A Survey of Automated Presentation Coaching: Systems, Methods, and Open Challenges

Synthesis: This survey provides the first systematic review of automated presentation coaching systems, organizing them along a five-dimensional task taxonomy: segmental pronunciation, lexical stress, suprasegmental prosody, pacing, and content faithfulness. The authors review systems spanning pronunciation tutors, fluency and prosody coaches, Multimodal AI trainers, and conference Q&A practice tools, identifying coverage gaps in pacing and content feedback dimensions.

The survey connects to intelligent tutoring systems by examining how real-time feedback delivery mechanisms in coaching systems mirror educational feedback principles. The taxonomy serves as a framework for professional skills training, particularly for second language learners, and highlights the role of scaffolded feedback in skill development.

Key open challenges include the scarcity of annotated presentation corpora, achieving accent-fair feedback across diverse L1 backgrounds, and delivering real-time multimodal feedback without overwhelming the learner — all of which connect to broader personalized learning and AI literacy concerns in education technology design.

What this means for practice

• Instructors. Check which of the five taxonomy dimensions a coaching tool actually scores before assigning it: of the 15 systems surveyed, only two deliver feedback in real time, and coverage clusters on segmental pronunciation while lexical stress remains the most underaddressed dimension even though misstressed words are a leading cause of reduced L2 comprehensibility.
• Instructors. Assign dimension-specific Feedback rather than a single holistic rating — for example "lower your pitch at the start of a new section" for discourse prosody, or "say algorithm with stress on the first syllable" for lexical stress — because current systems assess intonation only at the utterance level while presentations require section-level prosodic planning.
• Instructors. Personalize pacing and prosodic targets per speaker instead of enforcing fixed norms; the survey proposes adapting targets from a 2–3 minute calibration recording, so scaffolded coaching improves clarity without erasing the learner's accent identity.
• Designers. Build accent fairness into thresholds: calibrate them per L1 cohort, report subgroup precision and recall by L1, and avoid penalizing identity-marking features such as rhoticity patterns or vowel quality when intelligibility is unaffected.
• Researchers. Contribute to shared evaluation infrastructure: the field lacks standardized slide-aligned corpora, and the survey recommends mini-sets of 20–30 speakers from 5 or more L1 backgrounds each delivering a 5–10 minute technical talk with timestamped slide boundaries, phone-level transcriptions, and expert prosody ratings — plus validation that metrics predict actual learner improvement rather than reference similarity.

Limitations

• As a survey, coverage is deliberately not exhaustive: the authors emphasize work that informs pronunciation, prosody, pacing, and comprehensibility for slide-based L2 English presentations, and map 15 representative systems onto the taxonomy.
• English-only scope and a single genre: applicability to other languages and to presentation styles such as storytelling or persuasive speaking requires further investigation.
• The taxonomy and metrics were derived from existing literature and may not capture all dimensions relevant to real-world coaching; the datasets curated and recommended are limited in scale, since most CAPT corpora cover isolated words or short sentences without slide structure, domain terminology, or discourse-level features.
• No empirical comparison is provided between clone-and-compare references and expert references across speaker populations, the practical implementation of accent-fair thresholds and privacy-preserving systems still needs validation in deployed settings, and cross-system comparison is constrained because commercial platforms (ELSA Speak, Speechling, Yoodli, Orai) typically address only one or two taxonomy dimensions and rarely publish their technical approaches.

Citation

Wen Liang, Li Siyan, Zackary Rackauckas, Julia Hirschberg (2026). A Survey of Automated Presentation Coaching: Systems, Methods, and Open Challenges. Computation and Language (cs.CL).