Game Lecture: Albert Gatt

Our second Game Lecture in April features Albert Gatt presenting the topic of natural language generation, a method of summarising data (such as game playtraces) into human-readable language. The lecture is on Tuesday 19th of April 2016 at 7:00 pm, and is held at the Institute of Digital Games.

The topic and summary of the Game Lecture follows:

"Automatic Generation of Text from Data: From Informing to Narration"

Natural Language Generation (NLG) systems, which generate text from non-linguistic data, are typically designed for the purpose of summarising such data for human consumption. Yet, in all but the simplest texts, users of such information are often found to require more than mere factual information. Events discovered in data and reported in a text are ultimately related temporally, but also linked rhetorically in complex ways, depending on the purpose for which a text is produced. This raises the question whether we should see NLG as producing narratives.

In this talk, I will give a brief overview of NLG, focussing specifically on the branch known as 'data-to-text'. I will show how certain developments in NLG recall theoretical developments in the literary theory of narratology. They also link up with insights from computational narratology. This sheds a somewhat different light on the nature of NLG in general: I will argue that NLG can in many instances be thought of as a technology aimed at generating narratives, whether the narrative world is a fictional one, or whether it is rooted in real-world information. In the final part of the talk, I will suggest possible ways in which this perspective can also benefit research in digital games, specifically, where such research incorporates an explicit narrative, recounted linguistically or in a mixture of modalities.