The Open University of Hong Kong
Hong Kong
Pictures are known to elicit thoughts and inspire writing. Picture-prompt writing is a teaching strategy that facilitates writing practices with specific ideas stemming from pictures. The visual content – such as objects, characters, scenes, and locations – can arouse the personal experience of writers and provide the ingredients about which to write. In practice, a language teacher usually selects some suitable pictures and attaches to them prompts or hints that direct the focus of students, who will then compose one or several sentences about the pictures. Research has indicated that picture-prompt writing helps to develop expressive writing proficiency in beginner language learners. This paper reports a feasibility study on the development of an online self-directed picture-prompt writing platform. Language learners should be better motivated with the capacity for picking their own photos to write about and to practise anywhere and at any time. Such a platform is a novel and interesting proposition for teaching any language.
Providing writing advice to language learners is crucial for the effectiveness of self-directed picture-prompt writing. During the writing phase, the platform should offer tips, such as examples of words related to the picture and the next words in sentences. When the writing has been completed, the platform should review the sentences, generate a performance report, and offer alternative picture descriptions. The former requires visual object recognition and the latter needs deeper natural language understanding, both of which had been very technically challenging until the recent advance in deep learning technologies. Deep learning, which includes a range of technologies such as the renowned AlphaGo, has driven the latest resurgence of artificial intelligence and has enabled many novel applications. This study adopts a build-and-test methodology to evaluate the extent to which state-of-the-art deep-learning technologies can be utilized for the implementation of self-directed picture-prompt writing.
A prototype online platform for a picture-prompt writing exercise was developed for evaluation. The findings can be divided into three aspects. The first aspect is about the feasibility of deep learning technologies for supporting various functions of the platform; and the second aspect focuses specifically on the performance of the two types of advice, including the accuracy of tips and review. The final aspect is about the preliminary perceived usefulness of the platform.
The online self-directed picture-prompt writing is a very attractive and novel proposition. Language education will receive a significant boost if this approach is mature. This paper demonstrates that state-of-the-art deep learning technologies have the potential to bridge the gap in the enabling technologies.