Dr Yaji Sripada Senior Lecturer Dept. of Computing Science University of Aberdeen yaji dot sripada #AT# abdn dot ac _dot_ uk

Research

I am interested in Natural Language Generation (NLG) or Text Generation. NLG is a sub-field of Computing Science/Artificial Intelligence/Natural Language Processing focusing on developing computational models for the task of natural language production. Traditional NLG models expect input that is semantically well grounded and map input semantic structures to output natural language text. However, it is useful not to place any restrictions on the input and accept input data as it is available in a domain. For example, consider developing an NLG model for generating a textual weather forecast (as seen in newspapers) from Numerical Weather Prediction (NWP) data, direct output of a numerical weather simulator. Weather patterns in NWP are not explicit and therefore traditional NLG models need to be extended to first interpret input data to meaningful weather patterns before more traditional NLG modules take over. I am interested in developing such extended NLG models that are called data-to-text NLG models.

Data-to-text NLG models address a similar problem to Information Visualization (InfoViz) – communicating electronic information to humans. I am interested in extending NLG models to design information presentations that combine information visualizations and natural language text. An important aspect of modern information visualization techniques is that they are organized in terms of the input data types and the interactive tasks performed by users. Organizing NLG systems too based on their input data types and defining interactive tasks for the NLG context are interesting. Most of the datasets I worked with to date have been either time series or geospatial or spatio-temporal. Geospatial data are visualized in GIS (Geographical Information Systems) and I am particularly interested in adding textual presentations to a GIS environment.

A common feature of data-to-text NLG models is that they summarise input data. I use data mining/analysis techniques for data summarization. As a related field I am also interested in Visual Analytics.

• Publications - List of Papers
• Current Projects -
• Informed Rural Passenger
• Affecting People with Natural Language
• Completed Projects
• Atlas.txt
• CBR and Text Revision
• BabyTalk
• RoadSafe
• ScubaText
• SumTime
• Current PhD Students
• Donald Matheson
• Konstantinos Papengelis
• John Paul Varghese
• Completed PhD Students
• Ibrahim Adeyanju
• Feng Gao
• Ross Turner

Commercialization

I am a co-founding director of a university spin-off, Data2Text Ltd building NLG software systems for several different industries.

Teaching

• CS4031 – Data Mining and Visualization
• CS5012 – Data Mining and Visualization
• Undergraduate Projects - Level 4

Admin

• Director of Undergraduate Teaching
• College Teaching and Learning Committee Representative
• Adviser of Studies
• Degrees in Science
• Course Index by Subject
• Computing Science Student Handbook - General Information
• CAS Scale

Contact :- Room 226, Meston Building, King's College, Aberdeen, AB24 3UE, Scotland, UK, Phone +44 (0)1224 272597, Fax +44 (0)1224 273422.