Speakers
Description
Teaching trend-description language is an important part of IELTS Writing Task 1, but it is still commonly taught through vocabulary lists and model sentences. Students may remember words such as increase, decline, or significantly, but they often have difficulty combining these words in grammatically and collocationally appropriate ways. Teachers may also select language patterns mainly from commercial materials or their own experience, rather than from authentic language evidence. This article reports a classroom innovation that integrated corpus data, visual materials, and generative artificial intelligence to teach trend-description language to 16 Vietnamese university students at an intermediate IELTS level, with a target band of 5.5–6.5. The innovation followed five main stages. First, the teachers developed a small specialized corpus, called T1Corpus, which included 183 expert IELTS Writing Task 1 reports with a total of 35,107 words. A collection of the graphs related to these reports, called T1Visuals, was also prepared. Second, students received basic training in using AntConc and learned how to search for and interpret concordance lines. Third, they worked in groups to explore common verb-preposition, verb-adverb, and adjective-noun patterns. Students compared the corpus examples with the original graphs and made hypotheses about how particular language patterns represented starting points, ending points, amounts, and degrees of change. Fourth, students used ChatGPT to examine their hypotheses. They compared the AI explanations with corpus and visual evidence instead of simply accepting the answers. Finally, they completed controlled exercises and a writing task. ChatGPT was then used to reconstruct a graph from their written description, allowing them to check whether their language communicated the visual information clearly. Classroom observations and informal student feedback indicated several possible benefits. Many students became more active in exploring language and noticed patterns that they had not paid attention to before. The graph reconstruction activity was also engaging and helped reveal unclear or inaccurate descriptions. However, students showed different levels of interest in exploratory learning. Some found AntConc difficult to operate, while moving between corpus data, graphs, and AI responses created considerable cognitive demand. One student also preferred asking AI directly instead of examining corpus evidence first. The experience suggests that corpus inquiry, multimodal data, and generative AI can complement each other in IELTS writing instruction. However, the activities need careful scaffolding, sufficient technical training, and a clear “corpus-first, AI-second” sequence. AI should be positioned as a tool for comparison and reflection, rather than as the main source of linguistic knowledge.
