Faculty Publications and Conference Presentations

從「微軟小冰」出版詩集到生成式人工智能的崛起，人工智能參與詩歌創作的問題已有不少討論。然而，生成式AI的應用不僅侷限於創作層面，其還可以作為現代詩教學的輔助工具，為詩歌教育提供新的可能性。筆者近年任教坊間的一些現代詩寫作班時，曾嘗試引入生成式AI輔助教學。本文將結合個人教學經驗，淺談 生成式AI在現代詩教學中的應用潛力，並對其未來發展作出一些展望。

本研究以香港高校一门中文传意课程中粤语母语学生的普通话产出为材料，分析影响「言语可懂度（speech intelligibility）」（参 Smith & Nelson, 1985；Munro & Derwing, 1995）的语音要素。针对 19 位发音人 126 个可懂度低的音节，研究发现：（1）单一成分偏误中，韵母偏误占比最高，其次是声母偏误；（2）多种成分偏误组合中，「声母 +韵母」偏误占比最高；（3）声母偏误中，sh 误读最为多见；（4）韵母偏误中，i 误读最为多见，含介音韵母误读亦较多；（5）声调偏误中，第四声误读最为多见，与其他三声相混，二、三声也有相混情况。研究表明，在以传意为目标、面向粤语母语者的普通话教学中，可优先教学声韵母，声调教学稍列其后；重点及优先教学舌尖后音与舌面音（尤其是声母 sh 与 x）、声母 r、单韵母 i 与 u、以及含介音的音节，并围绕第四声进行声调教学。本研究为探索「大华语」（陆俭明，2005、2015；李宇明，2017）的语音「共核」（Common Core）（祝晓宏，2019）提供了实证研究的材料与观点，同时为以传意为目标的普通话语音教学观提供了理据。 This study investigates the Putonghua pronunciation of Cantonese-speaking students in a Chinese communication course at a Hong Kong university, focusing on the phonetic elements that impact “speech intelligibility” (Smith & Nelson, 1985; Munro & Derwing, 1995). The analysis of 126 low-intelligibility syllables produced by 19 speakers revealed that (1) Among single-component errors, vowel errors were the most prevalent, followed by consonant errors; (2) In error combinations, “consonant + vowel” were the most common; (3) Within consonant errors, mispronunciations of “sh” occurred most frequently, followed by errors involving “h,” “x,” “r,” “j,” and “zh”; (4) For vowel errors, “i” was the most commonly mispronounced, particularly in vowels with medial glides; (5) Tone errors were dominated by mispronunciations of the fourth tone, with occasional confusion between the second and third tones. The findings suggest that in Putonghua teaching aimed at Cantonese native speakers with communication as the goal, teaching initials and finals could take precedence over tones, focusing on teaching apical post-alveolars and alveolo-palatals (especially the initials “sh” and “x”), the initial “r”, the monophthongs “i” and “u” and syllables containing medial sounds. Tone teaching should revolve around the fourth tone. This research provides empirical insights and perspectives for exploring the phonetic “Common Core” (Zhu, 2019) under the context of Global Chinese (Lu, 2005, 2015; Li, 2017).

隨著生成式人工智能工具的風潮席捲全球，關於如何革新現有的教育方式和未來教育走向的探討，成為了近年來的熱門話題。香港科技大學鼓勵教學人員使用生成式人工智能工具，語文教育中心在2023年初已經開始了相關的探索。本文僅以筆者負責設計開發的中文核心課程「職場實務中文運用」為例，探討以ChatGPT為代表的GenAI工具在實務中文寫作教學中的運用，並結合學生對在課程中使用ChatGPT效果的反饋進行反思，展望課程以及實務中文教學未來的發展方向。

《詩學綱要》是陳去病（1874－1933）重要的詩學及詩歌史著作。此書本是詩歌史教材，系統論述了中國詩歌自上古至清代之發展與流變，亦旨在匡正時人論詩之流弊。本文集中考察其對宋代與明清詩史的論述，探討陳氏之詩歌史觀及其時代意義。陳氏基於儒家詩教，以性情與氣運為詩歌價值評判之標準；同時又主張「詩以異為體」，強調呈現各代詩家獨特性及差異性之必要。陳氏批評晚清「宗尚蘇、黃」之風氣，持尊唐抑宋立場，延續南社宗唐派對宋詩派之排斥。然其不同於柳亞子（1887－1958）之激烈攻詆黃庭堅（1045－1105），在肯定黃氏師古而能自成一家的同時，通過質疑呂本中（1084－1145）《江西詩社宗派圖》所建構的江西詩派譜系，回應宋詩派的詩歌史話語策略。在明清詩史部分，陳氏持崇明抑清立場，從性情出發，前則肯定復古派嘗試重建文化盛世之氣象，後則表彰明末幾、復諸子之風骨，同時感歎明詩發展遭逢厄運，展現出其對政治盛世下文化理想圖景的嚮往

<p>With the growth of education in multilingual setting, language teachers who teach in multilingual settings are now facing challenges in adapting new linguistic environments where multiple language contacts take place. It is therefore necessary to educate pre-service and in-service teachers on how multilingualism influence language learning, and one of the key areas is L2 pronunciation teaching, in which teachers could conduct their own mini-research, or even guide their students to conduct ones, to facilitate their understanding and acquisition. Traditional phonetic research relies mainly on sound collection which requires labour forces and equipment. Conventionally, to obtain sound samples from/under a controlled environment, good quality microphones are essentials for both on-site and off-site data collection (Munro and Derwing, Pronunciation in Second Language Learning and Teaching Proceedings, 11(1), 2020) and the assistance from other research team members is needed for off-site collection. However, with the development of the Internet, there have been multiple studies using online sources for phonetic research (e.g. Sewell and Chan, English World-Wide, 31(2):138–161, 2010; Chan and Chan, Journal of Universal Language, 22(2):1–44, 2021). Internet-based corpora are usually composed of sound clips recorded under uncontrolled environments with varied recording quality. Studies using Internet-based corpora have demonstrated the advantage of having an easier data collection process, especially under the pandemic era when on-site data collection is sometimes not possible. Internet-based corpora are also effective in long-distance data collection for individual researchers with limited help. It is surely one interesting way for (pre-service and/or in-service) teachers to conduct their own research in order to deepen their understanding in multilingualism. Despite the obvious advantages, researchers may still question the validity and quality of such data source in phonetic studies. Therefore, this chapter aims to suggest an interesting plan for teacher educators, in which (pre-service and/or in-service) teachers could conduct their own research in L2 phonetics and phonology by creating their own corpora with online sources or using existing well-established spoken corpora such as the Asian Corpus of English (ACE. The Asian corpus of English. Director: Andy Kirkpatrick; Researchers: Wang Lixun, John Patkin, Sophiann Subhan. https://corpus.eduhk.hk/ace/index.html, 2020). In that way, teachers could have a better understanding in how multilingualism may affect L2 pronunciation. This chapter will also investigate the challenges over these phonetic studies which utilised internet sources. In addition, recommendations on how to use these Internet-based corpora for future research opportunities will be discussed. This chapter aims to provide insights for researchers by examining the risks and potentials of Internet-based corpora, and suggesting some ideas of how to use such corpora for future language teacher education.</p>