Research Output

Abstract: In today's technology-driven era, humanities are evolving with emerging technologies. To promote the humanities in an engaging way, a university in Hong Kong attempted to develop a digital playhouse (DP) using big data and generative AI. The DP project aims to create an interactive virtual learning platform with eight themes to promote sustainable teaching, research and knowledge transfer in the humanities. It targets the public and addresses personalised learning needs at varied ages and is divided into three main categories: game-based education, interactive art appreciation, and a database encompassing Chinese textbooks, local stories, and an AI-generated dictionary. Although the DP is still in construction, the project team intends to use big data technologies to capture and analyse learning behaviours, aiming to identify the evolution of learning pathways, characteristics of interaction patterns, variations in cognitive styles, and drivers of learning motivation. Using Theatre One as an example, where players can accumulate herbs and obtain access to prescriptions, this paper will illustrate characteristics of the DP and demonstrate how the backend system functions to assess the players’ accomplishments, enabling the project team to determine necessary adjustments in difficulty, showcasing a data-driven design and its support for learner customisation and platform maintenance. Such design helps ensure optimal educational outcomes and system stability. In gist, the DP brings new perspectives to the development of disciplines and promotes interdisciplinary research through the integration of knowledge and innovation. It will inspire future advances in humanities education advancements and serve as a reference for similar projects.

Abstract: “Cantonese Master Challenge” is a public-facing online quiz game on Google Forms centred around the Cantonese language. Its twofold purpose is to popularize knowledge about Cantonese linguistics and collect research data using a gamification approach. Launched on March 28, 2022, the game attracted over forty thousand participants in around one and a half months. This article, after introducing the background and design of the game, reports on its multifaceted achievements, including: (1) serving as a widespread popularization of Cantonese linguistics; (2) examining the results of past popularization efforts; (3) facilitating evaluation of whether certain linguistic tasks are suitable for crowdsourcing; (4) obtaining a large amount of quantitative data for specific research topics; (5) exploring public perceptions of Cantonese and linguistics. This attempt to integrate and gamify academic knowledge popularization and crowdsourced data collection was quite successful, possibly marking the largest Cantonese linguistics popularization event to date, and generating extensive data that offers insights into a wide range of topics.

Abstract: There is a widely observed increase in Cantonese vernacular writing in Hong Kong, which may point to the strengthening of a local (linguistic or political) identity. Linguistic data reveal that the development of vernacular writing is not a one-way increase of localness in writing driven by a homogenous group, but the result of competing writing norms created and upheld by local users with subtle ideological differences. This chapter identifies two distinct practices of Cantonese vernacularisation that align with the distinction between translanguaging and code-switching. The first form of Cantonese vernacularisation (through translanguaging) is an act of breaking the existing boundary between the written standard and the vernacular by exerting Cantoneseness in writing. The second form, (through code-switching), on the other hand, maintains two separate codes, with “hardwired” school-taught vernacular-to-standard conversion rules being either fully upheld or suppressed. The writings generated via the two practices differ in word choice, linguistic structures, formality and genre distribution. This divergence is attributed to users holding different views on the ideals and dynamics of the spoken and written lects. This chapter explains this divergence through individual differences in perceived psychological distance between spoken and written lects, and proposes an equilibrium that allows the coexistence of a strengthened translanguaging-induced diglossia and a new endoglossic standard.

Abstract: Hong Kong lacks an official set of standardized Chinese character glyphs. Despite the convergence of multiple handwritten variants to one agreed form for many characters over the past years, individual variation still exists in many cases. There is a trend in recent years in primary school Chinese education to use the standard glyphs found in List of Graphemes of Commonly-used Chinese Characters as the correct form, and rejecting other character variants. This is however problematic, due to this list not being timely revised, with many of its requirements for strokes and components deviating from common usage, which negatively impacts the teaching of Chinese literacy. This study aims to review the basic principles of judgment in the practice of glyph selection, systematically extract various glyph issues based on existing lists and tables and discuss directions for narrowing the gap between the designated standard glyphs and common practices. The chapter categorizes the issues based on the relation between the standard glyph and the variant(s) for each character, which fall into four major types: difference in character selection, difference in structural/component composition, difference in stroke/writing, and component disambiguation, with a total of 180 identified issues. These issues were then further sorted, based on character frequency, the number of affected characters, and distribution in online datasets, into 66 items as investigation points for handwriting preference. This list will provide a foundation for future construction of a new character glyph standard.

Abstract: This paper addresses the current state of Cantonese proficiency assessments and introduces a new test aligned with the Common European Framework of Reference for Languages (CEFR). This new test fills gaps in existing evaluations by focusing on Cantonese oral proficiency for less advanced learners (CEFR levels A1, A2, and B1) without requiring prior knowledge of English, Written Chinese, or Mandarin. Its design also considers cultural neutrality and avoids assumptions about Hong Kong-specific knowledge, emphasizing oral communication without bias towards accent or precise pronunciation. Conducted via one-to-one video conferencing, the test features a variety of tasks to assess interactive communication skills. This test aims to meet the sociolinguistic needs of diverse Cantonese learners globally and to encourage further promotion and development of relevant learning materials.

Abstract: A public service built on a transformer-based text-to-speech engine for two endangered languages in Hong Kong: Waitau and Hakka, with g2p conversion and word segmentation support.

Abstract: A companion website showcasing stories in Waitau and Hakka collected from communities in the Hong Kong countryside, with multimedia representations including recordings, animations and illustrations.

Abstract: This paper presents a linguistically-informed, non-machine-learning tool for classifying Written Cantonese, Standard Written Chinese, and the intermediate varieties used by Cantonese-speaking users from Hong Kong, which are often grouped into a single “Traditional Chinese” label. Our approach addresses the lack of textual materials for Cantonese NLP, a consequence of a lower sociolinguistic status of Written Cantonese and the interchangeable use of these varieties by users without sufficient language labeling. The tool utilizes key lexical markers identified from past linguistic research to determine whether a segment is Cantonese, Standard Written Chinese, mixed or unmarked. The task is reduced into string operations to allow for a flexible and efficient extraction of high-quality Cantonese data from large datasets mixed with Standard Written Chinese. This implementation ensures that the tool can process large amounts of data at a low cost by bypassing model-inferencing, which is particularly significant for marginalized languages. The tool also aims to provide a baseline measure for future classification systems, and the approach may be applicable to other low-resource regional or diglossic languages.

Abstract: A Cantonese phonemic keyboard designed as both a computer-assisted language learning tool and a daily application for Chinese input. Supports Cantonese/Chinese-as-a-second-language learners in Hong Kong and overseas heritage speakers with multilingual cues in English, Urdu, Indonesian, Hindi, and Nepali.

Abstract: A community project supporting Cantonese learning for ethnic minorities in Hong Kong and overseas heritage speakers. Created over 250 original and adapted stories as online learning materials, videos, and 40 storybook titles.

Abstract: A large-scale open-licensed Cantonese (monolingual and Cantonese-English) online dictionary database with over 400 volunteer editors. Provides word lists, translation data, and language resources for NLP and standardization.