ANALYSIS OF AI TOOLS FOR ORAL SPEECH ASSESSMENT

ABSTRACT

The article examines the application of artificial intelligence (AI) technologies to oral speech assessment in English language teaching. Three functions of AI tools are considered: automatic speech recognition, automated scoring, and AI-supported feedback. The analysis draws on a research report on the SpeechRater engine published by Educational Testing Service (ETS), official materials from ETS and the British Council, and a methodological handbook on the use of AI in language teaching. The article describes how automated systems process spoken responses, which linguistic features they analyze, and what information they provide to learners and teachers. SpeechRater and AiBC are compared according to their purpose, the aspects of speech they evaluate, the type of assessment they support, and the role of the teacher. Particular attention is paid to the limitations of these technologies, including difficulties in recognizing some accents and varieties of English and the need to protect learner data. It is concluded that AI-based speaking assessment is most useful in formative assessment when it is combined with the teacher’s professional judgement.

Keywords: artificial intelligence, oral speech assessment, formative assessment, automated scoring, automatic speech recognition, AI-supported feedback, English language teaching.

Introduction

Artificial intelligence has created new opportunities for foreign language teaching and assessment. Speaking is a key component of communicative competence, requiring learners to express ideas clearly, use appropriate vocabulary, and construct grammatically accurate sentences. Its assessment involves several criteria, including pronunciation, fluency, accuracy, and the organization of ideas, making it a time-consuming task for teachers.

Assessment can be divided into formative and summative types. Formative assessment provides feedback to support learning and improvement, whereas summative assessment evaluates learning outcomes, usually through a grade or score [4, p. 58]. AI tools can support both types, but this article focuses on their use in formative assessment.

Educational Testing Service (ETS) has developed SpeechRater, an automated system for assessing spoken English [1; 2]. The British Council’s AiBC, in turn, supports speaking practice through AI-generated personalized feedback [3]. These tools represent different approaches: SpeechRater focuses on automated scoring, while AiBC emphasizes language practice and learner development.

This article aims to analyze AI tools for oral speech assessment by examining automatic speech recognition, automated scoring, and AI-supported feedback. The tools are compared in terms of their purpose, the aspects of speech they address, the types of assessment they support, and the role of the teacher.

Automated Scoring

Automated scoring uses computational models to analyze spoken responses and generate scores based on selected speech features. According to Chen et al., SpeechRater evaluates non-native spontaneous English speech through features related to pronunciation, prosody, vocabulary, grammar, content, and discourse [1]. These features provide information about speech production, fluency-related characteristics, language use, and the organization of spoken responses.

ETS identifies pronunciation, fluency, vocabulary, and grammar among the areas addressed by SpeechRater [2]. In formative assessment, automated scoring can help learners identify aspects of their speaking that require improvement. For example, students can record a short response about environmental protection, review the available assessment results, and focus on improving their vocabulary, grammatical accuracy, or fluency.

Automated speech assessment is also relevant to adaptive testing, including placement tests. In these tests, the difficulty of subsequent tasks depends on the learner’s previous responses. Advances in automated speech assessment make it possible to assess speaking alongside written language, providing a broader picture of language proficiency [4, p. 59].

AI-Supported Feedback

AI-supported feedback is particularly relevant to formative assessment because it helps learners identify areas for improvement and plan subsequent learning. The British Council’s AiBC supports English-speaking practice through simulated conversations and provides personalized feedback on grammar, vocabulary, fluency, and clarity [3].

In senior high school, this approach can be integrated into a cycle of speaking practice, feedback, reflection, and task repetition. For example, students can discuss their future career plans, review AI-generated recommendations, identify areas for improvement, and repeat the task. Teachers can then discuss their progress and help them apply the feedback to subsequent activities.

Automated assessment can also generate data about individual learners and groups. Analyzing these data can help identify common learning difficulties and inform the planning of targeted teaching activities [4, pp. 60–61]. The British Council presents AiBC as a tool that complements teacher-led learning rather than replacing the teacher, providing additional speaking practice while teachers continue to guide learners’ development [3].

Comparison of SpeechRater and AiBC

Table 1.

Key characteristics of the two tools, in accordance with the criteria defined in the introduction.

The comparison shows that the three functions considered in this article perform related but distinct roles. Speech recognition converts spoken language into text; automated scoring estimates the quality of a spoken response according to selected criteria; AI-supported feedback gives learners information that helps them review their performance and plan further practice. SpeechRater covers the widest range of speech features and is linked to high-stakes testing, while AiBC is oriented towards regular practice and repeated attempts. In both cases, the results become educationally meaningful only when the teacher connects them to learning objectives.

Conclusion

Artificial intelligence technologies offer several approaches to supporting oral speech assessment in English language teaching. Research by ETS shows how automated systems analyze spoken responses through speech recognition, linguistic feature extraction, and scoring models, while the British Council’s AiBC demonstrates how AI supports speaking practice and personalized feedback. These tools serve different purposes: SpeechRater measures performance, whereas AiBC supports practice and improvement. Both can be used for formative assessment, encouraging learners to review their performance, practise speaking, and work towards learning objectives.

However, these tools have limitations, including difficulties with some accents and varieties of English, limited assessment of spoken interaction, and concerns about learner data protection. Their educational value depends on how teachers integrate them into the teaching process and interpret the results. Further research could examine their accuracy for learners in Kazakhstan and their impact on speaking skill development.

References:

1. Chen, L., Zechner, K., Yoon, S.-Y., Evanini, K., Wang, X., Loukina, A., Tao, J., Davis, L., Lee, C. M., Ma, M., Mundkowsky, R., Lu, C., Leong, C. W., & Gyawali, B. (2018). Automated Scoring of Nonnative Speech Using the SpeechRater v. 5.0 Engine. ETS Research Report Series, RR-18-10. Educational Testing Service. https://doi.org/10.1002/ets2.12198
2. Educational Testing Service. The SpeechRater Service: Advanced Spoken-Response Scoring Application. Available at: https://www.ets.org/speechrater.html
3. British Council. (2025, October 30). British Council launches AiBC: A human-first AI engine transforming English language learning. Available at: https://www.britishcouncil.org/about/press/british-council-launches-aibc-human-first-ai-engine-transforming-english-language
4. Hockly, N. (2024). Nicky Hockly’s 30 Essentials for Using Artificial Intelligence. Cambridge: Cambridge University Press.