A Statistical Analysis of Award-Winning Speeches

The primary purpose of this research is to investigate statistical characteristics in the linguistic data of the speeches delivered by Japanese female university students at the annual intramural English oratorical contests sponsored by Mukogawa Gakuin, an educational institution in Hyogo, Japan.The original speech-contest corpus, which contains 80 speeches 1) in total presented in 2016-2018, consists of three groups of speakers: 62 in the entry round, 12 finalists, and six winners.The corpus was analyzed to highlight the statistical differences and similarities between these three groups, with a specific focus on the differences between the first group (entry round) and the other two groups (finalists and winners) .After the technical analysis, this study drew upon the literature on English speechmaking published in Japan to verify how the ideas and tips introduced for Japanese speakers were, or were not, reflected in the facts revealed in the statistical analyses of the corpus. BackgroundHolding English speech contests for students now appears to be a prevalent practice at various levels in English education (Nomura, 2009, p. 2) .Since speechmaking is part of speaking practice, preparing for the contest is considered a meaningful opportunity, especially to improve students English pronunciation (Kuwamoto, 2010, p. 106; Nishino & Nekota, 2014, p. 51) .It has been stated, however, that the content of the speech has recently come to matter more than the quality of pronunciation at contests (Nomura, 2009, p. 16) .As Mikuma (2003, pp.55-82) calls speechmaking a "social interface device," speeches with good content make people and society stay interconnected.Hayase (1992, p. 2) also claims that one of the purposes of making speeches is to develop personality, where speakers can broaden their views about daily lives, social issues, and various problems around them.Creating stronger content that can motivate people in society plays a substantial part in English speechmaking practice.The more people focus on the content, the more speakers need clear, specific guidelines that will help them to build compelling messages in their speeches.In order to help define and recognize good speeches, there are some studies on evaluation methods for speech contests (Terashima, 2000;Fukuda, 2018) and on analysis of individual speeches at a university-level speech contest (Underwood, 2003) .However, they do not discuss the importance of exploring the technical data behind the texts.Indeed, such detailed linguistic data cannot thoroughly be evaluated at speech contests, since such data can only be uncovered through statistical analysis.Although a statistical approach to analyzing the speech texts may supposedly provide clear, specific guidelines for making good speeches, it is unfortunate to have found few studies which conducted a technical analysis to show statistical proof of good speeches at contests.This study, therefore, analyzed the linguistic data of 80 speeches presented by students to clarify the statistical characteristics of the awardwinning speeches.Furthermore, the results were compared with the existing recommended data for speechmakers.The ultimate goal of this research was to seek out new specific numerical guidelines for making award-winning speeches, which had never been explored by the statistical approach before.1.