Creating comics from automatic summarization of sports stories

Li, Tao (2023). Creating comics from automatic summarization of sports stories. University of Birmingham. Ph.D.

Preview

Li2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (5MB) | Preview

Abstract

Automatic text summarisation offers a means of managing information overload by presenting the critical points in documents. This thesis proposed to improve this process by visualising such summaries. Sports match reports from websites are summarised in the form of comics. Instead of using machine learning to support summarisation, we used the story grammar to support automatic summarisation. We will define a sports match's 'story grammar' (in terms of events, such as scoring a goal, penalising a foul, etc., and Named Entities, such as teams, players, referees etc.). It frames the extractive-summarisation process. Then we visualised the summary into a comic. We conducted two user trials to evaluate the summary. In the first user trial, participants read reports of soccer games in 3 different conditions. Then they answered comprehension questions. The results showed that the participants read the comics more quickly than the text, resulting in superior comprehension (literal, inferential, and evaluative), and the participants preferred to choose comics as reading material. In the second study, we designed an eye-tracking experiment to explore reading strategies for different media. While reading time is faster for comics, the average fixation time per word is the same across all media. It suggests the texts were skim-read, which might explain the lower comprehension of the text. Surprisingly, the comic is highly supportive of evaluative and inferential comprehension. Overall, the comic produced significantly better performance on comprehension tests in both studies. As we hypothesised, comics could significantly reduce time consumption and improve comprehension. Therefore, we believe that the comic helps users access important information.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):