This event took place on 8th March 2006 at 9:00am (09:00 GMT)
Knowledge Media Institute, Berrill Building, The Open University, Milton Keynes, United Kingdom, MK7 6AA
Human text has very individual authorspecific characteristics or features. Even as authors write about identical topics within a genre, their text will be individual. The output of natural language generation (NLG) systems is, in contrast, more monotonous. To help readers understand automatically generated text better and make NLG outputs more acceptable, NLG systems should operate in a more ?human-like? fashion. We are currently developing a NLG system that can generate output based on the individual features of readers. To support a broad range of readers, we want to apply an approach that could analyse individual features automatically. In this paper, we discuss a few possible architectures that can be used for the implementation of our system. We use a specific domain - recipes - that we will investigate in our system.
Human text has very individual authorspecific characteristics or features. Even as authors write about identical topics within a genre, their text will be individual. The output of natural language generation (NLG) systems is, in contrast, more monotonous. To help readers understand automatically generated text better and make NLG outputs more acceptable, NLG systems should operate in a more ?human-like? fashion. We are currently developing a NLG system that can generate output based on the individual features of readers. To support a broad range of readers, we want to apply an approach that could analyse individual features automatically. In this paper, we discuss a few possible architectures that can be used for the implementation of our system. We use a specific domain - recipes - that we will investigate in our system.
Click below to play the event (27 minutes)