Pith. sign in

REVIEW 3 major objections 5 minor 42 references

Examining Spanish Counseling with MIDAS: a Motivational Interviewing Dataset in Spanish

T0 review · 3 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper introduces MIDAS, the first Spanish motivational interviewing dataset, and uses it to show that Spanish counselors differ measurably from English counselors in conversational style and that Spanish-trained classifiers…

desk verdict MIDAS is a genuinely useful first resource for Spanish MI research, but the paper's analytical claims are thinner than the dataset contribution. read the letter →

arxiv 2502.08458 v1 pith:6G2RWDNE submitted 2025-02-12 cs.CL

Aylin Gunal , Bowen Yi , John Piette , Rada Mihalcea , Verónica Pérez-Rosas This is my paper · ORCID
classification cs.CL
keywords motivationalinterviewingSpanishcounselingdatasetcounselorbehaviorclassificationreflectionsandquestionscross-lingualNLPmentalhealthMITIannotationsentimentanalysis
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

MIDAS is a new, publicly available dataset of 74 Spanish-language motivational interviewing sessions, transcribed from YouTube videos and annotated by three Spanish-speaking counselors with 884 questions and 415 reflections. The paper claims this is the first Spanish motivational interviewing dataset, filling a gap that has kept counseling NLP mostly English-only. Using MIDAS, the authors find that Spanish counselors show different conversational dynamics and word-choice patterns than English counselors, and that classifiers trained on Spanish data outperform both English-trained and multilingual models. If correct, the dataset gives researchers a reusable resource for studying and automating counselor-behavior coding in Spanish and a concrete starting point for culturally aware mental-health NLP.

What carries the argument

The central object is the MIDAS dataset itself, defined as a collection of 74 Spanish motivational interviewing conversations with expert annotations for counselor questions and reflections. The dataset carries the argument because it enables both the cross-language behavioral comparison and the monolingual classifier training. The surrounding machinery includes the ITEM 2 Spanish version of the MITI coding scheme for annotation, the reflection-to-question ratio used as a proficiency indicator, LIWC word-class analysis for language-use comparisons, a sentiment classifier for emotional tone, and transformer-based models fine-tuned in Spanish, English, and multilingual configurations for behavior classification.

What would settle it

Have trained coders score every MIDAS video with the standard MI proficiency instrument; if the counselors are rated low or the reflection-to-question ratio does not track those ratings, the claimed equivalence with English high-quality sessions and the cultural explanations would need to be revised.

Watch

Extended reading notes

Core claim

The central claim is that Spanish-language motivational interviewing exhibits measurable, language-specific counselor behaviors that English data cannot capture, and that a Spanish-specific dataset is therefore needed to model them. The authors construct MIDAS from 74 public videos of Spanish MI demonstrations and role-plays, transcribe and diarize the conversations, and have three Spanish-speaking counselors annotate every counselor turn as a question or a reflection using the Spanish version of the MITI coding scheme. Comparing MIDAS with an English dataset compiled with the same methodology, they report that Spanish counselors show a more variable counselor-to-client word exchange rate, use more future and inhibition words, and express positive and negative sentiment more often, while English counselors use more perception and hearing words. In classification experiments, a Spanish-trained monolingual model reaches about 0.92 accuracy and F1 for distinguishing questions from reflections and for three-way coding, outperforming an English-trained model and multilingual cross-lingual transfer.

Load-bearing premise

The main load-bearing premise is that the English comparison conversations and the Spanish MIDAS conversations were collected in comparable ways and that the Spanish counselors are skilled at motivational interviewing, even though the dataset never measures their skill.

Editorial extensions

If this is right

  • Spanish-language counselor-behavior coding can be automated with a Spanish-trained transformer, reaching roughly 0.92 accuracy and F1 for distinguishing questions, reflections, and other turns.
  • Cross-lingual transfer from English does not match in-language training in this setting, so language-specific datasets are needed for reliable counseling tools in non-English languages.
  • Spanish MI counselors show a more variable counselor-to-client word exchange rate and greater use of future and inhibition words, while English counselors use more perception and hearing words, consistent with a more directive clinical style in Spanish-speaking settings.
  • Spanish clients use more assent words and fewer discrepancy words than English clients, suggesting different conversational compliance patterns.
  • Reflections remain harder to classify than questions in Spanish, matching results in English and Hebrew, so reflection coding should be a focus for future annotation efforts.

Reading between the lines

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

  • This goes beyond the paper: because the videos are educational role-plays, the observed asymmetry in word-exchange rate may be amplified relative to real sessions; testing on naturally occurring Spanish therapy would show whether the paternalistic pattern persists.
  • This goes beyond the paper: the finding that monolingual Spanish models outperform multilingual transfer suggests that multilingual pretraining alone is not a substitute for in-language counseling data; a direct ablation controlling for training-set size could quantify how much of the gap is data-specific.
  • This goes beyond the paper: the reflection-to-question ratio could be validated as a proficiency proxy by correlating it with full MITI global scores on the same transcripts, which would give future Spanish MI datasets a cheaper quality label.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript introduces MIDAS, a new Spanish-language Motivational Interviewing dataset constructed from 74 public YouTube videos. Three paid Spanish-speaking counselors with MI experience annotated counselor utterances for questions (884) and reflections (415) using the Spanish version of the MITI coding scheme. The authors report cross-linguistic comparisons between MIDAS and an English MI dataset (Pérez-Rosas et al., 2018) in terms of word-exchange rates, LIWC categories, and sentiment, and they train BERT-based classifiers to predict counselor behaviors in monolingual and multilingual settings. The paper claims that this is the first Spanish MI dataset and that language-specific models outperform multilingual models for this task.

Significance. The dataset fills a concrete gap: public, annotated non-English counseling corpora are rare, so a curated Spanish MI corpus with expert annotations is potentially valuable for NLP research and counselor training. The use of an established coding instrument (Spanish ITEM 2/MITI), the public release, and the direct comparison with an existing English corpus are strengths. The classification experiments provide a reproducible benchmark for question/reflection detection, and the explicit limitations about scripted content show appropriate caution. However, the reliability evidence and the sentiment-classifier validation are weaker than the uses made of them; the value of the resource depends on resolving those issues.

major comments (3)
  1. [3.2] The reliability analysis in §3.2 rests on a single ICC computed from five conversations for aggregate counts of reflections and questions. An ICC on per-conversation counts can be high even when annotators select different spans, so it does not establish that the spans released in the dataset are reliably labeled. The paper does not state whether the remaining 69 conversations were double-coded, adjudicated, or coded by a single annotator; if the latter, the corpus inherits unmeasured annotator bias. Because §5 trains and evaluates classifiers against these labels, unmeasured label noise would make the reported F1 scores optimistic and weaken the 'expert annotations' characterization. Please report span-level agreement (e.g., token-level or utterance-level kappa/F1) on a sufficiently large double-coded sample and describe the full annotation workflow.
  2. [4 (Sentiment Trends)] The sentiment validation in §4 is insufficient for the cross-linguistic claim that follows. On 300 utterances, the two human annotators reached Cohen's kappa of 0.45 and raw agreement of 0.64, and the PySentimiento classifier achieved accuracy of 0.64, with 69 of 109 misclassifications being neutral predictions. Characterizing this as 'reasonable accuracy scores' and then using the classifier outputs to compare positive/negative sentiment across languages is not justified. Please report per-class precision and recall for the classifier, and either treat the sentiment comparison as exploratory or rerun the cross-linguistic comparison on human labels.
  3. [4 (cross-lingual comparison)] The cross-lingual comparison assumes that MIDAS and the Pérez-Rosas et al. (2018) English corpus are comparable in sourcing, transcription, and counselor skill. The paper states that the English corpus was compiled with the same methodology and that MIDAS lacks proficiency evaluations, using the R:Q ratio as a substitute. This is an explicit assumption, but it is not validated; the English subset is restricted to 72 high-quality sessions, while MIDAS includes any video that met inclusion criteria. Without additional evidence on counselor proficiency and session structure, the observed Spanish-English differences in word exchange, LIWC categories, and sentiment could reflect differences in dataset construction or skill level rather than language or culture. Please either add a validity check (e.g., compare annotation distributions or have the Spanish annotators rate the English sessions) or frame the results as dataset-specific rather than language-specific.
minor comments (5)
  1. [4 (Conversation Word Exchange)] Figure 1 plots word-exchange-rate curves without confidence intervals or significance tests; the claim that Spanish varies more over the conversation should be supported by a statistical comparison rather than visual inspection.
  2. [Table 4] The F1-Reflection values in the multilingual 3-way condition are very low (0.29 for en-MLBERT and 0.22 for sp-MLBERT), but the discussion only says that reflections are harder to classify; please include an error analysis or at least acknowledge the magnitude of this gap.
  3. [4 (Language Usage)] The LIWC interpretation that Spanish clients use assent words 'suggesting greater compliance' goes beyond what word-class frequencies can establish; please soften this claim or provide additional evidence.
  4. [3.2] The Taguette URL is printed as '/www.taguette.org/' but should be a full URL such as 'https://www.taguette.org/'.
  5. [References] The related work text refers to 'GlobHCD' while the reference entry spells the dataset as 'GLoHBCD'; please harmonize the spelling.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: MIDAS is a manually annotated dataset resource; the cross-lingual comparison uses an external empirical benchmark, and the R:Q proficiency claim is an explicit assumption rather than a fitted prediction.

full rationale

MIDAS is a resource-construction paper, not a derivation from fitted parameters. The central artifact (74 Spanish videos, 884 questions, and 415 reflections) is collected and annotated manually; no equation or fitted constant is used to produce these counts. The cross-lingual analysis compares MIDAS against the authors' earlier English dataset (Pérez-Rosas et al., 2018), but that dataset is a published, externally available empirical resource; comparing against it does not make the Spanish findings true by construction, and the comparison could in principle falsify the paper's claims about language differences. The R:Q proficiency comparison is explicitly introduced as an assumption ("we assume that counselors exhibit desirable behaviors during conversations, designed to show MI skills. We instead use the reflection-to-question ratio (R:Q) as a proficiency indicator"), not as a fitted result, and it is anchored to an external MI metric (Moyers et al., 2016). The classifier experiments train standard BERT/BETO models on held-out splits; there is no input-label leakage equivalent to predicting a fitted quantity. The only self-citation is the English comparison dataset, which is not a uniqueness theorem or an unverified ansatz; it functions as an empirical benchmark. Consequently, no circular step can be quoted, and the paper's central claims stand or fall on data quality and comparability rather than on circular reasoning.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper's analyses rest on several domain assumptions: comparability of the English and Spanish datasets, proficiency of the MIDAS counselors, and accuracy of automatic transcription. No free parameters are fitted, and no new entities are postulated.

assumptions (3)
  • domain assumption The English comparison dataset (Pérez-Rosas et al., 2018) is methodologically comparable to MIDAS.
    Section 4 states the English dataset was compiled with the same methodology as MIDAS. If sourcing or annotation differs, cross-lingual comparisons are confounded.
  • domain assumption Counselors in MIDAS exhibit desirable MI behaviors despite lacking proficiency evaluations.
    Section 4 explicitly assumes this and uses the reflection-to-question ratio as a proficiency indicator. If the videos are poor examples of MI, the linguistic comparisons may reflect skill differences rather than cultural differences.
  • domain assumption Automatic transcription and diarization are accurate enough for behavioral analysis.
    Section 3.1 relies on Amazon Transcribe with manual review; transcription errors could affect word-level analyses and classifier inputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Examining Spanish Counseling with MIDAS: a Motivational Interviewing Dataset in Spanish." pith.science (2026). https://pith.science/paper/6G2RWDNE

@misc{pith2026250208458,
  author       = {Pith},
  title        = {Pith review of: Examining Spanish Counseling with MIDAS: a Motivational Interviewing Dataset in Spanish},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6G2RWDNE}},
  note         = {Machine review of arXiv:2502.08458}
}
read the original abstract

Cultural and language factors significantly influence counseling, but Natural Language Processing research has not yet examined whether the findings of conversational analysis for counseling conducted in English apply to other languages. This paper presents a first step towards this direction. We introduce MIDAS (Motivational Interviewing Dataset in Spanish), a counseling dataset created from public video sources that contains expert annotations for counseling reflections and questions. Using this dataset, we explore language-based differences in counselor behavior in English and Spanish and develop classifiers in monolingual and multilingual settings, demonstrating its applications in counselor behavioral coding tasks.

Figures

Figures reproduced from arXiv: 2502.08458 by the authors.

Figure 1
Figure 1. Mean word exchange rates across Spanish and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Counselor sentiment across languages sentiment using the same categories. The annota￾tion is conducted by two native Spanish speakers, achieving a Cohen kappa of 0.45 and a raw agree￾ment of 0.64, indicating moderate agreement. A third native speaker conducted further attribution on 107 utterances with disagreement. Among the 300 utterances, the classifier correctly classifies 192, yielding an accuracy of 0.64. Nota… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 26 canonical work pages

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Tim Althoff, Kevin Clark, and Jure Leskovec. 2016. https://doi.org/10.1162/tacl_a_00111 Large-scale analysis of counseling conversations: An application of natural language processing to mental health . Transactions of the Association for Computational Linguistics, 4:463--476

    doi:10.1162/tacl_a_00111 2016

  4. [4]

    José Cañete, Gabriel Chaperon, Rodrigo Fuentes, Jou-Hui Ho, Hojin Kang, and Jorge Pérez. 2023. http://arxiv.org/abs/2308.02976 Spanish pre-trained bert model and evaluation data

    arXiv 2023

  5. [5]

    Stevie Chancellor and Munmun De Choudhury. 2020. Methods in predictive techniques for mental health status on social media: a critical review. NPJ digital medicine, 3(1):43

    work page 2020

  6. [6]

    Yu Ying Chiu, Ashish Sharma, Inna Wanyin Lin, and Tim Althoff. 2024. http://arxiv.org/abs/2401.00820 A computational framework for behavioral assessment of llm therapists

    arXiv 2024

  7. [7]

    Angela Coulter and Helen Magee. 2003. The European patient of the future. McGraw-Hill Education (UK)

    work page 2003

  8. [9]

    Jacob Devlin, Ming - Wei Chang, Kenton Lee, and Kristina Toutanova. 2018 b . http://arxiv.org/abs/1810.04805 BERT: pre-training of deep bidirectional transformers for language understanding . CoRR, abs/1810.04805

    arXiv 2018

Show all 42 references
  1. [10]

    Steffen T Eberhardt, Jana Schaffrath, Danilo Moggia, Brian Schwartz, Martin Jaehde, Julian A Rubel, Tobias Baur, Elisabeth Andr \'e , and Wolfgang Lutz. 2024. Decoding emotions: Exploring the validity of sentiment analysis in psychotherapy. Psychotherapy Research, pages 1--16

    2024

  2. [11]

    Rosa Gim \'e nez-Moreno and Alicia Ricart-Vay \'a . 2022. https://api.semanticscholar.org/CorpusID:254671588 The expression of emotions in online medical consultations: a comprehensive spanish-english analysis . Ib \'e rica

    2022

  3. [12]

    Bhanu Prakash Reddy Guda, Aparna Garimella, and Niyati Chhaya. 2021. https://doi.org/10.18653/v1/2021.eacl-main.268 E mpath BERT : A BERT -based framework for demographic-aware empathy prediction . In Proceedings of the 16th Conference of the European Chapter of the Associatio...

    2021 doi

  4. [13]

    Nathan Hodson and Simon Williamson. 2024. https://doi.org/10.2196/52500 Can large language models replace therapists? evaluating performance at simple cognitive behavioral therapy tasks . JMIR AI, 3:e52500

    2024 doi

  5. [14]

    Anqi Li, Yu Lu, Nirui Song, Shuai Zhang, Lizhi Ma, and Zhenzhong Lan. 2024 a . Automatic evaluation for mental health counseling using llms. arXiv preprint arXiv:2402.11958

    2024 arXiv

  6. [15]

    Cheng Li, May Fung, Qingyun Wang, Chi Han, Manling Li, Jindong Wang, and Heng Ji. 2024 b . http://arxiv.org/abs/2410.06845 Mentalarena: Self-play training of language models for diagnosis and treatment of mental health disorders

    2024 arXiv

  7. [16]

    Alfonso Urena Lopez, and Maite Martin

    Pilar L \'o pez \'U beda, Flor Miriam Plaza del Arco, Manuel Carlos D \' az Galiano, L. Alfonso Urena Lopez, and Maite Martin. 2019. https://doi.org/10.26615/978-954-452-056-4_077 Detecting anorexia in S panish tweets . In Proceedings of the International Conference on Recent ...

    2019 doi

  8. [17]

    Kate Loveys, Jonathan Torrez, Alex Fine, Glen Moriarty, and Glen Coppersmith. 2018. https://doi.org/10.18653/v1/W18-0608 Cross-cultural differences in language markers of depression online . In Proceedings of the Fifth Workshop on Computational Linguistics and Clinical Psychol...

    2018 doi

  9. [18]

    Tobias Mayer, Neha Warikoo, Amir Eliassaf, Dana Atzil-Slonim, and Iryna Gurevych. 2024. https://aclanthology.org/2024.eacl-long.88 Predicting client emotions and therapist interventions in psychotherapy dialogues . In Proceedings of the 18th Conference of the European Chapter ...

    2024

  10. [19]

    Selina Meyer and David Elsweiler. 2022. https://aclanthology.org/2022.lrec-1.239 GL o HBCD : A naturalistic G erman dataset for language of health behaviour change on online support forums . In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 2...

    2022

  11. [20]

    Rada Mihalcea and Stephen Pulman. 2009. Linguistic ethnography: Identifying dominant word classes in text. In International Conference on Intelligent Text Processing and Computational Linguistics, pages 594--602. Springer

    2009

  12. [21]

    William R Miller and Stephen Rollnick. 2012. Motivational interviewing: Helping people change. Guilford press

    2012

  13. [22]

    Ishani Mondal, Kalika Bali, Mohit Jain, Monojit Choudhury, Ashish Sharma, Evans Gitau, Jacki O ' Neill, Kagonya Awori, and Sarah Gitau. 2021. https://doi.org/10.18653/v1/2021.law-1.7 A linguistic annotation framework to study interactions in multilingual healthcare conversatio...

    2021 doi

  14. [23]

    Theresa B Moyers, Tim Martin, Jennifer K Manuel, William R Miller, and D Ernst. 2003. The motivational interviewing treatment integrity (miti) code: Version 2.0. Retrieved from Verf \"u bar unter: www. casaa. unm. edu [01.03. 2005]

    2003

  15. [24]

    Theresa B Moyers, Lauren N Rowell, Jennifer K Manuel, Denise Ernst, and Jon M Houck. 2016. The motivational interviewing treatment integrity code (miti 4): rationale, preliminary reliability and validity. Journal of substance abuse treatment, 65:36--42

    2016

  16. [25]

    Jingping Nie, Hanya Shao, Yuang Fan, Qijia Shao, Haoxuan You, Matthias Preindl, and Xiaofan Jiang. 2024. http://arxiv.org/abs/2403.10779 Llm-based conversational ai therapist for daily functioning screening and psychotherapeutic intervention via everyday smart devices

    2024 arXiv

  17. [26]

    Hans Oh and Christina Lee. 2016. Culture and motivational interviewing. Patient education and counseling, 99(11):1914

    2016

  18. [27]

    Ai Ohtani, Takefumi Suzuki, Hiroyoshi Takeuchi, and Hiroyuki Uchida. 2015. https://api.semanticscholar.org/CorpusID:42684341 Language barriers and access to psychiatric care: A systematic review. Psychiatric services, 66 8:798--805

    2015

  19. [28]

    Pennebaker, Cindy K

    James W. Pennebaker, Cindy K. Chung, Molly Ireland, Amy L. Gonzales, and Roger John Booth. 2007. https://api.semanticscholar.org/CorpusID:180769814 The development and psychometric properties of liwc2007

    2007

  20. [29]

    Ver \'o nica P \'e rez-Rosas, Xuetong Sun, Christy Li, Yuchen Wang, Kenneth Resnicow, and Rada Mihalcea. 2018. https://aclanthology.org/L18-1591 Analyzing the quality of counseling conversations: the tell-tale signs of high-quality counseling . In Proceedings of the Eleventh I...

    2018

  21. [30]

    Ver \'o nica P \'e rez-Rosas, Xinyi Wu, Kenneth Resnicow, and Rada Mihalcea. 2019. https://doi.org/10.18653/v1/P19-1088 What makes a good counselor? learning to distinguish between high-quality and low-quality counseling conversations . In Proceedings of the 57th Annual Meetin...

    2019 doi

  22. [31]

    Khan, Timothy Baldwin, and Hisham Cholakkal

    Sara Pieri, Sahal Shaji Mullappilly, Fahad Shahbaz Khan, Rao Muhammad Anwer, Salman H. Khan, Timothy Baldwin, and Hisham Cholakkal. 2024. https://api.semanticscholar.org/CorpusID:267760013 Bimedix: Bilingual medical mixture of experts llm . In Conference on Empirical Methods i...

    2024

  23. [32]

    V \' ctor M Prieto, Sergio Matos, Manuel Alvarez, Fidel Cacheda, and Jos \'e Lu \' s Oliveira. 2014. Twitter: a good place to detect health conditions. PloS one, 9(1):e86191

    2014

  24. [33]

    Priyanshu Priya, Gopendra Singh, Mauajama Firdaus, Jyotsna Agrawal, and Asif Ekbal. 2024. https://doi.org/10.18653/v1/2024.findings-naacl.290 On the way to gentle AI counselor: Politeness cause elicitation and intensity tagging in code-mixed H inglish conversations for social ...

    2024 doi

  25. [34]

    Furman, Franco Luque, Laura Alonso Alemany, and María Vanina Martínez

    Juan Manuel Pérez, Mariela Rajngewerc, Juan Carlos Giudici, Damián A. Furman, Franco Luque, Laura Alonso Alemany, and María Vanina Martínez. 2023. http://arxiv.org/abs/2106.09462 pysentimiento: A python toolkit for opinion mining and social nlp tasks

    2023 arXiv

  26. [35]

    Huachuan Qiu and Zhenzhong Lan. 2024. http://arxiv.org/abs/2408.15787 Interactive agents: Simulating counselor-client psychological counseling via role-playing llm-to-llm interactions

    2024 arXiv

  27. [36]

    Ashish Sharma, Inna W Lin, Adam S Miner, David C Atkins, and Tim Althoff. 2023. Human--ai collaboration enables more empathic conversations in text-based peer-to-peer mental health support. Nature Machine Intelligence, 5(1):46--57

    2023

  28. [37]

    Siqi Shen, Veronica Perez-Rosas, Charles Welch, Soujanya Poria, and Rada Mihalcea. 2022. https://doi.org/10.18653/v1/2022.acl-long.221 Knowledge enhanced reflection generation for counseling dialogues . In Proceedings of the 60th Annual Meeting of the Association for Computati...

    2022 doi

  29. [38]

    Siqi Shen, Charles Welch, Rada Mihalcea, and Ver \'o nica P \'e rez-Rosas. 2020. https://doi.org/10.18653/v1/2020.sigdial-1.2 Counseling-style reflection generation using generative pretrained transformers with augmented context . In Proceedings of the 21th Annual Meeting of t...

    2020 doi

  30. [39]

    Hao Sun, Zhenru Lin, Chujie Zheng, Siyang Liu, and Minlie Huang. 2021. https://doi.org/10.18653/v1/2021.findings-acl.130 P sy QA : A C hinese dataset for generating long counseling text for mental health support . In Findings of the Association for Computational Linguistics: A...

    2021 doi

  31. [40]

    Dobber, and Jos A

    Xin Sun, Jiahuan Pei, Jan de Wit, Mohammad Aliannejadi, Emiel Krahmer, Jos T.P. Dobber, and Jos A. Bosch. 2024. https://aclanthology.org/2024.lrec-main.498 Eliciting motivational interviewing skill codes in psychotherapy with LLM s: A bilingual dataset and analytical study . I...

    2024

  32. [41]

    Gregory A Thompson, Jonathan Segura, Dianne Cruz, Cassie Arnita, and Leeann H Whiffen. 2022. Cultural differences in patients’ preferences for paternalism: comparing mexican and american patients’ preferences for and experiences with physician paternalism and patient autonomy....

    2022

  33. [42]

    Anuradha Welivita and Pearl Pu. 2022. https://aclanthology.org/2022.coling-1.293 Curating a large-scale motivational interviewing dataset using peer support forums . In Proceedings of the 29th International Conference on Computational Linguistics, pages 3315--3330, Gyeongju, R...

    2022

  34. [43]

    Artur Zygadło. 2021. https://api.semanticscholar.org/CorpusID:245857997 A therapeutic dialogue agent for polish language . 2021 9th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW), pages 1--5

    2021

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.