Evaluation Of Dialogue Systems Research Articles

The Art of Dialogue: Synergistic Effects of Language Adaptation and Emotional Intelligence in AI Systems

Artificial intelligence (AI) systems are increasingly applied in areas such as voice assistants, customer service, and healthcare. However, current AI systems often lack semantic understanding and emotional expression capabilities when interacting with humans, leading to poor user experiences. To enable AI systems to converse more naturally with humans, it is necessary to synergistically optimize two key capabilities: language adaptation and emotional intelligence. This paper reviews the latest research progress in AI dialogue systems regarding language adaptation and emotional intelligence. In terms of language adaptation, this paper summarizes technologies such as personalized dialogue, knowledge-enhanced dialogue, and multi-turn contextual dialogue. In terms of emotional intelligence, this paper outlines work in areas such as emotion recognition, emotional dialogue generation, and empathetic dialogue. Additionally, this paper discusses strategies for synergistic optimization of language adaptation and emotional intelligence, as well as challenges in dialogue system evaluation. Finally, this paper looks ahead to future directions for AI dialogue systems.

Interactive probes: Towards action-level evaluation for dialogue systems

Measures of ‘humanness’, ‘coherence’ or ‘fluency’ are the mainstay of dialogue system evaluation, but they don’t target system capabilities and rarely offer actionable feedback. Reviewing recent work in this domain, we identify an opportunity for evaluation at the level of action sequences, rather than the more commonly targeted levels of whole conversations or single responses. We introduce interactive probes, an evaluation framework inspired by empirical work on social interaction that can help to systematically probe the capabilities of dialogue systems. We sketch some first probes in the domains of tellings and repair, two sequence types ubiquitous in human interaction and challenging for dialogue systems. We argue interactive probing can offer the requisite flexibility to keep up with developments in interactive language technologies and do justice to the open-endedness of action formation and ascription in interaction.

Contextual Conversational Agent to Address Vaccine Hesitancy: Protocol for a Design-Based Research Study.

BackgroundSince the beginning of the COVID-19 pandemic, people have been exposed to misinformation, leading to many myths about SARS-CoV-2 and the vaccines against it. As this situation does not seem to end soon, many authorities and health organizations, including the World Health Organization (WHO), are utilizing conversational agents (CAs) in their fight against it. Although the impact and usage of these novel digital strategies are noticeable, the design of the CAs remains key to their success.ObjectiveThis study describes the use of design-based research (DBR) for contextual CA design to address vaccine hesitancy. In addition, this protocol will examine the impact of DBR on CA design to understand how this iterative process can enhance accuracy and performance.MethodsA DBR methodology will be used for this study. Each phase of analysis, design, and evaluation of each design cycle inform the next one via its outcomes. An anticipated generic strategy will be formed after completing the first iteration. Using multiple research studies, frameworks and theoretical approaches are tested and evaluated through the different design cycles. User perception of the CA will be analyzed or collected by implementing a usability assessment during every evaluation phase using the System Usability Scale. The PARADISE (PARAdigm for Dialogue System Evaluation) method will be adopted to calculate the performance of this text-based CA.ResultsTwo phases of the first design cycle (design and evaluation) were completed at the time of this writing (April 2022). The research team is currently reviewing the natural-language understanding model as part of the conversation-driven development (CDD) process in preparation for the first pilot intervention, which will conclude the CA’s first design cycle. In addition, conversational data will be analyzed quantitatively and qualitatively as part of the reflection and revision process to inform the subsequent design cycles. This project plans for three rounds of design cycles, resulting in various studies spreading outcomes and conclusions. The results of the first study describing the entire first design cycle are expected to be submitted for publication before the end of 2022.ConclusionsCAs constitute an innovative way of delivering health communication information. However, they are primarily used to contribute to behavioral change or educate people about health issues. Therefore, health chatbots’ impact should be carefully designed to meet outcomes. DBR can help shape a holistic understanding of the process of CA conception. This protocol describes the design of VWise, a contextual CA that aims to address vaccine hesitancy using the DBR methodology. The results of this study will help identify the strengths and flaws of DBR’s application to such innovative projects.

Maintaining Common Ground in Dynamic Environments

Abstract Common grounding is the process of creating and maintaining mutual understandings, which is a critical aspect of sophisticated human communication. While various task settings have been proposed in existing literature, they mostly focus on creating common ground under a static context and ignore the aspect of maintaining them overtime under dynamic context. In this work, we propose a novel task setting to study the ability of both creating and maintaining common ground in dynamic environments. Based on our minimal task formulation, we collected a large-scale dataset of 5,617 dialogues to enable fine-grained evaluation and analysis of various dialogue systems. Through our dataset analyses, we highlight novel challenges introduced in our setting, such as the usage of complex spatio-temporal expressions to create and maintain common ground. Finally, we conduct extensive experiments to assess the capabilities of our baseline dialogue system and discuss future prospects of our research.

Survey on evaluation methods for dialogue systems

In this paper, we survey the methods and concepts developed for the evaluation of dialogue systems. Evaluation, in and of itself, is a crucial part during the development process. Often, dialogue systems are evaluated by means of human evaluations and questionnaires. However, this tends to be very cost- and time-intensive. Thus, much work has been put into finding methods which allow a reduction in involvement of human labour. In this survey, we present the main concepts and methods. For this, we differentiate between the various classes of dialogue systems (task-oriented, conversational, and question-answering dialogue systems). We cover each class by introducing the main technologies developed for the dialogue systems and then present the evaluation methods regarding that class.

Predictive Engagement: An Efficient Metric for Automatic Evaluation of Open-Domain Dialogue Systems

User engagement is a critical metric for evaluating the quality of open-domain dialogue systems. Prior work has focused on conversation-level engagement by using heuristically constructed features such as the number of turns and the total time of the conversation. In this paper, we investigate the possibility and efficacy of estimating utterance-level engagement and define a novel metric, predictive engagement, for automatic evaluation of open-domain dialogue systems. Our experiments demonstrate that (1) human annotators have high agreement on assessing utterance-level engagement scores; (2) conversation-level engagement scores can be predicted from properly aggregated utterance-level engagement scores. Furthermore, we show that the utterance-level engagement scores can be learned from data. These scores can be incorporated into automatic evaluation metrics for open-domain dialogue systems to improve the correlation with human judgements. This suggests that predictive engagement can be used as a real-time feedback for training better dialogue models.

Evaluation of Dialogue Systems with Different End of Sentences Using Seq2Seq Model

Evaluation of Dialogue Systems with Different End of Sentences Using Seq2Seq Model

Re-Evaluating ADEM: A Deeper Look at Scoring Dialogue Responses

Automatically evaluating the quality of dialogue responses for unstructured domains is a challenging problem. ADEM (Lowe et al. 2017) formulated the automatic evaluation of dialogue systems as a learning problem and showed that such a model was able to predict responses which correlate significantly with human judgements, both at utterance and system level. Their system was shown to have beaten word-overlap metrics such as BLEU with large margins. We start with the question of whether an adversary can game the ADEM model. We design a battery of targeted attacks at the neural network based ADEM evaluation system and show that automatic evaluation of dialogue systems still has a long way to go. ADEM can get confused with a variation as simple as reversing the word order in the text! We report experiments on several such adversarial scenarios that draw out counterintuitive scores on the dialogue responses. We take a systematic look at the scoring function proposed by ADEM and connect it to linear system theory to predict the shortcomings evident in the system. We also devise an attack that can fool such a system to rate a response generation system as favorable. Finally, we allude to future research directions of using the adversarial attacks to design a truly automated dialogue evaluation system.

An Evaluation of Chinese Human-Computer Dialogue Technology

The human-computer dialogue has recently attracted extensive attention from both academia and industry as an important branch in the field of artificial intelligence (AI). However, there are few studies on the evaluation of large-scale Chinese human-computer dialogue systems. In this paper, we introduce the Second Evaluation of Chinese Human-Computer Dialogue Technology, which focuses on the identification of a user's intents and intelligent processing of intent words. The Evaluation consists of user intent classification (Task 1) and online testing of task-oriented dialogues (Task 2), the data sets of which are provided by iFLYTEK Corporation. The evaluation tasks and data sets are introduced in detail, and meanwhile, the evaluation results and the existing problems in the evaluation are discussed.

Interactive Hesitation Synthesis: Modelling and Evaluation

Conversational spoken dialogue systems that interact with the user rather than merely reading the text can be equipped with hesitations to manage dialogue flow and user attention. Based on a series of empirical studies, we elaborated a hesitation synthesis strategy for dialogue systems, which inserts hesitations of a scalable extent wherever needed in the ongoing utterance. Previously, evaluations of hesitation systems have shown that synthesis quality is affected negatively by hesitations, but that they result in improvements of interaction quality. We argue that due to its conversational nature, hesitation synthesis needs interactive evaluation rather than traditional mean opinion score (MOS)-based questionnaires. To validate this claim, we dually evaluate our system’s speech synthesis component, on the one hand, linked to the dialogue system evaluation, and on the other hand, in a traditional MOS way. We are thus able to analyze and discuss differences that arise due to the evaluation methodology. Our results suggest that MOS scales are not sufficient to assess speech synthesis quality, leading to implications for future research that are discussed in this paper. Furthermore, our results indicate that synthetic hesitations are able to increase task performance and that an elaborated hesitation strategy is necessary to avoid likability issues.

Usefulness, localizability, humanness, and language-benefit: additional evaluation criteria for natural language dialogue systems

Human---computer dialogue systems interact with human users using natural language. We used the ALICE/AIML chatbot architecture as a platform to develop a range of chatbots covering different languages, genres, text-types, and user-groups, to illustrate qualitative aspects of natural dialogue system evaluation. We present some of the different evaluation techniques used in natural dialogue systems, including black box and glass box, comparative, quantitative, and qualitative evaluation. Four aspects of NLP dialogue system evaluation are often overlooked: usefulness in terms of a user's qualitative needs, localizability to new genres and languages, humanness or naturalness compared to human---human dialogues, and language benefit compared to alternative interfaces. We illustrated these aspects with respect to our work on machine-learnt chatbot dialogue systems; we believe these aspects are worthwhile in impressing potential new users and customers.

Unsupervised Spoken Language Understanding for a Multi-Domain Dialog System

This paper proposes an unsupervised spoken language understanding (SLU) framework for a multi-domain dialog system. Our unsupervised SLU framework applies a non-parametric Bayesian approach to dialog acts, intents and slot entities, which are the components of a semantic frame. The proposed approach reduces the human effort necessary to obtain a semantically annotated corpus for dialog system development. In this study, we analyze clustering results using various evaluation metrics for four dialog corpora. We also introduce a multi-domain dialog system that uses the unsupervised SLU framework. We argue that our unsupervised approach can help overcome the annotation acquisition bottleneck in developing dialog systems. To verify this claim, we report a dialog system evaluation, in which our method achieves competitive results in comparison with a system that uses a manually annotated corpus. In addition, we conducted several experiments to explore the effect of our approach on reducing development costs. The results show that our approach be helpful for the rapid development of a prototype system and reducing the overall development costs.

A domain-independent statistical methodology for dialog management in spoken dialog systems

This paper proposes a domain-independent statistical methodology to develop dialog managers for spoken dialog systems. Our methodology employs a data-driven classification procedure to generate abstract representations of system turns taking into account the previous history of the dialog. A statistical framework is also introduced for the development and evaluation of dialog systems created using the methodology, which is based on a dialog simulation technique. The benefits and flexibility of the proposed methodology have been validated by developing statistical dialog managers for four spoken dialog systems of different complexity, designed for different languages (English, Italian, and Spanish) and application domains (from transactional to problem-solving tasks). The evaluation results show that the proposed methodology allows rapid development of new dialog managers as well as to explore new dialog strategies, which permit developing new enhanced versions of already existing systems.

Predicting User Satisfaction in Spoken Dialog System Evaluation With Collaborative Filtering

We propose a collaborative filtering (CF) model to predict user satisfaction in SDS evaluation. Inspired by the use of CF in recommendation systems, where a user's preference for a new item is assume to resemble that for similar items rated previously, we adapt the idea to predict user evaluations of unrated dialogs based on the ratings received by similar dialogs. Ratings of dialogs are gathered by crowdsourcing through Amazon Mechanical Turk. A reference baseline is provided by a linear regression model (LRM) based on the PARADISE framework. We present two versions of the CF model. First, the item-based collaborative filtering model (ICFM) clusters rated dialogs and builds an LRM for each cluster. The rating of an unseen dialog is predicted by the LRM of its most similar cluster. Second, the extended ICFM (EICFM) separates dialog features into user-related and system-related groups, to build LRMs for these separately. Experimental results on dialogs from the Let's Go! system show both ICFM and EICFM can significantly improve the proportion of variability explained by the LRM. We also demonstrate the generalizability of the CF model to a new dialog corpus from the systems in the Spoken Dialog Challenge (SDC) 2010.

Recent Approaches to Dialog Management for Spoken Dialog Systems

A field of spoken dialog systems is a rapidly growing research area because the performance improvement of speech technologies motivates the possibility of building systems that a human can easily operate in order to access useful information via spoken languages. Among the components in a spoken dialog system, the dialog management plays major roles such as discourse analysis, database access, error handling, and system action prediction. This survey covers design issues and recent approaches to the dialog management techniques for modeling the dialogs. We also explain the user simulation techniques for automatic evaluation of spoken dialog systems.

Non-verbal communication strategies to improve robustness in dialogue systems: a comparative study

This paper explores the use of embodied conversational agents (ECAs) and their visual communicative ability to improve interaction with spoken language dialogue systems (SLDSs) through an experimental case study in the application context of secure access by speaker verification followed by remote home automation control. After identifying a set of typical interaction problems with SLDSs and associated with each of them a particular ECA gesture or behaviour, we conducted a comparative evaluation based on ITU recommendations for the evaluation of spoken dialogue systems. User tests were carried out dividing the test users into two groups, each facing a different interface setup: one with an ECA, and the other only with voice output. The ECA group encountered fewer interaction problems. Users’ impressions, however, were similar in both groups, with a slight advantage observed for the ECA group. In particular, the ECA seems to help users to better understand the flow of the dialogue and reduce confusion. Results also suggest that rejection (based on privacy and security concerns) is a dimension in its own right that may influence subjective evaluation parameters closely related to user acceptance.

Analysis of a new simulation approach to dialog system evaluation

The evaluation of spoken dialog systems still relies on subjective interaction experiments for quantifying interaction behavior and user-perceived quality. In this paper, we present a simulation approach replacing subjective tests in early system design and evaluation phases. The simulation is based on a model of the system, and a probabilistic model of user behavior. Probabilities for the next user action vary in dependence of system features and user characteristics, as defined by rules. This way, simulations can be conducted before data have been acquired. In order to evaluate the simulation approach, characteristics of simulated interactions are compared to interaction corpora obtained in subjective experiments. As was previously proposed in the literature, we compare interaction parameters for both corpora and calculate recall and precision of user utterances. The results are compared to those from a comparison of real user corpora. While the real corpora are not equal, they are more similar than the simulation is to the real data. However, the simulations can predict differences between system versions and user groups quite well on a relative level. In order to derive further requirements for the model, we conclude with a detailed analysis of utterances missing in the simulated corpus and consider the believability of entire dialogs.

Data-driven user simulation for automated evaluation of spoken dialog systems

This paper proposes a novel integrated dialog simulation technique for evaluating spoken dialog systems. A data-driven user simulation technique for simulating user intention and utterance is introduced. A novel user intention modeling and generating method is proposed that uses a linear-chain conditional random field, and a two-phase data-driven domain-specific user utterance simulation method and a linguistic knowledge-based ASR channel simulation method are also presented. Evaluation metrics are introduced to measure the quality of user simulation at intention and utterance. Experiments using these techniques were carried out to evaluate the performance and behavior of dialog systems designed for car navigation dialogs and a building guide robot, and it turned out that our approach was easy to set up and showed similar tendencies to real human users.

Evaluating user simulations with the Cramér–von Mises divergence

User simulations are increasingly employed in the development and evaluation of spoken dialog systems. However, there is no accepted method for evaluating user simulations, which is problematic because the performance of new dialog management techniques is often evaluated on user simulations alone, not on real people. In this paper, we propose a novel method of evaluating user simulations. We view a user simulation as a predictor of the performance of a dialog system, where per-dialog performance is measured with a domain-specific scoring function. The divergence between the distribution of dialog scores in the real and simulated corpora provides a measure of the quality of the user simulation, and we argue that the Cramér–von Mises divergence is well-suited to this task. To demonstrate this technique, we study a corpus of callers with real information needs and show that Cramér-von Mises divergence conforms to expectations. Finally, we present simple tools which enable practitioners to interpret the statistical significance of comparisons between user simulations.

Relations between de-facto criteria in the evaluation of a spoken dialogue system

Evaluation of spoken dialogue systems has been traditionally carried out in terms of instrumentally or expert-derived measures (usually called “objective” evaluation) and quality judgments of users who have previously interacted with the system (also called “subjective” evaluation). Different research efforts have been made to extract relationships between these evaluation criteria. In this paper we report empirical results obtained from statistical studies, which were carried out on interactions of real users with our spoken dialogue system. These studies have rarely been exploited in the literature. Our results show that they can indicate important relationships between criteria, which can be used as guidelines for refinement of the systems under evaluation, as well as contributing to the state-of-the-art knowledge about how quantitative aspects of the systems affect the user’s perceptions about them.