Speaking of Language Technology

Abstract

This dissertation investigates speech technology and educational technology (EdTech) in relation to second language (L2) learner pronunciation. As the speech technology and EdTech fields evolve and impact of millions of people internationally (Chen, 2015; Google, 2019; Singer, 2017), this study aims to fill gaps in knowledge about how these novel tools affect L2 learners and compare to traditional L2 pedagogical practices. In this pretest-posttest study, the experiences of 134 L2 French beginning learners are investigated in relation to ASR and virtual reality (VR) technology, ImmerseMe (Cardwell, 2017). Through ASR and VR technology, L2 learners can practice speaking the L2 through the software and potentially improve their L2 speaking skills. ImmerseMe is currently licensed to 24,000+ language learners in New Zealand and Australia. In this study, participants used the software for a 5-week period. Before and after usage of the software, participants provided demographic, survey, interview, speech, and backend analytic data. The overall study is segmented into three articles to explore the subtopics of 1) EdTech ASR, VR technology and time spent speaking the L2 as compared to typical practice, 2) participant comfort levels speaking in ASR and L2 classroom contexts, and 3) L2 French pronunciation improvement and beliefs of improvement through usage of the software. Article 1 considers how much time L2 French learners spend speaking the L2 outside of class with traditional pedagogy and how VR technology ASR EdTech software can increase the time spent on task practicing the L2. Typically, L2 learners practiced speaking infrequently outside of the L2 classroom, but through usage of the ImmerseMe software, participants spoke French for an estimated average of 114 minutes during the five weeks of the study. Across all of the participants, the range was between 0-10 hours. These findings support how EdTech software of this kind can increase L2 speaking practice and complement traditional pedagogy in the classroom. Article 2 examines L2 learner comfort levels speaking in L2 classroom and ASR contexts. Anxiety has been documented as detrimental to L2 acquisition (Horwitz, 2016). To consider whether ASR EdTech is a comfortable setting for L2 speaking practice, this article compared reported comfort levels in different contexts through repeated Chi-Squared tests. While smart speakers and ASR are becoming more prevalent in society, 65% of participants reported using typical ASR relatively infrequently, a few times a month or less. Findings suggest that speaking to ImmerseMe was comparably comfortable to speaking to an L2 classmate, L2 instructor, and to the L2 classroom. ImmerseMe was found to be a more comfortable context than typical ASR. However, speaking to a native speaker or typical ASR were found to be more anxiety-producing settings. The low-anxiety environment bolsters support for ASR EdTech to be a comfortable context for learners to practice speaking in a private setting. Article 3 assesses whether participants improved their L2 pronunciation through usage of the software and whether they believed that they were improving their L2 speaking skills. Whether learners perceive that they are improving in language study contributes to whether they will continue to study (Dugartsyrenova & Sardegna, 2017). Examining the often studied L2 vowel contrast /y/ and /u/ (Flege, 1987; Levy & Strange, 2008; Liakin, Cardoso, & Liakina, 2015, 2017b; Strange, Levy, & Law, 2009), pretest and posttest participant production were compared. Through 2x2x5 and 2x2x4 ANOVAs tests analyzing F2 Barks values, findings indicate that while no significant improvement was found in L2 French learner production of /y/ and /u/ vowels, most participants (79%) believed they were improving their L2 speaking skills through using the software. While participants produced the /y/ and /u/ vowels with distinct F2 values for both pretest and posttest values, these values were still substantially different from native speaker norms. Interpretations of these data include that the five-week period was not enough time to see improvement, but also that participants may have improved in other ways besides this vowel contrast, i.e. speech rate, fewer pauses, and intonation patterns. As these beginning L2 learners perceive that they improve, this belief in their developing skills may help them invest more time learning the language overall and add to how much time they spend studying the language. Together, these three articles enrich research in the fields of speech technology, L2 EdTech, and L2 pronunciation. Conclusions frame findings not only in how software can complement traditional pedagogy, but also how ASR can be a powerful tool to integrate into software generally. This research might be of interest to L2 pedagogues, learners, instructors, administrators, curriculum designers, human language technologists, technology consumers, researchers, and designers of products worldwide.