Generative language models have changed the way we interact with computers using natural language. With the release of increasingly advanced GPT models, systems are able to correctly respond to questions in various domains. However, they still have important limitations, such as hallucinations, lack of substance in answers, inability to justify responses, or showing high confidence with fabricated content. In digital mental health, every decision must be traceable and based on scientific evidence and these shortcomings are hindering the integration of LLMs into clinical practice. In this paper, we provide a novel automated method to develop evidence-based question answering systems. Powerful state-of-the-art generalist language models are used and forced to employ only contents in validated clinical guidelines, tracking the source of the evidence for each generated response. This way, the system is able to protect users from hallucinatory responses. As a proof of concept, we present the results obtained building question-answering systems circumscribed to the clinical practice guidelines of the Spanish National Health System about the management of depression and attention deficit hyperactivity disorder. The coherence, veracity, and evidence supporting the responses have been evaluated by human experts obtaining high reliability, clarity, completeness, and traceability of evidence results.

Large Language Models (LLMs) have attracted interest due to their sophisticated natural language understanding capacities. Nevertheless, the employment of their d-vectors remains barely explored in the mental health domain despite LLMs’ potential. In this article, we perform feature selection strategies over the embeddings extracted from LLama-2 and MentaLlama models to solve fine-grain mental health topic classification by removing redundant features. All the proposals were evaluated across two realistic datasets with unbalanced topics, in which the use of the feature sets containing 1,000 of the 4,096 initial dimensions resulted in the most efficient solution, achieving a reduction of 75% on the complexity with minimal damage in W-F1 performance of the systems in the test set (a decrease of 1.51 percentage points for 7Cups dataset for Llama-2, and of 0.27 percentage points for Counsel-Chat for MentaLlama). Our findings suggest that applying feature selection approaches over the d-vectors extracted from LLMs could be beneficial, especially in cases with scarce computational resources.

3D models (or assets) that are present in many of modern software applications are first modeled by graphic designers using dedicated computer graphic tools and then integrated into such software applications or apps by software developers. This simple workflow/procedure requires developers to have a basic grounding in computer graphics, since 3D engines, libraries and third-party software are needed for this kind of integrations. Oftentimes, 3D designers are also required to customize or produce versions of a 3D model and thus, they must re-model all the assets before they are returned back to the developers for integration into the applications. This procedure also occurs whenever a modification or customization is requested. One possible significant improvement to this traditional, poorly automated workflow is to use services-oriented technology and features servitization to carry out the customization of 3D assets on-demand. In this article, we introduce$ \mu $S3D, an open-source microservices-based platform designed to support features relating to the customization of 3D models.$ \mu $S3D not only enables 3D assets to be customized without the need for computer graphic tools or designers, but also allows 3D models to be visualized through web technologies (e.g., HTML, Javascript and web component to visualize and interact with 3D models), thereby avoiding the development of computer graphics libraries or components in final software products. The article describes the elements that$ \mu $S3D comprises, explains how it works and presents a series of load tests to compare the performance (time consumption, CPU and memory utilization) of$ \mu $S3D when implemented and deployed as a microservices platform against a monolithic-based implementation, showing similar results with a low number of users (and requests) but reducing, on average, 64.32% the response time in the microservice-based implementation for a large number of users; reducing CPU utilization on microservice-based implementation and remaining the memory usage more or less constant in both implementations.

Abstract Conversational systems have become an element of everyday life for billions of users who use speech‐based interfaces to services, engage with personal digital assistants on smartphones, social media chatbots, or smart speakers. One of the most complex tasks in the development of these systems is to design the dialogue model, the logic that provided a user input selects the next answer. The dialogue model must also consider mechanisms to adapt the response of the system and the interaction style according to different groups and user profiles. Rule‐based systems are difficult to adapt to phenomena that were not taken into consideration at design‐time. However, many of the systems that are commercially available are based on rules, and so are the most widespread tools for the development of chatbots and speech interfaces. In this article, we present a proposal to: (a) automatically generate the dialogue rules from a dialogue corpus through the use of evolving algorithms, (b) adapt the rules according to the detected user intention. We have evaluated our proposal with several conversational systems of different application domains, from which our approach provided an efficient way for adapting a set of dialogue rules considering user utterance clusters.

T e-Coaches are promising intelligent systems that aims at supporting human everyday life, dispatching advices through different interfaces, such as apps, conversational interfaces and augmented reality interfaces. This workshop aims at exploring how e-coaches might benefit from spatially and time-multiplexed interfaces and from different communication modalities (e.g., text, visual, audio, etc.) according to the context of the interaction.

Abstract As life expectancy increases, it has become more necessary to find ways to support healthy ageing. A number of active ageing initiatives are being developed nowadays to foster healthy habits in the population. This paper presents our contribution to these initiatives in the form of a multimodal conversational coach that acts as a coach for physical activities. The agent can be developed as an Android app running on smartphones and coupled with cheap widely available sport sensors in order to provide meaningful coaching. It can be employed to prepare exercise sessions, provide feedback during the sessions, and discuss the results after the exercise. It incorporates an affective component that informs dynamic user models to produce adaptive interaction strategies.

Mobile health (m-Health) has emerged as a rapidly developing area that is transforming clinical research and health care on a global scale. In this paper, we describe a conversational app for the therapy of stroke rehabilitation. The main objective of the conversational app is to help recovering cognitive abilities of patients by means of a set of proposed exercises, which are divided into 8 categories focused on specific abilities. These categories have been defined after a detailed review of the guidelines for rehabilitation and training therapies. In addition, the application integrates a multimodal conversational interface to facilitate human-computer interaction, which has been specially designed for the elderly and patients with motor or visual or disabilities. The exercises provided by the application can be easily adapted to the specific users' requirements and preferences by means of the incorporation, deletion or modification of routines stored into a specific database isolated from the logic of the application.

Designing the rules for the dialog management process is oneof the most resources-consuming tasks when developing a dialog system. Although statistical approaches to dialog management are becoming mainstream in research and industrial contexts, still many systems are being developed following the rule-based or hybrid paradigms. For example, when developers require deterministic system responses to keep total control on the decisions made by the system, or because the infrastructure employed is designed for rule-based systems using technologies currently used in commercial platforms. In this paper, we propose the use of evolutionary algorithms to automatically obtain the dialog rules that are implicit in a dialog corpus. Our proposal makes it possible to exploit the benefits of statistical approaches to build rule-based systems. Our proposal has been evaluated with a practical spoken dialog system, for which we have automatically obtained a set of fuzzy rules to successfully manage the dialog.

Abstract Smart mobile devices have fostered new learning scenarios that demand sophisticated interfaces. Multimodal conversational agents have became a strong alternative to develop human‐machine interfaces that provide a more engaging and human‐like relationship between students and the system. The main developers of operating systems for such devices have provided application programming interfaces for developers to implement their own applications, including different solutions for developing graphical interfaces, sensor control and voice interaction. Despite the usefulness of such resources, there are no strategies defined for coupling the multimodal interface with the possibilities that these devices offer to enhance mobile educative apps with intelligent communicative capabilities and adaptation to the user needs. In this paper, we present a practical m‐learning application that integrates features of Android application programming interfaces on a modular architecture that emphasizes interaction management and context‐awareness to foster user‐adaptively, robustness and maintainability.

This paper proposes a statistical framework to develop user-adapted spoken dialog systems. The proposed framework integrates two main models. The first model is used to predict the user's intention during the dialog. The second model uses this prediction and the history of dialog up to the current moment to predict the next system response. This prediction is performed with an ensemble-based classifier trained for each of the tasks considered, so that a better selection of the next system can be attained weighting the outputs of these specialized classifiers. The codification of the information and the definition of data structures to store the data supplied by the user throughout the dialog makes the estimation of the models from the training data and practical domains manageable. We describe our proposal and its application and detailed evaluation in a practical spoken dialog system.

In this paper, we present an approach to spoken dialog management based on the use of a Stochastic Finite-State Transducer estimated from a dialog corpus. The states of the Stochastic Finite-State Transducer represent the dialog states, the input alphabet includes all the possible user utterances, without considering specific values, and the set of system answers constitutes the output alphabet. Then, a dialog describes a path in the transducer model from the initial state to the final one. An automatic dialog generation technique was used in order to generate the dialog corpus from which the transducer parameters are estimated. Our proposal for dialog management has been evaluated in a sport facilities booking task.

This paper proposes a new technique to enhance the performance of spoken dialogue systems which presents one novel contribution: the automatic correction of some ASR errors by using language models dependent on dialogue states, in conjunction with grammatical rules. These models are optimally selected by computing similarity scores between patterns obtained from uttered sentences and patterns learnt during training. Experimental results with a spoken dialogue system designed for the fast food domain show that our technique allows enhancing word accuracy, speech understanding and task completion rates of a spoken dialogue system by 8.5%, 16.54% and 44.17% absolute, respectively.

In this paper, we present an approach for the development of a statistical dialog manager, in which the system response is selected by means of a classification process which considers all the previous history of the dialog to select the next system response. In particular, we use decision trees for its implementation. The statistical model is automatically learned from training data which are labeled in terms of different SLU features. This methodology has been applied to develop a dialog manager within the framework of the European LUNA project, whose main goal is the creation of a robust natural spoken language understanding system. We present an evaluation of this approach for both human machine and human-human conversations acquired in this project. We demonstrate that a statistical dialog manager developed with the proposed technique and learned from a corpus of human-machine dialogs can successfully infer the task-related topics present in spontaneous humanhuman dialogs.

We are interested in the problem of understanding human conversation structure in the context of human-machine and human-human interaction. We present a statistical methodol-ogy for detecting the structure of spoken dialogs based on a generative model learned using decision trees. To evaluate our approach we have used the LUNA corpora, collected from real users engaged in problem solving tasks. The results of the evaluation show that automatic segmentation of spoken dialogs is very effective not only with models built using separately human-machine dialogs or human-human dialogs, but it is also possible to infer the task-related structure of human-human di-alogs with a model learned using only human-machine dialogs.

In this paper, we present the results of the comparison between three corpora acquired by means of different techniques. The first corpus was acquired using the Wizard of Oz technique. A statistical user simulation technique has been developed for the acquisition of the second corpus. In this technique, the next user answer is selected by means of a classification process that takes into account the previous user turns, the last system answer and the objective of the dialog. Finally, a dialog simulation technique has been developed for the acquisition of the third corpus. This technique uses a random selection of the user and system turns, defining stop conditions for automatically deciding if the simulated dialog is successful or not. We use several evaluation measures proposed in previous research to compare between our three acquired corpora, and then discuss the similarities and differences with regard to these measures.

In this paper, we present a platform for the development of dialog systems. The architecture is organized in modules, each of which accomplishes a specific objective. Communication among modules is done by means of transmission of XML data packages through sockets. This platform allows for multimodal communication and permits an easy interchange of modules, so that the change of language or task is straightforward. In our system, most of the modules use stochastic models for the representation of the different knowledge sources involved in the dialog process. These models are automatically learnt from corpora of training data. We also present a dialog system developed with this platform. The task of the dialog system presented is the access to information about train timetables and prices of trains in Spanish.

In this article, we present an approach for the construction of a stochastic dialog manager, in which the system answer is selected by means of a classification procedure. In particular, we use neural networks for the implementation of this classification process, which takes into account the data supplied by the user and the last system turn. The stochastic model is automatically learnt from training data which are labeled in terms of dialog acts. An important characteristic of this approach is the introduction of a partition in the space of sequences of dialog acts in order to deal with the scarcity of available training data. This system has been developed in the DIHANA project, whose goal is the design and development of a dialog system to access a railway information system using spontaneous speech in Spanish. An evaluation of this approach is also presented. 1.

This paper presents a Passage Retrieval-based approach for the development of spoken Question Answering systems. Question Answering can be seen as a particular aspect of Information Retrieval where the user's information need are not satisfied by means of a document, but by a portion of text. Currently, the performance of typical Question Answering systems is rather poor, if compared with other Information Retrieval tasks, however, it is predictable that these system will be used in combination with some speech interface when high precision will be achieved. The most important evaluation competitions for Question Answering, such as TREC and CLEF, still do not have a spoken Question Answering track, therefore this work presents a novel approach to this task in order to study the influence of recognition errors over Question Answering systems.