Tatsuya Kawahara

Ryuichi Sumida, Koji Inoue, Tatsuya Kawahara

While Retrieval-Augmented Generation (RAG) has shown promise in enhancing long-term conversations, the increasing memory load as conversations progress degrades retrieval accuracy. Drawing on psychological insights, we propose LUFY, a simple yet effective method that focuses on emotionally arousing memories and retains less than 10% of the conversation. In the user experiment, participants interacted with three types of RAG chatbots, each for 2 hours over 4 sessions, marking the most extensive assessment of a chatbot's long-term capabilities to date -- more than four times longer than any existing benchmark. The results demonstrate that prioritizing arousing memories while forgetting the majority of the conversation significantly enhances user experience. This study pushes the frontier of long-term conversations and highlights the importance of forgetting unimportant parts of conversations. Code and Dataset: https://github.com/ryuichi-sumida/LUFY, Hugginface Dataset:https://huggingface.co/datasets/RuiSumida/LUFY

Hayato Futami, Hirofumi Inaguma, Sei Ueno et al.

Connectionist temporal classification (CTC) -based models are attractive in automatic speech recognition (ASR) because of their non-autoregressive nature. To take advantage of text-only data, language model (LM) integration approaches such as rescoring and shallow fusion have been widely used for CTC. However, they lose CTC's non-autoregressive nature because of the need for beam search, which slows down the inference speed. In this study, we propose an error correction method with phone-conditioned masked LM (PC-MLM). In the proposed method, less confident word tokens in a greedy decoded output from CTC are masked. PC-MLM then predicts these masked word tokens given unmasked words and phones supplementally predicted from CTC. We further extend it to Deletable PC-MLM in order to address insertion errors. Since both CTC and PC-MLM are non-autoregressive models, the method enables fast LM integration. Experimental evaluations on the Corpus of Spontaneous Japanese (CSJ) and TED-LIUM2 in domain adaptation setting shows that our proposed method outperformed rescoring and shallow fusion in terms of inference speed, and also in terms of recognition accuracy on CSJ.

Jumon Nozaki, Tatsuya Kawahara, Kenkichi Ishizuka et al.

Conventional automatic speech recognition systems do not produce punctuation marks which are important for the readability of the speech recognition results. They are also needed for subsequent natural language processing tasks such as machine translation. There have been a lot of works on punctuation prediction models that insert punctuation marks into speech recognition results as post-processing. However, these studies do not utilize acoustic information for punctuation prediction and are directly affected by speech recognition errors. In this study, we propose an end-to-end model that takes speech as input and outputs punctuated texts. This model is expected to predict punctuation robustly against speech recognition errors while using acoustic information. We also propose to incorporate an auxiliary loss to train the model using the output of the intermediate layer and unpunctuated texts. Through experiments, we compare the performance of the proposed model to that of a cascaded system. The proposed model achieves higher punctuation prediction accuracy than the cascaded system without sacrificing the speech recognition error rate. It is also demonstrated that the multi-task learning using the intermediate output against the unpunctuated text is effective. Moreover, the proposed model has only about 1/7th of the parameters compared to the cascaded system.

Hayato Futami, Hirofumi Inaguma, Masato Mimura et al.

Connectionist temporal classification (CTC) -based models are attractive because of their fast inference in automatic speech recognition (ASR). Language model (LM) integration approaches such as shallow fusion and rescoring can improve the recognition accuracy of CTC-based ASR by taking advantage of the knowledge in text corpora. However, they significantly slow down the inference of CTC. In this study, we propose to distill the knowledge of BERT for CTC-based ASR, extending our previous study for attention-based ASR. CTC-based ASR learns the knowledge of BERT during training and does not use BERT during testing, which maintains the fast inference of CTC. Different from attention-based models, CTC-based models make frame-level predictions, so they need to be aligned with token-level predictions of BERT for distillation. We propose to obtain alignments by calculating the most plausible CTC paths. Experimental evaluations on the Corpus of Spontaneous Japanese (CSJ) and TED-LIUM2 show that our method improves the performance of CTC-based ASR without the cost of inference speed.

Yahui Fu, Koji Inoue, Chenhui Chu et al.

Recent approaches to empathetic response generation try to incorporate commonsense knowledge or reasoning about the causes of emotions to better understand the user's experiences and feelings. However, these approaches mainly focus on understanding the causalities of context from the user's perspective, ignoring the system's perspective. In this paper, we propose a commonsense-based causality explanation approach for diverse empathetic response generation that considers both the user's perspective (user's desires and reactions) and the system's perspective (system's intentions and reactions). We enhance ChatGPT's ability to reason for the system's perspective by integrating in-context learning with commonsense knowledge. Then, we integrate the commonsense-based causality explanation with both ChatGPT and a T5-based model. Experimental evaluations demonstrate that our method outperforms other comparable methods on both automatic and human evaluations.

Yuka Ko, Sheng Li, Chao-Han Huck Yang et al.

With the strong representational power of large language models (LLMs), generative error correction (GER) for automatic speech recognition (ASR) aims to provide semantic and phonetic refinements to address ASR errors. This work explores how LLM-based GER can enhance and expand the capabilities of Japanese language processing, presenting the first GER benchmark for Japanese ASR with 0.9-2.6k text utterances. We also introduce a new multi-pass augmented generative error correction (MPA GER) by integrating multiple system hypotheses on the input side with corrections from multiple LLMs on the output side and then merging them. To the best of our knowledge, this is the first investigation of the use of LLMs for Japanese GER, which involves second-pass language modeling on the output transcriptions generated by the ASR system (e.g., N-best hypotheses). Our experiments demonstrated performance improvement in the proposed methods of ASR quality and generalization both in SPREDS-U1-ja and CSJ data.

Sota Kobuki, Katie Seaborn, Seiki Tokunaga et al.

Japan faces many challenges related to its aging society, including increasing rates of cognitive decline in the population and a shortage of caregivers. Efforts have begun to explore solutions using artificial intelligence (AI), especially socially embodied intelligent agents and robots that can communicate with people. Yet, there has been little research on the compatibility of these agents with older adults in various everyday situations. To this end, we conducted a user study to evaluate a robot that functions as a facilitator for a group conversation protocol designed to prevent cognitive decline. We modified the robot to use backchannelling, a natural human way of speaking, to increase receptiveness of the robot and enjoyment of the group conversation experience. We conducted a cross-generational study with young adults and older adults. Qualitative analyses indicated that younger adults perceived the backchannelling version of the robot as kinder, more trustworthy, and more acceptable than the non-backchannelling robot. Finally, we found that the robot's backchannelling elicited nonverbal backchanneling in older participants.

Koji Inoue, Mikey Elmers, Yahui Fu et al.

We present a multilingual, continuous backchannel prediction model for Japanese, English, and Chinese, and use it to investigate cross-linguistic timing behavior. The model is Transformer-based and operates at the frame level, jointly trained with auxiliary tasks on approximately 300 hours of dyadic conversations. Across all three languages, the multilingual model matches or surpasses monolingual baselines, indicating that it learns both language-universal cues and language-specific timing patterns. Zero-shot transfer with two-language training remains limited, underscoring substantive cross-lingual differences. Perturbation analyses reveal distinct cue usage: Japanese relies more on short-term linguistic information, whereas English and Chinese are more sensitive to silence duration and prosodic variation; multilingual training encourages shared yet adaptable representations and reduces overreliance on pitch in Chinese. A context-length study further shows that Japanese is relatively robust to shorter contexts, while Chinese benefits markedly from longer contexts. Finally, we integrate the trained model into a real-time processing software, demonstrating CPU-only inference. Together, these findings provide a unified model and empirical evidence for how backchannel timing differs across languages, informing the design of more natural, culturally-aware spoken dialogue systems.

Koji Inoue, Divesh Lala, Keiko Ochi et al.

This paper tackles the challenging task of evaluating socially situated conversational robots and presents a novel objective evaluation approach that relies on multimodal user behaviors. In this study, our main focus is on assessing the human-likeness of the robot as the primary evaluation metric. While previous research often relied on subjective evaluations from users, our approach aims to evaluate the robot's human-likeness based on observable user behaviors indirectly, thus enhancing objectivity and reproducibility. To begin, we created an annotated dataset of human-likeness scores, utilizing user behaviors found in an attentive listening dialogue corpus. We then conducted an analysis to determine the correlation between multimodal user behaviors and human-likeness scores, demonstrating the feasibility of our proposed behavior-based evaluation method.

Jianan Chen, Xiaoxue Gao, Tatsuya Kawahara et al.

Large language models (LLMs) have driven substantial advances in speech language models (SpeechLMs), yielding strong performance in automatic speech recognition (ASR) under high-resource conditions. However, existing benchmarks predominantly focus on high-resource languages, leaving the ASR behavior of SpeechLMs in low-resource languages insufficiently understood. This gap is critical, as practical ASR systems must reliably support low-resource languages and generalize across diverse language families, and it directly hinders the deployment of SpeechLM-based ASR in real-world multilingual scenarios. As a result, it is essential to evaluate SpeechLMs on low-resource languages to ensure their generalizability across different language families. To address this problem, we propose \textbf{LoASR-Bench}, a comprehensive benchmark designed to evaluate \textbf{lo}w-resource \textbf{a}utomatic \textbf{s}peech \textbf{r}ecognition (\textbf{ASR}) of the latest SpeechLMs across diverse language families. LoASR-Bench comprises 25 languages from 9 language families, featuring both Latin and non-Latin scripts, enabling cross-linguistic and cross-script assessment of ASR performance of current SpeechLMs. Experimental results highlight the limitations of the latest SpeechLMs in handling real-world low-resource languages.

Hao Shi, Yuan Gao, Zhaoheng Ni et al.

Serialized output training (SOT) attracts increasing attention due to its convenience and flexibility for multi-speaker automatic speech recognition (ASR). However, it is not easy to train with attention loss only. In this paper, we propose the overlapped encoding separation (EncSep) to fully utilize the benefits of the connectionist temporal classification (CTC) and attention hybrid loss. This additional separator is inserted after the encoder to extract the multi-speaker information with CTC losses. Furthermore, we propose the serialized speech information guidance SOT (GEncSep) to further utilize the separated encodings. The separated streams are concatenated to provide single-speaker information to guide attention during decoding. The experimental results on LibriMix show that the single-speaker encoding can be separated from the overlapped encoding. The CTC loss helps to improve the encoder representation under complex scenarios. GEncSep further improved performance.

Yahui Fu, Chenhui Chu, Tatsuya Kawahara

Recent approaches for empathetic response generation mainly focus on emotional resonance and user understanding, without considering the system's personality. Consistent personality is evident in real human expression and is important for creating trustworthy systems. To address this problem, we propose StyEmp, which aims to stylize the empathetic response generation with a consistent personality. Specifically, it incorporates a multi-grained prefix mechanism designed to capture the intricate relationship between a system's personality and its empathetic expressions. Furthermore, we introduce a personality reinforcement module that leverages contrastive learning to calibrate the generation model, ensuring that responses are both empathetic and reflective of a distinct personality. Automatic and human evaluations on the EMPATHETICDIALOGUES benchmark show that StyEmp outperforms competitive baselines in terms of both empathy and personality expressions.

Zi Haur Pang, Yahui Fu, Yuan Gao et al.

Emotional Validation is a psychotherapy communication technique that involves recognizing, understanding, and explicitly acknowledging another person's feelings and actions, which strengthens alliance and reduces negative affect. To maximize the emotional support provided by validation, it is crucial to deliver it with appropriate timing and frequency. This study investigates validation timing detection from the speech perspective. Leveraging both paralinguistic and emotional information, we propose a paralinguistic- and emotion-aware model for validation timing detection without relying on textual context. Specifically, we first conduct continued self-supervised training and fine-tuning on different HuBERT backbones to obtain (i) a paralinguistics-aware Self-Supervised Learning (SSL) encoder and (ii) a multi-task speech emotion classification encoder. We then fuse these encoders and further fine-tune the combined model on the downstream validation timing detection task. Experimental evaluations on the TUT Emotional Storytelling Corpus (TESC) compare multiple models, fusion mechanisms, and training strategies, and demonstrate that the proposed approach achieves significant improvements over conventional speech baselines. Our results indicate that non-linguistic speech cues, when integrated with affect-related representations, carry sufficient signal to decide when validation should be expressed, offering a speech-first pathway toward more empathetic human-robot interaction.

Hao Shi, Yuan Gao, Xugang Lu et al.

Large Language Models (LLMs) are strong decoders for Serialized Output Training (SOT) in two-talker Automatic Speech Recognition (ASR), yet their performance degrades substantially in challenging conditions such as three-talker mixtures. A key limitation is that current systems inject acoustic evidence only through a projected prefix, which can be lossy and imperfectly aligned with the LLM input space, providing insufficient fine-grained grounding during decoding. Addressing this limitation is crucial for robust multi-talker ASR, especially in three-talker mixtures. This paper improves LLM-based multi-talker ASR by explicitly injecting talker-aware acoustic evidence into the decoder. We first revisit Connectionist Temporal Classification (CTC)-derived prefix prompting and compare three variants with increasing acoustic content. The CTC information is obtained using the serialized CTC proposed in our previous works. While acoustic-enriched prompts outperform the SOT-only baseline, prefix-only conditioning remains inadequate for three-talker mixtures. We therefore propose a lightweight gated residual cross-attention adapter and design a two-stage acoustic adaptation framework based on low-rank updates (LoRA). In Stage 1, we insert gated cross-attention adapters after the self-attention sub-layer to stably inject acoustic embeddings as external memory. In Stage 2, we refine both the cross-attention adapters and the pretrained LLM's self-attention projections using parameter-efficient LoRA, improving robustness for large backbones under limited data; the learned updates are merged into the base weights for inference. Experiments on Libri2Mix/Libri3Mix under clean and noisy conditions show consistent gains, with particularly large improvements in three-talker settings.

Zi Haur Pang, Xiaoxue Gao, Tatsuya Kawahara et al.

Speech emotion recognition (SER) systems can exhibit gender-related performance disparities, but how such bias manifests in multilingual speech LLMs across languages and modalities is unclear. We introduce a novel multilingual, multimodal benchmark built on MELD-ST, spanning English, Japanese, and German, to quantify language-specific SER performance and gender gaps. We find bias is strongly language-dependent, and multimodal fusion does not reliably improve fairness. To address these, we propose ERM-MinMaxGAP, a fairness-informed training objective, which augments empirical risk minimization (ERM) with a proposed adaptive fairness weight mechanism and a novel MinMaxGAP regularizer on the maximum male-female loss gap within each language and modality. Building upon the Qwen2-Audio backbone, our ERM-MinMaxGAP approach improves multilingual SER performance by 5.5% and 5.0% while reducing the overall gender bias gap by 0.1% and 1.4% in the unimodal and multimodal settings, respectively.

Tianyu Zhao, Tatsuya Kawahara

In open-domain dialogue response generation, a dialogue context can be continued with diverse responses, and the dialogue models should capture such one-to-many relations. In this work, we first analyze the training objective of dialogue models from the view of Kullback-Leibler divergence (KLD) and show that the gap between the real world probability distribution and the single-referenced data's probability distribution prevents the model from learning the one-to-many relations efficiently. Then we explore approaches to multi-referenced training in two aspects. Data-wise, we generate diverse pseudo references from a powerful pretrained model to build multi-referenced data that provides a better approximation of the real-world distribution. Model-wise, we propose to equip variational models with an expressive prior, named linear Gaussian model (LGM). Experimental results of automated evaluation and human evaluation show that the methods yield significant improvements over baselines. We will release our code and data in https://github.com/ZHAOTING/dialog-processing.

Tianyu Zhao, Divesh Lala, Tatsuya Kawahara

Automatic dialogue response evaluator has been proposed as an alternative to automated metrics and human evaluation. However, existing automatic evaluators achieve only moderate correlation with human judgement and they are not robust. In this work, we propose to build a reference-free evaluator and exploit the power of semi-supervised training and pretrained (masked) language models. Experimental results demonstrate that the proposed evaluator achieves a strong correlation (> 0.6) with human judgement and generalizes robustly to diverse responses and corpora. We open-source the code and data in https://github.com/ZHAOTING/dialog-processing.

Yahui Fu, Zi Haur Pang, Tatsuya Kawahara

User satisfaction in dialogue systems is inherently subjective. When the same response strategy is applied across users, minority users may assign different satisfaction ratings than majority users due to variations in individual intents and preferences. However, existing alignment methods typically train one-size-fits-all models that aim for broad consensus, often overlooking minority perspectives and user-specific adaptation. We propose a unified framework that models both individual- and group-level preferences for user satisfaction estimation. First, we introduce Chain-of-Personalized-Reasoning (CoPeR) to capture individual preferences through interpretable reasoning chains. Second, we propose an expectation-maximization-based Majority-Minority Preference-Aware Clustering (M2PC) algorithm that discovers distinct user groups in an unsupervised manner to learn group-level preferences. Finally, we integrate these components into a preference-adaptive reinforcement learning framework (PAda-PPO) that jointly optimizes alignment with both individual and group preferences. Experiments on the Emotional Support Conversation dataset demonstrate consistent improvements in user satisfaction estimation, particularly for underrepresented user groups.

Koji Inoue, Bing'er Jiang, Erik Ekstedt et al.

A demonstration of a real-time and continuous turn-taking prediction system is presented. The system is based on a voice activity projection (VAP) model, which directly maps dialogue stereo audio to future voice activities. The VAP model includes contrastive predictive coding (CPC) and self-attention transformers, followed by a cross-attention transformer. We examine the effect of the input context audio length and demonstrate that the proposed system can operate in real-time with CPU settings, with minimal performance degradation.

Koji Inoue, Bing'er Jiang, Erik Ekstedt et al.

This paper investigates the application of voice activity projection (VAP), a predictive turn-taking model for spoken dialogue, on multilingual data, encompassing English, Mandarin, and Japanese. The VAP model continuously predicts the upcoming voice activities of participants in dyadic dialogue, leveraging a cross-attention Transformer to capture the dynamic interplay between participants. The results show that a monolingual VAP model trained on one language does not make good predictions when applied to other languages. However, a multilingual model, trained on all three languages, demonstrates predictive performance on par with monolingual models across all languages. Further analyses show that the multilingual model has learned to discern the language of the input signal. We also analyze the sensitivity to pitch, a prosodic cue that is thought to be important for turn-taking. Finally, we compare two different audio encoders, contrastive predictive coding (CPC) pre-trained on English, with a recent model based on multilingual wav2vec 2.0 (MMS).

Koji Inoue, Divesh Lala, Gabriel Skantze et al.

In human conversations, short backchannel utterances such as "yeah" and "oh" play a crucial role in facilitating smooth and engaging dialogue. These backchannels signal attentiveness and understanding without interrupting the speaker, making their accurate prediction essential for creating more natural conversational agents. This paper proposes a novel method for real-time, continuous backchannel prediction using a fine-tuned Voice Activity Projection (VAP) model. While existing approaches have relied on turn-based or artificially balanced datasets, our approach predicts both the timing and type of backchannels in a continuous and frame-wise manner on unbalanced, real-world datasets. We first pre-train the VAP model on a general dialogue corpus to capture conversational dynamics and then fine-tune it on a specialized dataset focused on backchannel behavior. Experimental results demonstrate that our model outperforms baseline methods in both timing and type prediction tasks, achieving robust performance in real-time environments. This research offers a promising step toward more responsive and human-like dialogue systems, with implications for interactive spoken dialogue applications such as virtual assistants and robots.

Hao Shi, Tatsuya Kawahara

Adapting an automatic speech recognition (ASR) system to unseen noise environments is crucial. Integrating adapters into neural networks has emerged as a potent technique for transfer learning. This study thoroughly investigates adapter-based ASR adaptation in noisy environments. We conducted experiments using the CHiME--4 dataset. The results show that inserting the adapter in the shallow layer yields superior effectiveness, and there is no significant difference between adapting solely within the shallow layer and adapting across all layers. The simulated data helps the system to improve its performance under real noise conditions. Nonetheless, when the amount of data is the same, the real data is more effective than the simulated data. Multi-condition training is still useful for adapter training. Furthermore, integrating adapters into speech enhancement-based ASR systems yields substantial improvements.

Zi Haur Pang, Yahui Fu, Divesh Lala et al.

This paper introduces the human-like embodied AI interviewer which integrates android robots equipped with advanced conversational capabilities, including attentive listening, conversational repairs, and user fluency adaptation. Moreover, it can analyze and present results post-interview. We conducted a real-world case study at SIGDIAL 2024 with 42 participants, of whom 69% reported positive experiences. This study demonstrated the system's effectiveness in conducting interviews just like a human and marked the first employment of such a system at an international conference. The demonstration video is available at https://youtu.be/jCuw9g99KuE.

Koji Inoue, Divesh Lala, Mikey Elmers et al.

Handling multi-party dialogues represents a significant step for advancing spoken dialogue systems, necessitating the development of tasks specific to multi-party interactions. To address this challenge, we are constructing a multi-modal multi-party dialogue corpus of triadic (three-participant) discussions. This paper focuses on the task of addressee recognition, identifying who is being addressed to take the next turn, a critical component unique to multi-party dialogue systems. A subset of the corpus was annotated with addressee information, revealing that explicit addressees are indicated in approximately 20% of conversational turns. To evaluate the task's complexity, we benchmarked the performance of a large language model (GPT-4o) on addressee recognition. The results showed that GPT-4o achieved an accuracy only marginally above chance, underscoring the challenges of addressee recognition in multi-party dialogue. These findings highlight the need for further research to enhance the capabilities of large language models in understanding and navigating the intricacies of multi-party conversational dynamics.

Zi Haur Pang, Yahui Fu, Divesh Lala et al.

In the realm of human-AI dialogue, the facilitation of empathetic responses is important. Validation is one of the key communication techniques in psychology, which entails recognizing, understanding, and acknowledging others' emotional states, thoughts, and actions. This study introduces the first framework designed to engender empathetic dialogue with validating responses. Our approach incorporates a tripartite module system: 1) validation timing detection, 2) users' emotional state identification, and 3) validating response generation. Utilizing Japanese EmpatheticDialogues dataset - a textual-based dialogue dataset consisting of 8 emotional categories from Plutchik's wheel of emotions - the Task Adaptive Pre-Training (TAPT) BERT-based model outperforms both random baseline and the ChatGPT performance, in term of F1-score, in all modules. Further validation of our model's efficacy is confirmed in its application to the TUT Emotional Storytelling Corpus (TESC), a speech-based dialogue dataset, by surpassing both random baseline and the ChatGPT. This consistent performance across both textual and speech-based dialogues underscores the effectiveness of our framework in fostering empathetic human-AI communication.

Yahui Fu, Haiyue Song, Tianyu Zhao et al.

Personality recognition is useful for enhancing robots' ability to tailor user-adaptive responses, thus fostering rich human-robot interactions. One of the challenges in this task is a limited number of speakers in existing dialogue corpora, which hampers the development of robust, speaker-independent personality recognition models. Additionally, accurately modeling both the interdependencies among interlocutors and the intra-dependencies within the speaker in dialogues remains a significant issue. To address the first challenge, we introduce personality trait interpolation for speaker data augmentation. For the second, we propose heterogeneous conversational graph networks to independently capture both contextual influences and inherent personality traits. Evaluations on the RealPersonaChat corpus demonstrate our method's significant improvements over existing baselines.

Zi Haur Pang, Yahui Fu, Divesh Lala et al.

We investigate the impact of robot appearance on users' spoken behavior during real-world interactions by comparing a human-like android, ERICA, with a less anthropomorphic humanoid, TELECO. Analyzing data from 42 participants at SIGDIAL 2024, we extracted linguistic features such as disfluencies and syntactic complexity from conversation transcripts. The results showed moderate effect sizes, suggesting that participants produced fewer disfluencies and employed more complex syntax when interacting with ERICA. Further analysis involving training classification models like Naïve Bayes, which achieved an F1-score of 71.60\%, and conducting feature importance analysis, highlighted the significant role of disfluencies and syntactic complexity in interactions with robots of varying human-like appearances. Discussing these findings within the frameworks of cognitive load and Communication Accommodation Theory, we conclude that designing robots to elicit more structured and fluent user speech can enhance their communicative alignment with humans.

Koji Inoue, Divesh Lala, Keiko Ochi et al.

Establishing evaluation schemes for spoken dialogue systems is important, but it can also be challenging. While subjective evaluations are commonly used in user experiments, objective evaluations are necessary for research comparison and reproducibility. To address this issue, we propose a framework for indirectly but objectively evaluating systems based on users' behaviors. In this paper, to this end, we investigate the relationship between user behaviors and subjective evaluation scores in social dialogue tasks: attentive listening, job interview, and first-meeting conversation. The results reveal that in dialogue tasks where user utterances are primary, such as attentive listening and job interview, indicators like the number of utterances and words play a significant role in evaluation. Observing disfluency also can indicate the effectiveness of formal tasks, such as job interview. On the other hand, in dialogue tasks with high interactivity, such as first-meeting conversation, behaviors related to turn-taking, like average switch pause length, become more important. These findings suggest that selecting appropriate user behaviors can provide valuable insights for objective evaluation in each social dialogue task.

Koji Inoue, Yuki Okafuji, Jun Baba et al.

Turn-taking is a crucial aspect of human-robot interaction, directly influencing conversational fluidity and user engagement. While previous research has explored turn-taking models in controlled environments, their robustness in real-world settings remains underexplored. In this study, we propose a noise-robust voice activity projection (VAP) model, based on a Transformer architecture, to enhance real-time turn-taking in dialogue robots. To evaluate the effectiveness of the proposed system, we conducted a field experiment in a shopping mall, comparing the VAP system with a conventional cloud-based speech recognition system. Our analysis covered both subjective user evaluations and objective behavioral analysis. The results showed that the proposed system significantly reduced response latency, leading to a more natural conversation where both the robot and users responded faster. The subjective evaluations suggested that faster responses contribute to a better interaction experience.

Koji Inoue, Mikey Elmers, Divesh Lala et al.

Laughter serves as a multifaceted communicative signal in human interaction, yet its identification within dialogue presents a significant challenge for conversational AI systems. This study addresses this challenge by annotating laughable contexts in Japanese spontaneous text conversation data and developing a taxonomy to classify the underlying reasons for such contexts. Initially, multiple annotators manually labeled laughable contexts using a binary decision (laughable or non-laughable). Subsequently, an LLM was used to generate explanations for the binary annotations of laughable contexts, which were then categorized into a taxonomy comprising ten categories, including "Empathy and Affinity" and "Humor and Surprise," highlighting the diverse range of laughter-inducing scenarios. The study also evaluated GPT-4o's performance in recognizing the majority labels of laughable contexts, achieving an F1 score of 43.14%. These findings contribute to the advancement of conversational AI by establishing a foundation for more nuanced recognition and generation of laughter, ultimately fostering more natural and engaging human-AI interactions.

Haruki Kawai, Divesh Lala, Koji Inoue et al.

The handling of communication breakdowns and loss of engagement is an important aspect of spoken dialogue systems, particularly for chatting systems such as attentive listening, where the user is mostly speaking. We presume that a human is best equipped to handle this task and rescue the flow of conversation. To this end, we propose a semi-autonomous system, where a remote operator can take control of an autonomous attentive listening system in real-time. In order to make human intervention easy and consistent, we introduce automatic detection of low interest and engagement to provide explicit takeover prompts to the remote operator. We implement this semi-autonomous system which detects takeover points for the operator and compare it to fully tele-operated and fully autonomous attentive listening systems. We find that the semi-autonomous system is generally perceived more positively than the autonomous system. The results suggest that identifying points of conversation when the user starts to lose interest may help us improve a fully autonomous dialogue system.

Weihua Zheng, Zhengyuan Liu, Tanmoy Chakraborty et al.

Large language models (LLMs) are now used worldwide, yet their multimodal understanding and reasoning often degrade outside Western, high-resource settings. We propose MMA-ASIA, a comprehensive framework to evaluate LLMs' cultural awareness with a focus on Asian contexts. MMA-ASIA centers on a human-curated, multilingual, and multimodally aligned multiple-choice benchmark covering 8 Asian countries and 10 languages, comprising 27,000 questions; over 79 percent require multi-step reasoning grounded in cultural context, moving beyond simple memorization. To our knowledge, this is the first dataset aligned at the input level across three modalities: text, image (visual question answering), and speech. This enables direct tests of cross-modal transfer. Building on this benchmark, we propose a five-dimensional evaluation protocol that measures: (i) cultural-awareness disparities across countries, (ii) cross-lingual consistency, (iii) cross-modal consistency, (iv) cultural knowledge generalization, and (v) grounding validity. To ensure rigorous assessment, a Cultural Awareness Grounding Validation Module detects "shortcut learning" by checking whether the requisite cultural knowledge supports correct answers. Finally, through comparative model analysis, attention tracing, and an innovative Vision-ablated Prefix Replay (VPR) method, we probe why models diverge across languages and modalities, offering actionable insights for building culturally reliable multimodal LLMs.

Mikey Elmers, Koji Inoue, Divesh Lala et al.

Turn-taking is a fundamental component of spoken dialogue, however conventional studies mostly involve dyadic settings. This work focuses on applying voice activity projection (VAP) to predict upcoming turn-taking in triadic multi-party scenarios. The goal of VAP models is to predict the future voice activity for each speaker utilizing only acoustic data. This is the first study to extend VAP into triadic conversation. We trained multiple models on a Japanese triadic dataset where participants discussed a variety of topics. We found that the VAP trained on triadic conversation outperformed the baseline for all models but that the type of conversation affected the accuracy. This study establishes that VAP can be used for turn-taking in triadic dialogue scenarios. Future work will incorporate this triadic VAP turn-taking model into spoken dialogue systems.

Koji Inoue, Mikey Elmers, Yahui Fu et al.

Turn-taking prediction models are essential components in spoken dialogue systems and conversational robots. Recent approaches leverage transformer-based architectures to predict speech activity continuously and in real-time. In this study, we propose a novel model that enables turn-taking prediction to be dynamically controlled via textual prompts. This approach allows intuitive and explicit control through instructions such as "faster" or "calmer" adapting dynamically to conversational partners and contexts. The proposed model builds upon a transformer-based voice activity projection (VAP) model, incorporating textual prompt embeddings into both channel-wise transformers and a cross-channel transformer. We evaluated the feasibility of our approach using over 950 hours of human-human spoken dialogue data. Since textual prompt data for the proposed approach was not available in existing datasets, we utilized a large language model (LLM) to generate synthetic prompt sentences. Experimental results demonstrated that the proposed model improved prediction accuracy and effectively varied turn-taking timing behaviors according to the textual prompts.

Hao Shi, Kazuki Shimada, Masato Hirano et al.

Diffusion-based generative speech enhancement (SE) has recently received attention, but reverse diffusion remains time-consuming. One solution is to initialize the reverse diffusion process with enhanced features estimated by a predictive SE system. However, the pipeline structure currently does not consider for a combined use of generative and predictive decoders. The predictive decoder allows us to use the further complementarity between predictive and diffusion-based generative SE. In this paper, we propose a unified system that use jointly generative and predictive decoders across two levels. The encoder encodes both generative and predictive information at the shared encoding level. At the decoded feature level, we fuse the two decoded features by generative and predictive decoders. Specifically, the two SE modules are fused in the initial and final diffusion steps: the initial fusion initializes the diffusion process with the predictive SE to improve convergence, and the final fusion combines the two complementary SE outputs to enhance SE performance. Experiments conducted on the Voice-Bank dataset demonstrate that incorporating predictive information leads to faster decoding and higher PESQ scores compared with other score-based diffusion SE (StoRM and SGMSE+).

Hayato Futami, Hirofumi Inaguma, Masato Mimura et al.

In automatic speech recognition (ASR) rescoring, the hypothesis with the fewest errors should be selected from the n-best list using a language model (LM). However, LMs are usually trained to maximize the likelihood of correct word sequences, not to detect ASR errors. We propose an ASR rescoring method for directly detecting errors with ELECTRA, which is originally a pre-training method for NLP tasks. ELECTRA is pre-trained to predict whether each word is replaced by BERT or not, which can simulate ASR error detection on large text corpora. To make this pre-training closer to ASR error detection, we further propose an extended version of ELECTRA called phone-attentive ELECTRA (P-ELECTRA). In the pre-training of P-ELECTRA, each word is replaced by a phone-to-word conversion model, which leverages phone information to generate acoustically similar words. Since our rescoring method is optimized for detecting errors, it can also be used for word-level confidence estimation. Experimental evaluations on the Librispeech and TED-LIUM2 corpora show that our rescoring method with ELECTRA is competitive with conventional rescoring methods with faster inference. ELECTRA also performs better in confidence estimation than BERT because it can learn to detect inappropriate words not only in fine-tuning but also in pre-training.

Hirofumi Inaguma, Yosuke Higuchi, Kevin Duh et al.

This article describes an efficient end-to-end speech translation (E2E-ST) framework based on non-autoregressive (NAR) models. End-to-end speech translation models have several advantages over traditional cascade systems such as inference latency reduction. However, conventional AR decoding methods are not fast enough because each token is generated incrementally. NAR models, however, can accelerate the decoding speed by generating multiple tokens in parallel on the basis of the token-wise conditional independence assumption. We propose a unified NAR E2E-ST framework called Orthros, which has an NAR decoder and an auxiliary shallow AR decoder on top of the shared encoder. The auxiliary shallow AR decoder selects the best hypothesis by rescoring multiple candidates generated from the NAR decoder in parallel (parallel AR rescoring). We adopt conditional masked language model (CMLM) and a connectionist temporal classification (CTC)-based model as NAR decoders for Orthros, referred to as Orthros-CMLM and Orthros-CTC, respectively. We also propose two training methods to enhance the CMLM decoder. Experimental evaluations on three benchmark datasets with six language directions demonstrated that Orthros achieved large improvements in translation quality with a very small overhead compared with the baseline NAR model. Moreover, the Conformer encoder architecture enabled large quality improvements, especially for CTC-based models. Orthros-CTC with the Conformer encoder increased decoding speed by 3.63x on CPU with translation quality comparable to that of an AR model.

Hirofumi Inaguma, Tatsuya Kawahara

In this work, we propose novel decoding algorithms to enable streaming automatic speech recognition (ASR) on unsegmented long-form recordings without voice activity detection (VAD), based on monotonic chunkwise attention (MoChA) with an auxiliary connectionist temporal classification (CTC) objective. We propose a block-synchronous beam search decoding to take advantage of efficient batched output-synchronous and low-latency input-synchronous searches. We also propose a VAD-free inference algorithm that leverages CTC probabilities to determine a suitable timing to reset the model states to tackle the vulnerability to long-form data. Experimental evaluations demonstrate that the block-synchronous decoding achieves comparable accuracy to the label-synchronous one. Moreover, the VAD-free inference can recognize long-form speech robustly for up to a few hours.

Hirofumi Inaguma, Tatsuya Kawahara

While attention-based encoder-decoder (AED) models have been successfully extended to the online variants for streaming automatic speech recognition (ASR), such as monotonic chunkwise attention (MoChA), the models still have a large label emission latency because of the unconstrained end-to-end training objective. Previous works tackled this problem by leveraging alignment information to control the timing to emit tokens during training. In this work, we propose a simple alignment-free regularization method, StableEmit, to encourage MoChA to emit tokens earlier. StableEmit discounts the selection probabilities in hard monotonic attention for token boundary detection by a constant factor and regularizes them to recover the total attention mass during training. As a result, the scale of the selection probabilities is increased, and the values can reach a threshold for token emission earlier, leading to a reduction of emission latency and deletion errors. Moreover, StableEmit can be combined with methods that constraint alignments to further improve the accuracy and latency. Experimental evaluations with LSTM and Conformer encoders demonstrate that StableEmit significantly reduces the recognition errors and the emission latency simultaneously. We also show that the use of alignment information is complementary in both metrics.

Etsuko Ishii, Genta Indra Winata, Samuel Cahyawijaya et al.

Over the past year, research in various domains, including Natural Language Processing (NLP), has been accelerated to fight against the COVID-19 pandemic, yet such research has just started on dialogue systems. In this paper, we introduce an end-to-end dialogue system which aims to ease the isolation of people under self-quarantine. We conduct a control simulation experiment to assess the effects of the user interface, a web-based virtual agent called Nora vs. the android ERICA via a video call. The experimental results show that the android offers a more valuable user experience by giving the impression of being more empathetic and engaging in the conversation due to its nonverbal information, such as facial expressions and body gestures.

Tatsuya Kawahara, Koji Inoue, Divesh Lala

Following the success of spoken dialogue systems (SDS) in smartphone assistants and smart speakers, a number of communicative robots are developed and commercialized. Compared with the conventional SDSs designed as a human-machine interface, interaction with robots is expected to be in a closer manner to talking to a human because of the anthropomorphism and physical presence. The goal or task of dialogue may not be information retrieval, but the conversation itself. In order to realize human-level "long and deep" conversation, we have developed an intelligent conversational android ERICA. We set up several social interaction tasks for ERICA, including attentive listening, job interview, and speed dating. To allow for spontaneous, incremental multiple utterances, a robust turn-taking model is implemented based on TRP (transition-relevance place) prediction, and a variety of backchannels are generated based on time frame-wise prediction instead of IPU-based prediction. We have realized an open-domain attentive listening system with partial repeats and elaborating questions on focus words as well as assessment responses. It has been evaluated with 40 senior people, engaged in conversation of 5-7 minutes without a conversation breakdown. It was also compared against the WOZ setting. We have also realized a job interview system with a set of base questions followed by dynamic generation of elaborating questions. It has also been evaluated with student subjects, showing promising results.

Hirofumi Inaguma, Tatsuya Kawahara, Shinji Watanabe

A conventional approach to improving the performance of end-to-end speech translation (E2E-ST) models is to leverage the source transcription via pre-training and joint training with automatic speech recognition (ASR) and neural machine translation (NMT) tasks. However, since the input modalities are different, it is difficult to leverage source language text successfully. In this work, we focus on sequence-level knowledge distillation (SeqKD) from external text-based NMT models. To leverage the full potential of the source language information, we propose backward SeqKD, SeqKD from a target-to-source backward NMT model. To this end, we train a bilingual E2E-ST model to predict paraphrased transcriptions as an auxiliary task with a single decoder. The paraphrases are generated from the translations in bitext via back-translation. We further propose bidirectional SeqKD in which SeqKD from both forward and backward NMT models is combined. Experimental evaluations on both autoregressive and non-autoregressive models show that SeqKD in each direction consistently improves the translation performance, and the effectiveness is complementary regardless of the model capacity.

Hirofumi Inaguma, Tatsuya Kawahara

This article describes an efficient training method for online streaming attention-based encoder-decoder (AED) automatic speech recognition (ASR) systems. AED models have achieved competitive performance in offline scenarios by jointly optimizing all components. They have recently been extended to an online streaming framework via models such as monotonic chunkwise attention (MoChA). However, the elaborate attention calculation process is not robust for long-form speech utterances. Moreover, the sequence-level training objective and time-restricted streaming encoder cause a nonnegligible delay in token emission during inference. To address these problems, we propose CTC synchronous training (CTC-ST), in which CTC alignments are leveraged as a reference for token boundaries to enable a MoChA model to learn optimal monotonic input-output alignments. We formulate a purely end-to-end training objective to synchronize the boundaries of MoChA to those of CTC. The CTC model shares an encoder with the MoChA model to enhance the encoder representation. Moreover, the proposed method provides alignment information learned in the CTC branch to the attention-based decoder. Therefore, CTC-ST can be regarded as self-distillation of alignment knowledge from CTC to MoChA. Experimental evaluations on a variety of benchmark datasets show that the proposed method significantly reduces recognition errors and emission latency simultaneously. The robustness to long-form and noisy speech is also demonstrated. We compare CTC-ST with several methods that distill alignment knowledge from a hybrid ASR system and show that the CTC-ST can achieve a comparable tradeoff of accuracy and latency without relying on external alignment information. The best MoChA system shows recognition accuracy comparable to that of RNN-transducer (RNN-T) while achieving lower emission latency.

Hirofumi Inaguma, Yosuke Higuchi, Kevin Duh et al.

Fast inference speed is an important goal towards real-world deployment of speech translation (ST) systems. End-to-end (E2E) models based on the encoder-decoder architecture are more suitable for this goal than traditional cascaded systems, but their effectiveness regarding decoding speed has not been explored so far. Inspired by recent progress in non-autoregressive (NAR) methods in text-based translation, which generates target tokens in parallel by eliminating conditional dependencies, we study the problem of NAR decoding for E2E-ST. We propose a novel NAR E2E-ST framework, Orthros, in which both NAR and autoregressive (AR) decoders are jointly trained on the shared speech encoder. The latter is used for selecting better translation among various length candidates generated from the former, which dramatically improves the effectiveness of a large length beam with negligible overhead. We further investigate effective length prediction methods from speech inputs and the impact of vocabulary sizes. Experiments on four benchmarks show the effectiveness of the proposed method in improving inference speed while maintaining competitive translation quality compared to state-of-the-art AR E2E-ST systems.

Hayato Futami, Hirofumi Inaguma, Sei Ueno et al.

Attention-based sequence-to-sequence (seq2seq) models have achieved promising results in automatic speech recognition (ASR). However, as these models decode in a left-to-right way, they do not have access to context on the right. We leverage both left and right context by applying BERT as an external language model to seq2seq ASR through knowledge distillation. In our proposed method, BERT generates soft labels to guide the training of seq2seq ASR. Furthermore, we leverage context beyond the current utterance as input to BERT. Experimental evaluations show that our method significantly improves the ASR performance from the seq2seq baseline on the Corpus of Spontaneous Japanese (CSJ). Knowledge distillation from BERT outperforms that from a transformer LM that only looks at left context. We also show the effectiveness of leveraging context beyond the current utterance. Our method outperforms other LM application approaches such as n-best rescoring and shallow fusion, while it does not require extra inference cost.

Hirofumi Inaguma, Masato Mimura, Tatsuya Kawahara

We investigate a monotonic multihead attention (MMA) by extending hard monotonic attention to Transformer-based automatic speech recognition (ASR) for online streaming applications. For streaming inference, all monotonic attention (MA) heads should learn proper alignments because the next token is not generated until all heads detect the corresponding token boundaries. However, we found not all MA heads learn alignments with a naïve implementation. To encourage every head to learn alignments properly, we propose HeadDrop regularization by masking out a part of heads stochastically during training. Furthermore, we propose to prune redundant heads to improve consensus among heads for boundary detection and prevent delayed token generation caused by such heads. Chunkwise attention on each MA head is extended to the multihead counterpart. Finally, we propose head-synchronous beam search decoding to guarantee stable streaming inference.

Kohei Matsuura, Masato Mimura, Shinsuke Sakai et al.

It is important to transcribe and archive speech data of endangered languages for preserving heritages of verbal culture and automatic speech recognition (ASR) is a powerful tool to facilitate this process. However, since endangered languages do not generally have large corpora with many speakers, the performance of ASR models trained on them are considerably poor in general. Nevertheless, we are often left with a lot of recordings of spontaneous speech data that have to be transcribed. In this work, for mitigating this speaker sparsity problem, we propose to convert the whole training speech data and make it sound like the test speaker in order to develop a highly accurate ASR system for this speaker. For this purpose, we utilize a CycleGAN-based non-parallel voice conversion technology to forge a labeled training data that is close to the test speaker's speech. We evaluated this speaker adaptation approach on two low-resource corpora, namely, Ainu and Mboshi. We obtained 35-60% relative improvement in phone error rate on the Ainu corpus, and 40% relative improvement was attained on the Mboshi corpus. This approach outperformed two conventional methods namely unsupervised adaptation and multilingual training with these two corpora.

Hirofumi Inaguma, Masato Mimura, Tatsuya Kawahara

Monotonic chunkwise attention (MoChA) has been studied for the online streaming automatic speech recognition (ASR) based on a sequence-to-sequence framework. In contrast to connectionist temporal classification (CTC), backward probabilities cannot be leveraged in the alignment marginalization process during training due to left-to-right dependency in the decoder. This results in the error propagation of alignments to subsequent token generation. To address this problem, we propose CTC-synchronous training (CTC-ST), in which MoChA uses CTC alignments to learn optimal monotonic alignments. Reference CTC alignments are extracted from a CTC branch sharing the same encoder with the decoder. The entire model is jointly optimized so that the expected boundaries from MoChA are synchronized with the alignments. Experimental evaluations of the TEDLIUM release-2 and Librispeech corpora show that the proposed method significantly improves recognition, especially for long utterances. We also show that CTC-ST can bring out the full potential of SpecAugment for MoChA.

Viet-Trung Dang, Tianyu Zhao, Sei Ueno et al.

Spoken language understanding, which extracts intents and/or semantic concepts in utterances, is conventionally formulated as a post-processing of automatic speech recognition. It is usually trained with oracle transcripts, but needs to deal with errors by ASR. Moreover, there are acoustic features which are related with intents but not represented with the transcripts. In this paper, we present an end-to-end model which directly converts speech into dialog acts without the deterministic transcription process. In the proposed model, the dialog act recognition network is conjunct with an acoustic-to-word ASR model at its latent layer before the softmax layer, which provides a distributed representation of word-level ASR decoding information. Then, the entire network is fine-tuned in an end-to-end manner. This allows for stable training as well as robustness against ASR errors. The model is further extended to conduct DA segmentation jointly. Evaluations with the Switchboard corpus demonstrate that the proposed method significantly improves dialog act recognition accuracy from the conventional pipeline framework.

Kohei Matsuura, Sei Ueno, Masato Mimura et al.

Ainu is an unwritten language that has been spoken by Ainu people who are one of the ethnic groups in Japan. It is recognized as critically endangered by UNESCO and archiving and documentation of its language heritage is of paramount importance. Although a considerable amount of voice recordings of Ainu folklore has been produced and accumulated to save their culture, only a quite limited parts of them are transcribed so far. Thus, we started a project of automatic speech recognition (ASR) for the Ainu language in order to contribute to the development of annotated language archives. In this paper, we report speech corpus development and the structure and performance of end-to-end ASR for Ainu. We investigated four modeling units (phone, syllable, word piece, and word) and found that the syllable-based model performed best in terms of both word and phone recognition accuracy, which were about 60% and over 85% respectively in speaker-open condition. Furthermore, word and phone accuracy of 80% and 90% has been achieved in a speaker-closed setting. We also found out that a multilingual ASR training with additional speech corpora of English and Japanese further improves the speaker-open test accuracy.