Pith. sign in

REVIEW 3 major objections 6 minor 66 references

Multimodal Assessment of Classroom Discourse Quality: A Text-Centered Attention-Based Multi-Task Learning Approach

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read The paper claims that a text-centered multimodal multi-task model can assess classroom discourse quality from 16-minute lesson segments at a level of agreement comparable to trained human raters, with the text-audio configuration reaching…

desk verdict Solid and useful multimodal classroom-assessment paper, but its main 'surpasses human IRR' claim compares apples to oranges and should be re-benchmarked. read the letter →

arxiv 2505.07902 v1 pith:373WD4MS submitted 2025-05-12 cs.CY cs.AIcs.CLcs.LG

classification cs.CYcs.AIcs.CLcs.LG
keywords ClassroomobservationDiscoursepracticesMultimodalfusionAttentionmechanismsMulti-tasklearningOrdinalclassificationQuadraticWeightedKappaTeachingqualityassessment
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that a supervised multimodal model can score holistic classroom discourse quality from 16-minute lesson segments with consistency comparable to trained human raters. The authors build a text-centered attention architecture that fuses transcript, audio, and video embeddings and jointly predicts three GTI discourse components: Nature of Discourse, Questioning, and Explanations. Using 367 segments from 92 videotaped German math lessons, the text-audio model achieves an average Quadratic Weighted Kappa of 0.384 against double-rated ground truth, while the average human inter-rater reliability is 0.326. If the result is correct, automated scoring of discourse practices becomes feasible at scale, providing teachers with timely feedback and easing a costly manual coding process.

What carries the argument

The load-bearing mechanism is a text-centered attention stack: utterance-level embeddings from a long-context German text encoder, chunk-level embeddings from a German fine-tuned wav2vec-style speech model, and frame-level embeddings from a contrastive vision-language model are passed through two cross-attention encoders in which text serves as the query and audio or video as key-value, followed by a self-attention encoder; this module is repeated, and a CLS token becomes the global representation. The representation feeds three softmax heads, one per discourse component, trained jointly with a weighted ordinal log-loss that penalizes a prediction by its absolute distance from the true rating category, preserving the order of the 1-to-4 scale.

What would settle it

Recompute the text-plus-audio model's QWK against each individual rater's score instead of the averaged double-rater label on the same 367 segments; if the per-rater average drops to or below the 0.326 human benchmark, the claim of matching or exceeding human reliability does not survive. A related check is to compute human QWK on the same averaged labels, which should be higher than pairwise human IRR if averaging is the source of the gap.

Watch

Extended reading notes

Core claim

The central claim is that a text-centered multimodal multi-task learner can assess extended classroom discourse at a consistency level comparable to trained human observers. With transcripts alone the model reaches an average QWK of 0.356, above the human benchmark; adding acoustic features raises the average to 0.384, with the largest single-component gain in Nature of Discourse, and adding video lowers performance to 0.364. Ablations show that the attention encoder outperforms an LSTM baseline, multi-task learning trades a minor accuracy loss for a single shared model, and ordinal log-loss outperforms both L1 regression and standard cross-entropy. The model also reproduces parts of the human-rated correlation pattern with student outcomes, notably the negative association between discourse quality and student interest reported in the original GTI study.

Load-bearing premise

The comparison between model and human performance is only fair if a model's agreement with the average of two raters can be set against one rater's agreement with another; averaging smooths out rater noise, so the benchmark is tilted toward the model.

Editorial extensions

If this is right

  • Classroom observation protocols like the GTI Discourse domain can be automated at scale, since one multi-task model scores all three components in a single pass.
  • Video is not needed for good discourse-quality prediction: dropping the visual stream improves the text-audio model, so deployment could rely on audio recordings that preserve more privacy.
  • The high-inference components Questioning and Explanations are best predicted from text, suggesting transcripts carry the core evidence for rubric-based discourse scoring.
  • Ordinal classification with ordinal log-loss should be the default formulation for teaching-quality rating scales, since it outperformed both regression and standard cross-entropy in the ablations.
  • Model-predicted scores reproduce part of the human-rated correlation pattern with student outcomes, indicating automated scores can stand in for human ratings in large-scale studies.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The headline comparison likely overstates the model's true reliability: human inter-rater reliability is measured pairwise, while the model is scored against the average of two raters, so a fairer test is per-rater QWK; if that drops below 0.326, the claimed edge over human raters evaporates.
  • A fully automated deployment would need automatic speech recognition to produce transcripts, whereas this study uses human transcripts, so end-to-end performance under ASR errors remains an open empirical question.
  • The visual stream's poor showing may be an artifact of generic contrastive visual embeddings rather than evidence that classroom images carry no signal; scene-aware or gesture-specific visual encoders are a natural next test.
  • The counter-intuitive negative correlations between discourse quality and student interest reproduce the original GTI report, but the mechanism is left open; a human-in-the-loop interpretation study would be a concrete extension.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes a supervised multimodal architecture for scoring three classroom-discourse components from the GTI observation protocol: Nature of Discourse, Questioning, and Explanations. It uses frozen pre-trained encoders for transcript, audio, and video; text-guided cross-attention and self-attention fusion modules; multi-task learning; and ordinal log-loss. Experiments on 367 segments from 92 German mathematics lessons use teacher-independent nested 5-fold cross-validation and report an average QWK of 0.384 for the text-plus-audio configuration, which the authors describe as comparable to and sometimes surpassing human inter-rater reliability (0.326). Ablations examine modality contributions, attention versus LSTM encoders, single- versus multi-task learning, and ordinal versus standard losses. The paper also reports correlations between model-predicted and human-rated discourse quality and student outcomes.

Significance. If the headline comparison were valid, the contribution would be significant: it moves automated classroom observation beyond utterance-level classification toward holistic segment-level discourse scoring, and the text-audio result would suggest that reliable automated assessment is possible using privacy-preserving transcript/audio inputs. The paper's strengths are its realistic setting, the teacher-independent evaluation protocol, the systematic ablation study, and the inclusion of student-outcome validation. However, the central quantitative claim rests on a benchmark mismatch: the model is evaluated against averaged double-rater labels while human IRR is measured between individual raters, and the two QWKs are computed on different numbers of rating categories. This does not invalidate the architecture or the ablation findings, but the headline conclusion about surpassing human consistency is not currently established and requires a corrected comparison.

major comments (3)
  1. [Sects. 3.1.3, 4.2, 5.1.3; Eq. (3)] The headline comparison to human IRR is not on equal terms. Human IRR is computed in Sect. 4.2 as a leave-one-rater-out QWK between individual raters' integer 1-4 ratings, whereas the model QWK in Sect. 5.1.3 is computed against the averaged double-rater labels on the seven-point set {1, 1.5, ..., 4} introduced in Sect. 3.1.3. Averaging two independent raters removes rater-specific noise, and increasing the number of categories from K=4 to K=7 changes the chance correction in Eq. (3); both effects work in the model's favor. Consequently, the reported 0.384 versus 0.326 does not establish that the model is 'comparable to' or 'surpasses' human IRR as claimed in the Abstract and Sect. 6.1. The authors should recompute model QWK against each individual rater's original integer scores; if the averaged target is retained, they should compare against the projected reliability of the two-rater average rather than the single-rater IRR.
  2. [Sects. 5.1.3, 6.1; Tables 3 and 4] The claims of improvement from adding audio and from using the ordinal loss are not supported by any inferential statistic. Tables 3 and 4 report point estimates with standard errors but no confidence intervals or paired tests; for example, the overall T+A advantage over T (0.384 vs 0.356) and the OLL advantage over CE (0.384 vs 0.332) are described as notable without an assessment of uncertainty. Because all configurations are evaluated on the same teacher-independent folds, a paired bootstrap over folds or segments is straightforward and should be reported before the 'notable margin' wording in Sect. 6.1 is used.
  3. [Sect. 4.2; Eq. (1) and Eq. (3)] It is not specified how the model's probabilistic output over the seven rating categories is converted into the confusion matrix used in Eq. (3). The loss in Eq. (1) produces a soft probability distribution, but QWK is defined for hard counts. The authors should state whether they take the argmax category or use an expected-count/soft QWK and justify the choice. This decision affects comparability with the human IRR, which is necessarily based on hard integer ratings.
minor comments (6)
  1. [Sects. 3.3.1, 5.1.2, 6.1] The paper repeatedly writes 'textural' where 'textual' is meant; please correct the typo throughout.
  2. [Sect. 3.3.4] The sentence 'we sorely utilized its visual encoder' should read 'we solely utilized its visual encoder'.
  3. [Sect. 2.1] The phrase 'providing students with key phrases they can during learning' appears to be missing a verb and should be rewritten.
  4. [Sect. 3.3.4 and Table 2] The optimal number of fusion modules M from the grid search is never reported; this architectural hyperparameter should be stated for reproducibility.
  5. [Sect. 4.1] The fixed random seed is not given; for reproducibility the seed value should be reported.
  6. [Table 5] The large number of correlations is reported without correction for multiple testing or adjustment for students nested in classrooms; the interpretation of 'partial alignment' should be correspondingly cautious.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a supervised empirical evaluation against held-out human ratings, not a derivation whose conclusions are equivalent to its inputs.

full rationale

This is an empirical supervised-learning study, not a derivation chain. The model is trained on human-rated GTI segments and evaluated on teacher-independent held-out folds (Sect. 4.1), with QWK computed between held-out predictions and the averaged double-rater ground truth (Sect. 4.2, Table 3). Human inter-rater reliability is measured separately by leave-one-rater-out QWK. No equation in the method section reduces to a fitted parameter that is then renamed as a prediction; the ordinal log-loss, multi-task loss, and QWK metric are all standard evaluation components with no circular dependency. The claimed comparison to human IRR is a benchmark, not an input to training, and the correlation analysis with student outcomes is an independent external validation. Although the human-IRR comparison mixes a single-rater agreement target with an averaged-label model target, that is a measurement-validity concern about comparability, not a circularity in the sense of the paper's claims being true by construction. The authors' self-citations (e.g., prior work on classroom emotion and hand-raising detection) are contextual and not load-bearing for the central result. Therefore the appropriate circularity score is 0.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim rests on several domain assumptions about the validity of the observation protocol, the reliability of averaged human ratings, the suitability of frozen pretrained feature extractors, and the representativeness of the accessible subset. No new physical or conceptual entities are introduced.

free parameters (5)
  • Number of fusion modules M = selected from grid 1-5
    Tuned via inner cross-validation; controls model depth and capacity.
  • Audio chunk duration = 10 seconds
    Chosen experimentally (Sect. 3.3.2); affects the number of audio representations and temporal granularity.
  • Task weights mu_c = 1 for all tasks
    Hand-set in Eq. (2); treats all discourse components equally.
  • Segment combination threshold = 8 minutes
    Short last segments under 8 minutes are merged with previous segment (Sect. 3.1.3); affects segment count and label distribution.
  • Class weights w_c = inverse class frequency
    Computed from training set to handle imbalance; influences the loss function.
assumptions (6)
  • domain assumption GTI observation protocol ratings are valid measures of discourse quality
    The three components are adopted from the GTI protocol (Sect. 3.1.2) and treated as ground truth without independent validation.
  • domain assumption Averaging double-rated scores yields a reliable training target
    The mean of two raters is used as ground truth (Sect. 3.1.3); this assumes rater noise is symmetric and averaging improves accuracy.
  • domain assumption Pretrained frozen encoders produce suitable representations for classroom data
    JinaBERT, XLSR-German, and CLIP are used without fine-tuning (Sect. 3.3); their suitability for this specific domain is assumed.
  • domain assumption The ordered rating scale is linearly spaced for the ordinal loss
    Labels 1, 1.5, ..., 4 are treated as equally spaced in Eq. (1), so L1 distance is a meaningful penalty.
  • domain assumption Teacher-independent folds generalize to unseen teachers
    Nested 5-fold CV groups segments by teacher (Sect. 4.1); this assumes the 50 teachers represent the target population.
  • domain assumption Accessible 92 recordings are representative of the full GTI Germany sample
    Eight recordings were inaccessible due to data protection (Sect. 3.1.1); no analysis of selection bias is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multimodal Assessment of Classroom Discourse Quality: A Text-Centered Attention-Based Multi-Task Learning Approach." pith.science (2026). https://pith.science/paper/373WD4MS

@misc{pith2026250507902,
  author       = {Pith},
  title        = {Pith review of: Multimodal Assessment of Classroom Discourse Quality: A Text-Centered Attention-Based Multi-Task Learning Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/373WD4MS}},
  note         = {Machine review of arXiv:2505.07902}
}
read the original abstract

Classroom discourse is an essential vehicle through which teaching and learning take place. Assessing different characteristics of discursive practices and linking them to student learning achievement enhances the understanding of teaching quality. Traditional assessments rely on manual coding of classroom observation protocols, which is time-consuming and costly. Despite many studies utilizing AI techniques to analyze classroom discourse at the utterance level, investigations into the evaluation of discursive practices throughout an entire lesson segment remain limited. To address this gap, our study proposes a novel text-centered multimodal fusion architecture to assess the quality of three discourse components grounded in the Global Teaching InSights (GTI) observation protocol: Nature of Discourse, Questioning, and Explanations. First, we employ attention mechanisms to capture inter- and intra-modal interactions from transcript, audio, and video streams. Second, a multi-task learning approach is adopted to jointly predict the quality scores of the three components. Third, we formulate the task as an ordinal classification problem to account for rating level order. The effectiveness of these designed elements is demonstrated through an ablation study on the GTI Germany dataset containing 92 videotaped math lessons. Our results highlight the dominant role of text modality in approaching this task. Integrating acoustic features enhances the model's consistency with human ratings, achieving an overall Quadratic Weighted Kappa score of 0.384, comparable to human inter-rater reliability (0.326). Our study lays the groundwork for the future development of automated discourse quality assessment to support teacher professional development through timely feedback on multidimensional discourse practices.

Figures

Figures reproduced from arXiv: 2505.07902 by the authors.

Figure 1
Figure 1. Distributions of average double-coded human ratings for three discourse components. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Text-centered multimodal attention-based architecture for classroom discourse assessment through multi-task learning. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 53 canonical work pages

  1. [1]

    INTRODUCTION Classroom discourse serves as the medium through which teaching and learning take place [10]. Previous research has consistently shown that the quality of classroom discourse is positively associated with student learning [31] and that teachers’ skills in facilitating productive discourse influence students’ motivation, autonomy, and interest...

  2. [2]

    explaining

    RELATED WORK In this section, we review related research in two areas: the automated classroom utterance analysis and the holistic as- sessment of teaching effectiveness using classroom observa- tion protocols. 2.1 Automated Classroom Discourse Analysis To characterize discursive strategies that can promote effec- tive teaching, educational researchers ha...

  3. [3]

    L” for the teacher and “S01

    METHOD In this section, we present our methodology for automated classroom discourse assessment. We begin by describing the Global Teaching InSights (GTI) dataset used in this study, highlighting the three discourse components within its ob- servation protocol. Next, we provide a brief overview of our task formulation and present the key elements of our p...

  4. [4]

    Teacher-directed discourse is characterized by a monological lecture style or by the teacher controlling the communication flow through questions and answers

    Nature of Discourse distinguishes teacher-directed and more student-centered interactions. Teacher-directed discourse is characterized by a monological lecture style or by the teacher controlling the communication flow through questions and answers. In contrast, student- centered discourse involves students actively contribut- ing detailed insights on the...

  5. [5]

    The scale spans from basic recall or yes/no questions to more complex inquiries that require stu- dents to explain, classify, analyze, or synthesize infor- mation

    Questioning evaluates the types of questions the teacher poses, focusing on the level of cognitive demand placed on students. The scale spans from basic recall or yes/no questions to more complex inquiries that require stu- dents to explain, classify, analyze, or synthesize infor- mation

  6. [6]

    It assesses the depth of these explanations, from super- ficial or brief descriptions to more detailed and thor- ough explorations of the subject matter

    The Explanations component examines the extent to which teachers and students provide justifications or reasoning behind mathematical ideas and procedures. It assesses the depth of these explanations, from super- ficial or brief descriptions to more detailed and thor- ough explorations of the subject matter. Table 1 presents a summary of the key attribute...

  7. [7]

    This approach grouped all the segments from the same teacher into a single fold, en- suring the model’s generalizability to unseen teachers

    EXPERIMENTS 4.1 Implementation We implemented a teacher-independent nested 5-fold cross- validation for training and evaluation. This approach grouped all the segments from the same teacher into a single fold, en- suring the model’s generalizability to unseen teachers. Be- sides, the cross-validation process allowed us to collect infer- ences for the enti...

  8. [8]

    Unless otherwise specified, all models were trained and evaluated under the same condi- tions described in Sect

    RESULTS In this section, we present the results of automated class- room discourse assessment. Unless otherwise specified, all models were trained and evaluated under the same condi- tions described in Sect. 4.1. 5.1 Model Performance across Modalities Table 3 summarizes the results of our multimodal multi- task learning framework in comparison to human I...

Show all 66 references
  1. [9]

    In terms of overall performance, even the exclusive use of textural rep- resentations achieved a slightly higher agreement with the manual ratings than human IRR

    DISCUSSION 6.1 Main Findings Leveraging human IRR as a benchmark, we demonstrated the efficacy of our proposed architecture in assessing dis- cursive practices across various dimensions. In terms of overall performance, even the exclusive use of textural rep- resentations achi...

  2. [10]

    CONCLUSION The present study introduces an attention-based multimodal fusion architecture that utilizes multi-task learning to si- multaneously assess three classroom discourse dimensions (i.e., Nature of Discourse , Questioning, and Explanations), yielding a level of agreemen...

  3. [11]

    R. J. Alexander. Towards dialogic teaching: Rethinking classroom talk. Dialogos, 2008

  4. [12]

    S. Alic, D. Demszky, Z. Mancenido, J. Liu, H. Hill, and D. Jurafsky. Computationally identifying funneling and focusing questions in classroom discourse. arXiv preprint arXiv:2208.04715 , 2022

  5. [13]

    Ateia and U

    S. Ateia and U. Kruschwitz. Can open-source llms compete with commercial models? exploring the few-shot performance of current gpt models in biomedical tasks. arXiv preprint arXiv:2407.13511 , 2024

  6. [14]

    Baevski, Y

    A. Baevski, Y. Zhou, A. Mohamed, and M. Auli. wav2vec 2.0: A framework for self-supervised learning of speech representations. Advances in neural information processing systems, 33:12449–12460, 2020

  7. [15]

    C. Bell, Y. Qi, M. Witherspoon, M. Barragan, and H. Howell. Annex a: Talis video observation codes: Holistic domain ratings and components. In Global Teaching Insights: Technical Report. OECD, 2018

  8. [16]

    C. Bell, Y. Qi, M. Witherspoon, M. Barragan, and H. Howell. Annex a: Talis video training notes: Holistic domain ratings and components. In Global Teaching Insights: Technical Report. OECD, 2018

  9. [17]

    B ¨uhler, E

    B. B ¨uhler, E. Bozkir, P. Goldberg, ¨O. S¨umer, S. D’Mello, P. Gerjets, U. Trautwein, and E. Kasneci. From the lab to the wild: Examining generalizability of video-based mind wandering detection. International Journal of Artificial Intelligence in Education, pages 1–35, 2024

  10. [18]

    B ¨uhler, R

    B. B ¨uhler, R. Hou, E. Bozkir, P. Goldberg, P. Gerjets, U. Trautwein, and E. Kasneci. Automated hand-raising detection in classroom videos: A view-invariant and occlusion-robust machine learning approach. In International Conference on Artificial Intelligence in Education , p...

  11. [19]

    Castagnos, M

    F. Castagnos, M. Mihelich, and C. Dognin. A simple log-based loss function for ordinal text classification. In Proceedings of the 29th International Conference on Computational Linguistics, pages 4604–4609, 2022

  12. [20]

    C. B. Cazden. Classroom discourse: The language of teaching and learning. ERIC, 1988

  13. [21]

    C. Chin. Teacher questioning in science classrooms: Approaches that stimulate productive thinking. Journal of Research in Science Teaching: The Official Journal of the National Association for Research in Science Teaching, 44(6):815–843, 2007

  14. [22]

    Conneau, A

    A. Conneau, A. Baevski, R. Collobert, A. Mohamed, and M. Auli. Unsupervised cross-lingual representation learning for speech recognition. arXiv preprint arXiv:2006.13979, 2020

  15. [23]

    Demszky, J

    D. Demszky, J. Liu, H. C. Hill, D. Jurafsky, and C. Piech. Can automated feedback improve teachers’ uptake of student ideas? evidence from a randomized controlled trial in a large-scale online course. Educational Evaluation and Policy Analysis , 2023

  16. [24]

    Demszky, J

    D. Demszky, J. Liu, Z. Mancenido, J. Cohen, H. Hill, D. Jurafsky, and T. Hashimoto. Measuring conversational uptake: A case study on student-teacher interactions. arXiv preprint arXiv:2106.03873, 2021

  17. [25]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding, 2019

  18. [26]

    Dubey, A

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783 , 2024

  19. [27]

    Frank and M

    E. Frank and M. Hall. A simple approach to ordinal classification. In Machine Learning: ECML 2001: 12th European Conference on Machine Learning Freiburg, Germany, September 5–7, 2001 Proceedings 12 , pages 145–156. Springer, 2001

  20. [28]

    M. L. Franke, E. Kazemi, and D. Battey. Mathematics teaching and classroom practice. Second handbook of research on mathematics teaching and learning , 1(1):225–256, 2007

  21. [29]

    Goldberg, ¨O

    P. Goldberg, ¨O. S¨umer, K. St¨urmer, W. Wagner, R. G¨ollner, P. Gerjets, E. Kasneci, and U. Trautwein. Attentive or not? toward a machine learning approach to assessing students’ visible engagement in classroom instruction. Educational Psychology Review, 33:27–49, 2021

  22. [30]

    Grossman, S

    P. Grossman, S. Loeb, J. Cohen, and J. Wyckoff. Measure for measure: The relationship between measures of instructional practice in middle school english language arts and teachers’ value-added scores. American Journal of Education , 119(3):445–470, 2013

  23. [31]

    Gr ¨unkorn, E

    J. Gr ¨unkorn, E. Klieme, A.-K. Praetorius, and P. Schreyer. Mathematikunterricht im internationalen vergleich. ergebnisse aus der talis-videostudie deutschland. DIPF| Leibniz-Institut f¨ur Bildungsforschung und Bildungsinformation , 2020

  24. [32]

    H. C. Hill, B. Rowan, and D. L. Ball. Effects of teachers’ mathematical knowledge for teaching on student achievement. American educational research journal, 42(2):371–406, 2005

  25. [33]

    R. Hou, T. F ¨utterer, B. B¨uhler, E. Bozkir, P. Gerjets, U. Trautwein, and E. Kasneci. Automated assessment of encouragement and warmth in classrooms leveraging multimodal emotional features and chatgpt. In International Conference on Artificial Intelligence in Education, pag...

  26. [34]

    beautiful work, you’re rock stars!

    N. Hunkins, S. Kelly, and S. D’Mello. “beautiful work, you’re rock stars!”: Teacher analytics to uncover discourse that supports or undermines student motivation, identity, and belonging in classrooms. In LAK22: 12th International Learning Analytics and Knowledge Conference, p...

  27. [35]

    James, M

    A. James, M. Kashyap, Y. H. V. Chua, T. Maszczyk, A. M. N´ u˜ nez, R. Bull, and J. Dauwels. Inferring the climate in classrooms from audio and video recordings: a machine learning approach. In IEEE International Conference on Teaching, Assessment, and Learning for Engineering,...

  28. [36]

    T. J. Kane, D. F. McCaffrey, T. Miller, and D. O. Staiger. Have we identified effective teachers? validating measures of effective teaching using random assignment. Research Paper. MET Project. Bill & Melinda Gates Foundation, 2013

  29. [37]

    Kiemer, A

    K. Kiemer, A. Gr ¨oschner, A.-K. Pehmer, and T. Seidel. Effects of a classroom discourse intervention on teachers’ practice and students’ motivation to learn mathematics and science. Learning and instruction, 35:94–103, 2015

  30. [38]

    Kupor, C

    A. Kupor, C. Morgan, and D. Demszky. Measuring five accountable talk moves to improve instruction at scale. arXiv preprint arXiv:2311.10749 , 2023

  31. [39]

    H. Liu, C. Li, Y. Li, B. Li, Y. Zhang, S. Shen, and Y. J. Lee. Llava-next: Improved reasoning, ocr, and world knowledge, 2024

  32. [40]

    Loshchilov and F

    I. Loshchilov and F. Hutter. Decoupled weight decay regularization, 2019

  33. [41]

    Mercer and L

    N. Mercer and L. Dawes. The study of talk between teachers and students, from the 1970s until the 2010s. Oxford review of education, 40(4):430–445, 2014

  34. [42]

    I. Mohr, M. Krimmel, S. Sturua, M. K. Akram, A. Koukounas, M. G ¨unther, G. Mastrapas, V. Ravishankar, J. F. Mart´ ınez, F. Wang, et al. Multi-task contrastive learning for 8192-token bilingual text embeddings. arXiv preprint arXiv:2402.17016, 2024

  35. [43]

    Muennighoff, N

    N. Muennighoff, N. Tazi, L. Magne, and N. Reimers. Mteb: Massive text embedding benchmark. arXiv preprint arXiv:2210.07316, 2022

  36. [44]

    Nguyen, H

    A. Nguyen, H. N. Ngo, Y. Hong, B. Dang, and B.-P. T. Nguyen. Ethical principles for artificial intelligence in education. Education and Information Technologies, 28(4):4221–4241, 2023

  37. [45]

    Global Teaching InSights: A Video Study of Teaching

    OECD. Global Teaching InSights: A Video Study of Teaching. OECD, Paris, 2020

  38. [46]

    O’Connor, S

    C. O’Connor, S. Michaels, and S. Chapin. Scaling down” to explore the role of talk in learning: From district intervention to controlled classroom study. Socializing intelligence through academic talk and dialogue, pages 111–126, 2015

  39. [47]

    Pepino, P

    L. Pepino, P. Riera, and L. Ferrer. Emotion Recognition from Speech Using wav2vec 2.0 Embeddings. In Proc. Interspeech, pages 3400–3404, 2021

  40. [48]

    R. C. Pianta, K. M. La Paro, and B. K. Hamre. Classroom Assessment Scoring System ™: Manual K-3. Paul H Brookes Publishing, 2008

  41. [49]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pages 8748–8763. PMLR, 2021

  42. [50]

    Ramakrishnan, B

    A. Ramakrishnan, B. Zylich, E. Ottmar, J. LoCasale-Crouch, and J. Whitehill. Toward automated classroom observation: Multimodal machine learning to estimate class positive climate and negative climate. IEEE Transactions on Affective Computing, 2021

  43. [51]

    D. L. Redfield and E. W. Rousseau. A meta-analysis of experimental research on teacher questioning behavior. Review of educational research, 51(2):237–245, 1981

  44. [52]

    Shi and S.-L

    T. Shi and S.-L. Huang. Multiemo: An attention-based correlation-aware multimodal fusion framework for emotion recognition in conversations. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14752–14766, 2023

  45. [53]

    S¨umer, P

    ¨O. S¨umer, P. Goldberg, S. D’Mello, P. Gerjets, U. Trautwein, and E. Kasneci. Multimodal engagement analysis from facial videos in the classroom. IEEE Transactions on Affective Computing, 2021

  46. [54]

    Suresh, J

    A. Suresh, J. Jacobs, C. Harty, M. Perkoff, J. H. Martin, and T. Sumner. The talkmoves dataset: K-12 mathematics lesson transcripts annotated for teacher and student discursive moves. arXiv preprint arXiv:2204.09652, 2022

  47. [55]

    Tamuleviˇ cius, G

    G. Tamuleviˇ cius, G. Korvel, A. B. Yayak, P. Treigys, J. Bernataviˇ cien˙ e, and B. Kostek. A study of cross-linguistic speech emotion recognition based on 2d feature spaces. Electronics, 9(10):1725, 2020

  48. [56]

    N. Tran, B. Pierce, D. Litman, R. Correnti, and L. C. Matsumura. Analyzing large language models for classroom discussion assessment. arXiv preprint arXiv:2406.08680, 2024

  49. [57]

    A. B. M. Tsui. Classroom Discourse: Approaches and Perspectives, pages 2013–2024. Springer US, Boston, MA, 2008

  50. [58]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin. Attention is all you need. In Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS’17, page 6000–6010, Red Hook, NY, USA, 2017. Curr...

  51. [59]

    S. Walsh. Investigating classroom discourse. Routledge, 2006

  52. [60]

    D. Wang, D. Shan, Y. Zheng, and G. Chen. Teacher talk moves in k12 mathematics lessons: Automatic identification, prediction explanation, and characteristic exploration. In International Conference on Artificial Intelligence in Education , pages 651–664. Springer, 2023

  53. [61]

    R. E. Wang and D. Demszky. Is chatgpt a good teacher coach? measuring zero-shot performance for scoring and providing actionable insights on classroom instruction. arXiv preprint arXiv:2306.03090 , 2023

  54. [62]

    N. M. Webb, M. L. Franke, M. Ing, J. Wong, C. H. Fernandez, N. Shin, and A. C. Turrou. Engaging with others’ mathematical ideas: Interrelationships among student participation, teachers’ instructional practices, and learning. International Journal of Educational Research, 63:7...

  55. [63]

    White and K

    M. White and K. Klette. What’s in a score? problematizing interpretations of observation scores. Studies in Educational Evaluation , 77:101238, 2023

  56. [64]

    Whitehill and J

    J. Whitehill and J. LoCasale-Crouch. Automated evaluation of classroom instructional support with llms and bows: Connecting global predictions to specific feedback. arXiv preprint arXiv:2310.01132 , 2023

  57. [65]

    Z. Wu, Z. Gong, J. Koo, and J. Hirschberg. Multimodal multi-loss fusion network for sentiment analysis. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers) , p...

  58. [66]

    Yang and Y

    J. Yang and Y. Zhang. Ncrf++: An open-source neural sequence labeling toolkit. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, 2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.