Paper Info
Title
Content selection from an ontology-based knowledge base for the generation of football summaries
Abstract
We present an approach to content selection that works on an ontology-based knowledge base developed independently from the task at hand, i.e., Natural Language Generation. Prior to content selection, a stage akin to signal analysis and data assessment used in the generation from numerical data is performed for identifying and abstracting patterns and trends, and identifying relations between individuals. This new information is modeled as an extended ontology on top of the domain ontology which is populated via inference rules. Content selection leverages the ontology-based description of the domain and is performed throughout the text planning at increasing levels of granularity. It includes a main topic selection phase that takes into account a simple user model, a set of heuristics, and semantic relations that link individuals of the KB. The heuristics are based on weights determined empirically by supervised learning on a corpus of summaries aligned with data. The generated texts are short football match summaries that take into account the user perspective.
Year
Venue
Keywords
2011
ENLG
data assessment,ontology-based knowledge base,domain ontology,numerical data,user perspective,ontology-based description,content selection,football summary,extended ontology,simple user model,main topic selection phase
DocType
Citations 
PageRank 
Conference
12
0.80
References 
Authors
17
3
Name
Order
Citations
PageRank
Nadjet Bouayad-Agha120620.28
Gerard Casamayor212112.47
Leo Wanner338965.54