Motivational Interviewing (MI) is a counseling technique that promotes behavioral change through reflective responses to mirror or refine client statements. While advanced Large Language Models (LLMs) can generate engaging dialogues, challenges remain for applying them in a sensitive context such as MI. This work assesses the potential of LLMs to generate MI reflections via three LLMs: GPT-4, Llama-2, and BLOOM, and explores the effect of dialogue context size and integration of MI strategies for reflection generation by LLMs. We conduct evaluations using both automatic metrics and human judges on four criteria: appropriateness, relevance, engagement, and naturalness, to assess whether these LLMs can accurately generate the nuanced therapeutic communication required in MI. While we demonstrate LLMs’ potential in generating MI reflections comparable to human therapists, content analysis shows that significant challenges remain. By identifying the strengths and limitations of LLMs in generating empathetic and contextually appropriate reflections in MI, this work contributes to the ongoing dialogue in enhancing LLM’s role in therapeutic counseling.

Large Language Models (LLMs) have shown promise in automating high-labor data tasks, but the adoption of LLMs in high-stake scenarios faces two key challenges: their tendency to answer despite uncertainty and their difficulty handling long input contexts robustly.We investigate commonly used off-the-shelf LLMs’ ability to identify low-confidence outputs for human review through “check set selection”–a process where LLMs prioritize information needing human judgment.Using a case study on social media monitoring for disaster risk management,we define the “check set” as a list of tweets escalated to the disaster manager when the LLM has the least confidence, enabling human oversight within budgeted effort.We test two strategies for LLM check set selection: *individual confidence elicitation* – LLMs assesses confidence for each tweet classification individually, requiring more prompts with shorter contexts, and *direct set confidence elicitation* – LLM evaluates confidence for a list of tweet classifications at once, using less prompts but longer contexts.Our results reveal that set selection via individual probabilities is more reliable but that direct set confidence merits further investigation.Direct set selection challenges include inconsistent outputs, incorrect check set size, and low inter-annotator agreement. Despite these challenges, our approach improves collaborative disaster tweet classification by outperforming random-sample check set selection, demonstrating the potential of human-LLM collaboration.

The adoption of large language models (LLMs) in high-stake scenarios continues to be a challenge due to lack of effective confidence calibration. Although LLMs are capable of providing convincing self-explanations and verbalizing confidence in NLP tasks, they tend to exhibit overconfidence when using generative or free-text rationales (e.g. Chain-of-Thought), where reasoning steps tend to lack verifiable grounding.In this paper, we investigate whether adding explanations in the form of extractive rationales –snippets of the input text that directly support the predictions, can improve the confidence calibration of LLMs in classification tasks.We examine two approaches for integrating these rationales: (1) a one-stage rationale-generation with prediction and (2) a two-stage rationale-guided confidence calibration.We evaluate these approaches on a disaster tweet classification task using four different off-the-shelf LLMs. Our results show that extracting rationales both before and after prediction can improve the confidence estimates of the LLMs. Furthermore, we find that replacing valid extractive rationales with irrelevant ones significantly lowers model confidence, highlighting the importance of rationale quality.This simple yet effective method improves LLM verbalized confidence and reduces overconfidence in possible hallucination.

This paper presents our system developed for the SemEval-2025 Task 9: The Food Hazard Detection Challenge. The shared task’s objective is to evaluate explainable classification systems for classifying hazards and products in two levels of granularity from web-collected food recall incident reports. In this work, we propose text augmentation techniques as a way to improve poor performance in minority classes and compare their effect for each category on various transformer and machine learning models. We apply three word-level data augmentation techniques, namely synonym replacement, random word swapping, and contextual word insertion utilizing BERT. The resultsshow that transformer models tend to have a better overall performance. Meanwhile, a statistically significant improvement (P 0.05) was observed in the fine-grained categories when using BERT to compare the baseline model with the three augmented models, which achieved a 6% increase in correct predictions for minority hazard classes. This suggests that targeted augmentation of minority classes can improve the performance of transformer models.

Motivational Interviewing is a counselling style that requires skillful usage of reflective listening and engaging in conversations about sensitive and personal subjects. In this paper, we investigate to what extent we can use generative large language models in motivational interviewing chatbots to generate precise and variable reflections on user responses. We conduct a two-step human evaluation where we first independently assess the generated reflections based on four criteria essential to health counseling; appropriateness, specificity, naturalness, and engagement. In the second step, we compare the overall quality of generated and human-authored reflections via a ranking evaluation. We use GPT-4, BLOOM, and FLAN-T5 models to generate motivational interviewing reflections, based on real conversational data collected via chatbots designed to provide support for smoking cessation and sexual health. We discover that GPT-4 can produce reflections of a quality comparable to human-authored reflections. Finally, we conclude that large language models have the potential to enhance and expand reflections in predetermined health counseling chatbots, but a comprehensive manual review is advised.

Proper recognition and interpretation of negation signals in text or communication is crucial for any form of full natural language understanding. It is also essential for computational approaches to natural language processing. In this study we focus on negation detection in Dutch spoken human-computer conversations. Since there exists no Dutch (dialogue) corpus annotated for negation we have annotated a Dutch corpus sample to evaluate our method for automatic negation detection. We use transfer learning and trained NegBERT (an existing BERT implementation used for negation detection) on English data with multilingual BERT to detect negation in Dutch dialogues. Our results show that adding in-domain training material improves the results. We show that we can detect both negation cues and scope in Dutch dialogues with high precision and recall. We provide a detailed error analysis and discuss the effects of cross-lingual and cross-domain transfer learning on automatic negation detection.

Digital recorded written and spoken dialogues are becoming increasingly available as an effect of the technological advances such as online messenger services and the use of chatbots. Summaries are a natural way of presenting the important information gathered from dialogues. We present a unique data set that consists of Dutch spoken human-computer conversations, an annotation layer of turn labels, and conversational abstractive summaries of user answers. The data set is publicly available for research purposes.

This paper makes the case for studying concreteness in language as a bridge that will allow language technology to support the understanding and improvement of ethnic inclusivity in job advertisements. We propose an annotation scheme that guides the assignment of sentences in job ads to classes that reflect concrete actions, i.e., what the employer needs people to do, and abstract dispositions, i.e., who the employer expects people to be. Using an annotated dataset of Dutch-language job ads, we demonstrate that machine learning technology is effectively able to distinguish these classes.

An important objective in health-technology is the ability to gather information about people’s well-being. Structured interviews can be used to obtain this information, but are time-consuming and not scalable. Questionnaires provide an alternative way to extract such information, though typically lack depth. In this paper, we present our first prototype of the BLISS agent, an artificial intelligent agent which intends to automatically discover what makes people happy and healthy. The goal of Behaviour-based Language-Interactive Speaking Systems (BLISS) is to understand the motivations behind people’s happiness by conducting a personalized spoken dialogue based on a happiness model. We built our first prototype of the model to collect 55 spoken dialogues, in which the BLISS agent asked questions to users about their happiness and well-being. Apart from a description of the BLISS architecture, we also provide details about our dataset, which contains over 120 activities and 100 motivations and is made available for usage.

We present REPORTS, an annotation scheme for the annotation of speech, attitude and perception reports. Such a scheme makes it possible to annotate the various text elements involved in such reports (e.g. embedding entity, complement, complement head) and their relations in a uniform way, which in turn facilitates the automatic extraction of information on, for example, complementation and vocabulary distribution. We also present the Ancient Greek corpus RAG (Thucydides’ History of the Peloponnesian War), to which we have applied this scheme using the annotation tool BRAT. We discuss some of the issues, both theoretical and practical, that we encountered, show how the corpus helps in answering specific questions, and conclude that REPORTS fitted in well with our needs.

We investigate modality in Portuguese and we combine a linguistic perspective with an application-oriented perspective on modality. We design an annotation scheme reflecting theoretical linguistic concepts and apply this schema to a small corpus sample to show how the scheme deals with real world language usage. We present two schemas for Portuguese, one for spoken Brazilian Portuguese and one for written European Portuguese. Furthermore, we use the annotated data not only to study the linguistic phenomena of modality, but also to train a practical text mining tool to detect modality in text automatically. The modality tagger uses a machine learning classifier trained on automatically extracted features from a syntactic parser. As we only have a small annotated sample available, the tagger was evaluated on 11 modal verbs that are frequent in our corpus and that denote more than one modal meaning. Finally, we discuss several valuable insights into the complexity of the semantic concept of modality that derive from the process of manual annotation of the corpus and from the analysis of the results of the automatic labeling: ambiguity and the semantic and syntactic properties typically associated to one modal meaning in context, and also the interaction of modality with negation and focus. The knowledge gained from the manual annotation task leads us to propose a new unified scheme for modality that applies to the two Portuguese varieties and covers both written and spoken data.

The present work is an overview of the TraMOOC (Translation for Massive Open Online Courses) research and innovation project, a machine translation approach for online educational content. More specifically, videolectures, assignments, and MOOC forum text is automatically translated from English into eleven European and BRIC languages. Unlike previous approaches to machine translation, the output quality in TraMOOC relies on a multimodal evaluation schema that involves crowdsourcing, error type markup, an error taxonomy for translation model comparison, and implicit evaluation via text mining, i.e. entity recognition and its performance comparison between the source and the translated text, and sentiment analysis on the students’ forum posts. Finally, the evaluation output will result in more and better quality in-domain parallel data that will be fed back to the translation engine for higher quality output. The translation service will be incorporated into the Iversity MOOC platform and into the VideoLectures.net digital library portal.

We present the process of building linguistic corpora of the Portuguese-related Gulf of Guinea creoles, a cluster of four historically related languages: Santome, Angolar, Principense and Fa d’Ambô. We faced the typical difficulties of languages lacking an official status, such as lack of standard spelling, language variation, lack of basic language instruments, and small data sets, which comprise data from the late 19th century to the present. In order to tackle these problems, the compiled written and transcribed spoken data collected during field work trips were adapted to a normalized spelling that was applied to the four languages. For the corpus compilation we followed corpus linguistics standards. We recorded meta data for each file and added morphosyntactic information based on a part-of-speech tag set that was designed to deal with the specificities of these languages. The corpora of three of the four creoles are already available and searchable via an online web interface.

We present our work in processing the Reference Corpus of Contemporary Portuguese and its publication online. After discussing how the corpus was built and our choice of meta-data, we turn to the processes and tools involved for the cleaning, preparation and annotation to make the corpus suitable for linguistic inquiries. The Web platform is described, and we show examples of linguistic resources that can be extracted from the platform for use in linguistic studies or in NLP.

We present a annotation scheme for modality in Portuguese. In our annotation scheme we have tried to combine a more theoretical linguistic viewpoint with a practical annotation scheme that will also be useful for NLP research but is not geared towards one specific application. Our notion of modality focuses on the attitude and opinion of the speaker, or of the subject of the sentence. We validated the annotation scheme on a corpus sample of approximately 2000 sentences that we fully annotated with modal information using the MMAX2 annotation tool to produce XML annotation. We discuss our main findings and give attention to the difficult cases that we encountered as they illustrate the complexity of modality and its interactions with other elements in the text.

Cet article présente une approche basée sur la comparaison fréquentielle de modèles lexicaux pour la segmentation automatique de textes historiques Portugais. Cette approche traite d’abord le problème de la segmentation comme un problème de classification, en attribuant à chaque élément lexical présent dans la phase d’apprentissage une valeur de saillance pour chaque type de segment. Ces modèles lexicaux permettent à la fois de produire une segmentation et de faire une analyse qualitative de textes historiques. Notre évaluation montre que l’approche adoptée permet de tirer de l’information sémantique que des approches se concentrant sur la détection des frontières séparant les segments ne peuvent acquérir.

We present the main outcomes of the COREA project: a corpus annotated with coreferential relations and a coreference resolution system for Dutch. In the project we developed annotation guidelines for coreference resolution for Dutch and annotated a corpus of 135K tokens. We discuss these guidelines, the annotation tool, and the inter-annotator agreement. We also show a visualization of the annotated relations. The standard approach to evaluate a coreference resolution system is to compare the predictions of the system to a hand-annotated gold standard test set (cross-validation). A more practically oriented evaluation is to test the usefulness of coreference relation information in an NLP application. We run experiments with an Information Extraction module for the medical domain, and measure the performance of this module with and without the coreference relation information. We present the results of both this application-oriented evaluation of our system and of a standard cross-validation evaluation. In a separate experiment we also evaluate the effect of coreference information produced by a simple rule-based coreference module in a Question Answering application.