk involved three modules:

BabyTalk Doctor: This module generated written summaries of clinical data over a relatively short period—roughly 45 minutes—to support decision making by doctors and nurses.

• BabyTalk Nurse: This module generated partial shift summaries, focusing on important events related to a baby’s respiration and circulation, in addition to background information, current status and potential problems. Handover is a complex process, and ineffective handover can endanger patient safety. There is no established best practice for handover. BabyTalk Nurse showed that the handover problem is solvable. It showed that an NLG computer system which automatically generates understandable and helpful shift summary texts from a complex, state-of-the-art patient information system holding a large amount of heterogeneous data is viable.

• BabyTalk Family: This module generated easy-to-understand, plain English reports on the medical condition of babies in neonatal care for families. These reports are available online to the infant’s parents, providing a simple summary of their child’s progress. When a newborn baby is admitted to a NICU, parents are frequently overwhelmed by the experience. The neonatal environment in which their baby is looked after can cause feelings of worry, confusion, and helplessness. Parents would often like more information about what is happening to their baby: the baby’s current weight, oxygen levels, milk feeding quantities, and so on. This, coupled with understanding, enables parents to adapt and cope with the situation. This sort of information is important because it helps parents to take on their parental role, as well as get involved with the care of their child’s delicate emotional state. BabyTalk Family solves this problem. It extracts the key facts of interest to parents (which are distinct from the key facts of interest to doctors and nurses), and generates a textual report which humanises the medical facts for parents, presenting them in a way which is both understandable and sensitive to parents. It automatically generates these easy-to-understand reports and makes them available to the infant’s parents online, thus providing easy and immediate access to a simple summary of their child’s progress.


Increasing volumes of data can
overwhelm medical practitioners


The availability of continuously monitored physiological data (e.g. heart rate) and detailed records of interventions and observations raises the expectation that medical staff will make best use of them to improve patient care, avoid errors and keep patients informed. If this expectation is to be realised, the recipients must not become overwhelmed by the volume of the data; it must be presented in a way that can be easily assimilated and understood.



Textual summaries of data can be more
effective than visualisations


Currently, computer-generated graphs and/or tables are often used to present large data sets to clinicians. However, the University of Aberdeen and Edinburgh's Neonate project (the predecessor of BabyTalk) showed that doctors make better treatment decisions when shown a textual summary of clinical data than when they are shown a visualisation of the data, even though they say they prefer visualisations. This discovery helped initiate and inform the BabyTalk programme.1





1 Source: Journal of Clinical Monitoring and Computing, 19, pp 183-194 (2005).