This session introduces a new AI-based speaking assessment platform designed to evaluate open-ended responses without requiring live interviewers. The system delivers consistent scoring, instant feedback, and full remote access.

Xreading is an online library that gives students access to thousands of graded readers and allows instructors to track their students’ reading progress. Since its launch in 2014, the system is being continuously being updated and improved with new features and more books. In this presentation, the founder of Xreading will explain the newest features and improvements, and review what is planned for the future, including our new AI powered “Book Chat” component.

With technology playing a growing role in language education, tools that adapt to learners’ needs can make practice more effective. AdaptLingo is a speech-based chatbot designed to help English learners improve their speaking skills through natural conversation. Learners speak directly to the chatbot, which listens to their speech and determines their English proficiency—beginner, intermediate, or advanced—and then adjusts its responses to match their level. Beginners receive simpler language and slower speech, while advanced learners engage in more natural, fluent conversation. To ensure responses are appropriate for each level, AdaptLingo selects words from a well-established vocabulary list suited to the learner’s proficiency, reinforcing level-appropriate language learning. The chatbot also adjusts its speaking speed to make conversations more accessible or challenging, depending on the user’s ability. By offering a fully speech-to-speech experience, AdaptLingo provides a personalized and interactive way for English learners to practice speaking in a low-pressure environment. This tool will be open-source, making English conversation practice more accessible to learners everywhere. AdaptLingo is a technological innovation that will be integrated into classroom environments to supplement students' curricula, supporting English learners from diverse native language backgrounds with additional speaking practice.

This study evaluates Synthesia, an AI-driven video platform, as a computer-assisted language learning (CALL) tool in Hong Kong university English courses. Focusing on its ability to address three challenges—differentiating materials for diverse classrooms, enhancing academic English engagement, and reducing instructor workload—the research involved 6 instructors and 120 students. Synthesia’s avatars and multilingual tools were used to create context-specific resources (e.g., academic presentation exemplars). Quantitative analysis of pre/post-speaking assessments showed moderate fluency improvements (Cohen’s d = 0.42), supported by qualitative feedback on confidence gains. However, limitations included Synthesia’s inability to model natural conversational turn-taking, reflecting broader AI pragmatism constraints (Hockly, 2024).

Instructors reported 35% time savings in material creation but required post-editing to meet local English for Academic Purposes standards. Adjustable speech pacing and visual annotations strongly correlated with listening comprehension gains (r = 0.67). While Synthesia improved accessibility (e.g., academic hedging tutorials), 22% of students noted reduced motivation with excessive avatar use, highlighting risks of over-reliance (Huang et al., 2021).

The study positions Synthesia as a supplemental CALL tool within blended learning frameworks, recommending strategic integration for preparatory/reinforcement tasks. Key suggestions include pairing AI materials with live communication practice and establishing ethical guidelines for AI in language education.