In this paper, we present an approach for extracting knowledge graph information for retrieval augmented generation in dialogue systems. Knowledge graphs are a rich source of background information, but the inclusion of more potentially useful information in a system prompt risks decreased model performance from excess context. We investigate a method of retrieving relevant subgraphs of maximum relevance and minimum size by framing this trade-off as a Prize-collecting Steiner Tree problem. The results of our user study and analysis indicate promising efficacy of a simple subgraph retrieval approach compared with a top-K retrieval model.

Engaging in discussions that involve diverse perspectives and exchanging arguments on a controversial issue is a natural way for humans to form opinions. In this process, the way arguments are presented plays a crucial role in determining how engaged users are, whether the interaction takes place solely among humans or within human-agent teams. This is of great importance as user engagement plays a crucial role in determining the success or failure of cooperative argumentative discussions. One main goal is to maintain the user’s motivation to participate in a reflective opinion-building process, even when addressing contradicting viewpoints. This work investigates how non-verbal agent behavior, specifically co-speech gestures, influences the user’s engagement and interest during an ongoing argumentative interaction. The results of a laboratory study conducted with 56 participants demonstrate that the agent’s co-speech gestures have a substantial impact on user engagement and interest and the overall perception of the system. Therefore, this research offers valuable insights for the design of future cooperative argumentative virtual agents.

Explainable artificial intelligence (XAI) is a rapidly evolving field that seeks to create AI systems that can provide humanunderstandable explanations for their decisionmaking processes.However, these explanations rely on model and data-specific information only.To support better human decisionmaking, integrating domain knowledge into AI systems is expected to enhance understanding and transparency.In this paper, we present an approach for combining XAI explanations with domain knowledge within a dialogue system.We concentrate on techniques derived from the field of computational argumentation to incorporate domain knowledge and corresponding explanations into human-machine dialogue.We implement the approach in a prototype system for an initial user evaluation, where users interacted with the dialogue system to receive predictions from an underlying AI model.The participants were able to explore different types of explanations and domain knowledge.Our results indicate that users tend to more effectively evaluate model performance when domain knowledge is integrated.On the other hand, we found that domain knowledge was not frequently requested by the user during dialogue interactions.

This paper investigates the enhancement of Dialogue Systems by integrating the creative capabilities of Large Language Models.While traditional Dialogue Systems focus on understanding user input and selecting appropriate system actions, Language Models excel at generating natural language text based on prompts.Therefore, we propose to improve controllability and coherence of interactions by guiding a Language Model with control signals that enable explicit control over the system behaviour.To address this, we tested and evaluated our concept in 815 conversations with over 3600 dialogue exchanges on a dataset.Our experiment examined the quality of generated system responses using two strategies: An unguided strategy where task data was provided to the models, and a controlled strategy in which a simulated Dialogue Controller provided appropriate system actions.The results show that the average BLEU score and the classification of dialogue acts improved in the controlled Natural Language Generation.

In this work, we study dialogue scenarios that start from chit-chat but eventually switch to task-related services, and investigate how a unified dialogue model, which can engage in both chit-chat and task-oriented dialogues, takes the initiative during the dialogue mode transition from chit-chat to task-oriented in a coherent and cooperative manner.We firstly build a transition info extractor (TIE) that keeps track of the preceding chit-chat interaction and detects the potential user intention to switch to a taskoriented service.Meanwhile, in the unified model, a transition sentence generator (TSG) is extended through efficient Adapter tuning and transition prompt learning.When the TIE successfully finds task-related information from the preceding chit-chat, such as a transition domain ("train" in Figure 1), then the TSG is activated automatically in the unified model to initiate this transition by generating a transition sentence under the guidance of transition information extracted by TIE.The experimental results show promising performance regarding the proactive transitions.We achieve an additional large improvement on TIE model by utilizing Conditional Random Fields (CRF).The TSG can flexibly generate transition sentences while maintaining the unified capabilities of normal chit-chat and task-oriented response generation.

Humans form opinions and justify different points of view by exchanging arguments and knowledge. Likewise to human-human interaction, the way arguments are presented influence the user's willingness to engage into a critical reflection. Especially when interacting with conversational agents the user's engagement and motivation are important factors and highly influence the success or failure of such a mixed team. To maintain the users' trust and satisfaction, the users' perception of the respective system is an important indicator. Thus, this work investigates the design of a cooperative argumentative dialogue system using a virtual avatar compared to a non-avatar interface by evaluating a crowdsourcing study conducted with 84 participants. The results indicate, that the avatar system is perceived as significantly more appealing and natural and thus, engaging which also influences the acceptance and perception of the quality of presented arguments. Furthermore, we found that the presence of the avatar often led to an increase in the anticipated level of conversational proficiency similar to that of a human interlocutor. Therefore, this work provides important insights for the design of future cooperative argumentative virtual avatar interfaces.

Human users tend to selectively ignore information that contradicts their pre-existing beliefs or opinions in their process of information seeking.These "self-imposed filter bubbles" (SFB) pose a significant challenge for cooperative argumentative dialogue systems aiming to build an unbiased opinion and a better understanding of the topic at hand.To address this issue, we develop a strategy for overcoming users' SFB within the course of the interaction.By continuously modeling the user's position in relation to the SFB, we are able to identify the respective arguments which maximize the probability to get outside the SFB and present them to the user.We implemented this approach in an argumentative dialogue system and evaluated in a laboratory user study with 60 participants to show its validity and applicability.The findings suggest that the strategy was successful in breaking users' SFBs and promoting a more reflective and comprehensive discussion of the topic.

Speech interfaces for argumentative dialogue systems (ADS) are rather scarce. The complex task they pursue hinders the application of common natural language understanding (NLU) approaches in this domain. To address this issue we include an adaption of a recently introduced NLU framework tailored to argumentative tasks into a complete ADS. We evaluate the likeability and motivation of users to interact with the new system in a user study. Therefore, we compare it to a solid baseline utilizing a drop-down menu. The results indicate that the integration of a flexible NLU framework enables a far more natural and satisfying interaction with human users in real-time. Even though the drop-down menu convinces regarding its robustness, the willingness to use the new system is significantly higher. Hence, the featured NLU framework provides a sound basis to build an intuitive interface which can be extended to adapt its behavior to the individual user.

Most systems helping to provide structured information and support opinion building, discuss with users without considering their individual interest. The scarce existing research on user interest in dialogue systems depends on explicit user feedback. Such systems require user responses that are not content-related and thus, tend to disturb the dialogue flow. In this paper, we present a novel model for implicitly estimating user interest during argumentative dialogues based on semantically clustered data. Therefore, an online user study was conducted to acquire training data which was used to train a binary neural network classifier in order to predict whether or not users are still interested in the content of the ongoing dialogue. We achieved a classification accuracy of 74.9% and furthermore investigated with different Artificial Neural Networks (ANN) which new argument would fit the user interest best.

To build a well-founded opinion it is natural for humans to gather and exchange new arguments. Especially when being confronted with an overwhelming amount of information, people tend to focus on only the part of the available information that fits into their current beliefs or convenient opinions. To overcome this “self-imposed filter bubble” (SFB) in the information seeking process, it is crucial to identify influential indicators for the former. Within this paper we propose and investigate indicators for the the user’s SFB, mainly their Reflective User Engagement (RUE), their Personal Relevance (PR) ranking of content-related subtopics as well as their False (FK) and True Knowledge (TK) on the topic. Therefore, we analysed the answers of 202 participants of an online conducted user study, who interacted with our argumentative dialogue system BEA (“Building Engaging Argumentation”). Moreover, also the influence of different input/output modalities (speech/speech and drop-down menu/text) on the interaction with regard to the suggested indicators was investigated.

Niklas Rach, Carolin Schindler, Isabel Feustel, Johannes Daxenberger, Wolfgang Minker, Stefan Ultes. Proceedings of the 22nd Annual Meeting of the Special Interest Group on Discourse and Dialogue. 2021.

Recently, principal reward components for dialogue policy reinforcement learning use task success and user satisfaction independently and neither the resulting learned behaviour has been analysed nor a suitable proper analysis method even existed.In this work, we employ both principal reward components jointly and propose a method to analyse the resulting behaviour through a structured way of probing the learned policy.We show that blending both reward components increases user satisfaction without sacrificing task success even in more hostile environments and provide insight about actions chosen by the learned policies.

This work introduces chained arguments into a dialogue game for argumentation to allow a more natural and intuitive interaction with a respective system. Thus, the turn taking rules of the game are improved while still preserving the general consistency that is ensured by the framework. The improved system is used to generate artificial dialogues between two virtual agents which are assessed in a user study. The results show a significant improvement in the perceived naturalness without violating the logical consistency.

We present a neural network approach to estimate the communication style of spoken interaction, namely the stylistic variations elaborateness and directness, and investigate which type of input features to the estimator are necessary to achive good performance. First, we describe our annotated corpus of recordings in the health care domain and analyse the corpus statistics in terms of agreement, correlation and reliability of the ratings. We use this corpus to estimate the elaborateness and the directness of each utterance. We test different feature sets consisting of dialogue act features, grammatical features and linguistic features as input for our classifier and perform classification in two and three classes. Our classifiers use only features that can be automatically derived during an ongoing interaction in any spoken dialogue system without any prior annotation. Our results show that the elaborateness can be classified by only using the dialogue act and the amount of words contained in the corresponding utterance. The directness is a more difficult classification task and additional linguistic features in form of word embeddings improve the classification results. Afterwards, we run a comparison with a support vector machine and a recurrent neural network classifier.

Paraphrasing is an important aspect of natural-language generation that can produce more variety in the way specific content is presented. Traditionally, paraphrasing has been focused on finding different words that convey the same meaning. However, in human-human interaction, we regularly express our intention with phrases that are vastly different regarding both word content and syntactic structure. Instead of exchanging only individual words, the complete surface realisation of a sentences is altered while still preserving its meaning and function in a conversation. This kind of contextual paraphrasing did not yet receive a lot of attention from the scientific community despite its potential for the creation of more varied dialogues. In this work, we evaluate several existing approaches to sentence encoding with regard to their ability to capture such context-dependent paraphrasing. To this end, we define a paraphrase classification task that incorporates contextual paraphrases, perform dialogue act clustering, and determine the performance of the sentence embeddings in a sentence swapping task.

We present an approach to evaluate argument search techniques in view of their use in argumentative dialogue systems by assessing quality aspects of the retrieved arguments. To this end, we introduce a dialogue system that presents arguments by means of a virtual avatar and synthetic speech to users and allows them to rate the presented content in four different categories (Interesting, Convincing, Comprehensible, Relation). The approach is applied in a user study in order to compare two state of the art argument search engines to each other and with a system based on traditional web search. The results show a significant advantage of the two search engines over the baseline. Moreover, the two search engines show significant advantages over each other in different categories, thereby reflecting strengths and weaknesses of the different underlying techniques.

The differences in decision making between behavioural models of voice interfaces are hard to capture using existing measures for the absolute performance of such models.For instance, two models may have a similar task success rate, but very different ways of getting there.In this paper, we propose a general methodology to compute the similarity of two dialogue behaviour models and investigate different ways of computing scores on both the semantic and the textual level.Complementing absolute measures of performance, we test our scores on three different tasks and show the practical usability of the measures.

Learning suitable and well-performing dialogue behaviour in statistical spoken dialogue systems has been in the focus of research for many years.While most work which is based on reinforcement learning employs an objective measure like task success for modelling the reward signal, we use a reward based on user satisfaction estimation.We propose a novel estimator and show that it outperforms all previous estimators while learning temporal dependencies implicitly.Furthermore, we apply this novel user satisfaction estimation model live in simulated experiments where the satisfaction estimation model is trained on one domain and applied in many other domains which cover a similar task.We show that applying this model results in higher estimated satisfaction, similar task success rates and a higher robustness to noise.

This work discusses the formulation of argumentative dialogue as Markov game. We show how formal systems for persuasive dialogues that adhere to a certain structure can be reformulated as Markov games and thus be addressed as Reinforcement Learning task in a multi-agent setting. We validate our approach on an implementation of a proof of principle scenario where we show that the optimal policy can be learned.

Paweł Budzianowski, Tsung-Hsien Wen, Bo-Hsiang Tseng, Iñigo Casanueva, Stefan Ultes, Osman Ramadan, Milica Gašić. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018.

This work introduces EVA, a multimodal argumentative Dialogue System that is capable of discussing controversial topics with the user. The interaction is structured as an argument game in which the user and the system select respective moves in order to convince their opponent. EVA's response is presented as a natural language utterance by a virtual agent that supports the respective content using characteristic gestures and mimic.

Reinforcement learning (RL) is a promising dialogue policy optimisation approach, but traditional RL algorithms fail to scale to large domains.Recently, Feudal Dialogue Management (FDM), has shown to increase the scalability to large domains by decomposing the dialogue management decision into two steps, making use of the domain ontology to abstract the dialogue state in each step.In order to abstract the state space, however, previous work on FDM relies on handcrafted feature functions.In this work, we show that these feature functions can be learned jointly with the policy model while obtaining similar performance, even outperforming the handcrafted features in several environments and domains.

In cooperative dialogues, identifying the intent of ones conversation partner and acting accordingly is of great importance.While this endeavour is facilitated by phrasing intentions as directly as possible, we can observe in human-human communication that a number of factors such as cultural norms and politeness may result in expressing one's intent indirectly.Therefore, in human-computer communication we have to anticipate the possibility of users being indirect and be prepared to interpret their actual meaning.Furthermore, a dialogue system should be able to conform to human expectations by adjusting the degree of directness it uses to improve the user experience.To reach those goals, we propose an approach to differentiate between direct and indirect utterances and find utterances of the opposite characteristic that express the same intent.In this endeavour, we employ dialogue vector models and recurrent neural networks.

Cross-domain natural language generation (NLG) is still a difficult task within spoken dialogue modelling.Given a semantic representation provided by the dialogue manager, the language generator should generate sentences that convey desired information.Traditional template-based generators can produce sentences with all necessary information, but these sentences are not sufficiently diverse.With RNN-based models, the diversity of the generated sentences can be high, however, in the process some information is lost.In this work, we improve an RNN-based generator by considering latent information at the sentence level during generation using the conditional variational autoencoder architecture.We demonstrate that our model outperforms the original RNN-based generator, while yielding highly diverse sentences.In addition, our model performs better when the training data is limited.

Stefan Ultes, Paweł Budzianowski, Iñigo Casanueva, Lina M. Rojas-Barahona, Bo-Hsiang Tseng, Yen-Chen Wu, Steve Young, Milica Gašić. Proceedings of the 19th Annual SIGdial Meeting on Discourse and Dialogue. 2018.

Tsung-Hsien Wen, David Vandyke, Nikola Mrkšić, Milica Gašić, Lina M. Rojas-Barahona, Pei-Hao Su, Stefan Ultes, Steve Young. Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers. 2017.

In a dialogue system, the dialogue manager selects one of several system actions and thereby determines the system’s behaviour. Defining all possible system actions in a dialogue system by hand is a tedious work. While efforts have been made to automatically generate such system actions, those approaches are mostly focused on providing functional system behaviour. Adapting the system behaviour to the user becomes a difficult task due to the limited amount of system actions available. We aim to increase the adaptability of a dialogue system by automatically generating variants of system actions. In this work, we introduce an approach to automatically generate action variants for elaborateness and indirectness. Our proposed algorithm extracts RDF triplets from a knowledge base and rates their relevance to the original system action to find suitable content. We show that the results of our algorithm are mostly perceived similarly to human generated elaborateness and indirectness and can be used to adapt a conversation to the current user and situation. We also discuss where the results of our algorithm are still lacking and how this could be improved: Taking into account the conversation topic as well as the culture of the user is likely to have beneficial effect on the user’s perception.

Learning suitable and well-performing dialogue behaviour in statistical spoken dialogue systems has been in the focus of research for many years. While most work which is based on reinforcement learning employs an objective measure like task success for modelling the reward signal, we propose to use a reward based on user satisfaction. We will show in simulated experiments that a live user satisfaction estimation model may be applied resulting in higher estimated satisfaction whilst achieving similar success rates. Moreover, we will show that one satisfaction estimation model which has been trained on one domain may be applied in many other domains which cover a similar task. We will verify our findings by employing the model to one of the domains for learning a policy from real users and compare its performance to policies using the user satisfaction and task success acquired directly from the users as reward.

Paweł Budzianowski, Stefan Ultes, Pei-Hao Su, Nikola Mrkšić, Tsung-Hsien Wen, Iñigo Casanueva, Lina M. Rojas-Barahona, Milica Gašić. Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue. 2017.

Kyusong Lee, Tiancheng Zhao, Yulun Du, Edward Cai, Allen Lu, Eli Pincus, David Traum, Stefan Ultes, Lina M. Rojas-Barahona, Milica Gasic, Steve Young, Maxine Eskenazi. Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue. 2017.

For estimating the Interaction Quality (IQ) in Spoken Dialogue Systems (SDS), the dialogue history is of significant importance.Previous works included this information manually in the form of precomputed temporal features into the classification process.Here, we employ a deep learning architecture based on Long Short-Term Memories (LSTM) to extract this information automatically from the data, thus estimating IQ solely by using current exchange features.We show that it is thereby possible to achieve competitive results as in a scenario where manually optimized temporal features have been included.

Deep reinforcement learning (RL) methods have significant potential for dialogue policy optimisation.However, they suffer from a poor performance in the early stages of learning.This is especially problematic for on-line learning with real users.Two approaches are introduced to tackle this problem.Firstly, to speed up the learning process, two sampleefficient neural networks algorithms: trust region actor-critic with experience replay (TRACER) and episodic natural actorcritic with experience replay (eNACER) are presented.For TRACER, the trust region helps to control the learning step size and avoid catastrophic model changes.For eNACER, the natural gradient identifies the steepest ascent direction in policy space to speed up the convergence.Both models employ off-policy learning with experience replay to improve sampleefficiency.Secondly, to mitigate the cold start issue, a corpus of demonstration data is utilised to pre-train the models prior to on-line reinforcement learning.Combining these two approaches, we demonstrate a practical approach to learning deep RLbased dialogue policies and demonstrate their effectiveness in a task-oriented information seeking domain.

Stefan Ultes, Paweł Budzianowski, Iñigo Casanueva, Nikola Mrkšić, Lina M. Rojas-Barahona, Pei-Hao Su, Tsung-Hsien Wen, Milica Gašić, Steve Young. Proceedings of the 18th Annual SIGdial Meeting on Discourse and Dialogue. 2017.

Spoken dialogue systems allow humans to interact with machines using natural speech. As such, they have many benefits. By using speech as the primary communication medium, a computer interface can facilitate swift, human-like acquisition of information. In recent years, speech interfaces have become ever more popular, as is evident from the rise of personal assistants such as Siri, Google Now, Cortana and Amazon Alexa. Recently, data-driven machine learning methods have been applied to dialogue modelling and the results achieved for limited-domain applications are comparable to or out-perform traditional approaches. Methods based on Gaussian processes are particularly effective as they enable good models to be estimated from limited training data. Furthermore, they provide an explicit estimate of the uncertainty which is particularly useful for reinforcement learning. This article explores the additional steps that are necessary to extend these methods to model multiple dialogue domains. We show that Gaussian process reinforcement learning is an elegant framework that naturally supports a range of methods, including prior knowledge, Bayesian committee machines and multi-agent learning, for facilitating extensible and adaptable dialogue systems.

The ability to compute an accurate reward function is essential for optimising a dialogue policy via reinforcement learning. In real-world applications, using explicit user feedback as the reward signal is often unreliable and costly to collect. This problem can be mitigated if the user's intent is known in advance or data is available to pre-train a task success predictor off-line. In practice neither of these apply for most real world applications. Here we propose an on-line learning framework whereby the dialogue policy is jointly trained alongside the reward model via active learning with a Gaussian process model. This Gaussian process operates on a continuous space dialogue representation generated in an unsupervised fashion using a recurrent neural network encoder-decoder. The experimental results demonstrate that the proposed framework is able to significantly reduce data annotation costs and mitigate noisy user feedback in dialogue policy learning.

This paper presents a deep learning architecture for the semantic decoder component of a Statistical Spoken Dialogue System. In a slot-filling dialogue, the semantic decoder predicts the dialogue act and a set of slot-value pairs from a set of n-best hypotheses returned by the Automatic Speech Recognition. Most current models for spoken language understanding assume (i) word-aligned semantic annotations as in sequence taggers and (ii) delexicalisation, or a mapping of input words to domain-specific concepts using heuristics that try to capture morphological variation but that do not scale to other domains nor to language variation (e.g., morphology, synonyms, paraphrasing ). In this work the semantic decoder is trained using unaligned semantic annotations and it uses distributed semantic representation learning to overcome the limitations of explicit delexicalisation. The proposed architecture uses a convolutional neural network for the sentence representation and a long-short term memory network for the context representation. Results are presented for the publicly available DSTC2 corpus and an In-car corpus which is similar to DSTC2 but has a significantly higher word error rate (WER).

Tsung-Hsien Wen, Milica Gašić, Nikola Mrkšić, Lina M. Rojas-Barahona, Pei-Hao Su, Stefan Ultes, David Vandyke, Steve Young. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016.

In this work, we investigate whether the cultural idiosyncrasies found in humanhuman interaction may be transferred to human-computer interaction.With the aim of designing a culture-sensitive dialogue system, we designed a user study creating a dialogue in a domain that has the potential capacity to reveal cultural differences.The dialogue contains different options for the system output according to cultural differences.We conducted a survey among Germans and Japanese to investigate whether the supposed differences may be applied in human-computer interaction.Our results show that there are indeed differences, but not all results are consistent with the cultural models.

Adapting Spoken Dialogue Systems to the user is supposed to result in more efficient and successful dialogues.In this work, we present an evaluation of a quality-adaptive strategy with a user simulator adapting the dialogue initiative dynamically during the ongoing interaction and show that it outperforms conventional non-adaptive strategies and a random strategy.Furthermore, we indicate a correlation between Interaction Quality and dialogue completion rate, task success rate, and average dialogue length.Finally, we analyze the correlation between task success and interaction quality in more detail identifying the usefulness of interaction quality for modelling the reward of reinforcement learning strategy optimization.

In this paper, we present novel work on speech-based adaptive emotion recognition through addition of speaker-specific information. We propose a two-stage approach of first determining the speaker and then using this information during the emotion recognition process. The proposed technique has been evaluated using five emotional speech databases of different languages using both artificial neural network-based speaker identifier and the ground truth. The addition of speaker-specific information improves the emotion recognition accuracy by up to +10.2%. Moreover, emotion recognition performance scores for all applied databases are improved.

Research trends on SDS evaluation are recently focusing on objective assessment methods. Most existing methods, which derive quality for each systemuser-exchange, do not consider temporal dependencies on the quality of previous exchanges. In this work, we investigate an approach for determining Interaction Quality for human-machine dialogue based on methods modeling the sequential characteristics using HMM modeling. Our approach significantly outperforms conventional approaches by up to 4.5% relative improvement based on Unweighted Average Recall metrics.

We present JaCHMM, a Java implementation of a conditioned Hidden Markov Model (CHMM), which is made available under BSD license. It is based on the open source library “Jahmm” and provides implementations of the Viterbi, Forward-Backward, Baum-Welch and K-Means algorithms, all adapted for the CHMM. Like the Hidden Markov Model (HMM), the CHMM may be applied to a wide range of uni- and multimodal classification problems. The library is intended for academic and scientific purposes but may be also used in commercial systems. As a proof of concept, the JaCHMM library is successfully applied to speech-based emotion recognition outperforming HMM- and SVM-based approaches.

This paper analyzes the human performance of recognizing drunk speakers merely by voice and compares the results with the performance of an automatic statistical classifier. The study is carried out within the Interspeech 2011 Speaker State Challenge [1] employing the Alcohol Language Corpus (ALC) [2]. The 79 subjects yielded an average performance of 55.8% unweighted accuracy on a balanced intoxicated/non-intoxicated sample set. The statistical classifier developed in this study reaches a performance of 66.6% unweighted accuracy on the test set. In comparison, the subject with the highest performance yielded 70.0%. Our classifier is based on 4368 acoustic and prosodic features. Incorporating linguistic features along with feature selection using Information Gain Ratio (IGR) ranking added 0.7% absolute improvement with resulting in a 29% smaller feature space size. Copyright © 2011 ISCA.