The rapid growth of mobile networks and the need for resilient infrastructure pose challenges to the scalability and fault tolerance of the mobile core network (CN). This study presents Premature-Cloud5GC (P-Cloud5GC), a novel 5G CN architecture designed for public cloud environments. Extending the serverless control plane design of Cloud5GC, P-Cloud5GC incorporates user plane functionality and web-application-based technologies to improve scalability, session management, and fault isolation. Using serverless computing, distributed messaging, and cloud-native databases, P-Cloud5GC deploys a micro Mobile Core per UE procedure, preventing large-scale failures and enhancing redundancy. Performance evaluation during PDU session setup shows that P-Cloud5GC can handle a sufficient volume of UE connection requests within practical time constraints. This work demonstrates the viability of cloud-based mobile core networks, offering a scalable, fault-tolerant, and costeffective alternative to traditional 3GPP-based 5G cores.

Zero-Touch is a broad concept, but it is understood as a means to limit human exposure to the data-plane states, and therefore reduces the potential for error. By combining the various layers from the administration level down to the switching fabric, introducing a new top to bottom architecture. This paper presents our operational research project HolistIX, a full-stack management solution enabling zero-touch networking within IXP infrastructure networks. HolistIX achieves this by first bringing diagramming to the forefront of topology planning and automated network configuration generation. This is then followed by verifying the proposed network configuration through emulation before deployment to the production network. By utilizing Umbrella SDN switching fabric, HolistIX enables full-stack management, with minimal human interaction required. We also briefly present the results of deploying our HolistIX approach in two IXPs, one in France (TouIX) and the other in Japan (DIX-IE).

This paper proposes a referential vowel duration ratio for a pair of vowels in consecutive syllables and a weighted mean of the referential vowel duration ratios on a logarithmic scale as a feature for automatic assessment of second-language (L2) word prosody. In addition to contours of fundamental frequency (F0) and energy i.e. suprasegmental information of speech, segmental duration of syllables or phonemes provides important information for assessing L2 prosody. For L2 learners, the first step of learning prosody is to put accents or stresses on appropriate syllables in words. A syllable with a stress should be produced longer and one without a stress should be produced shorter. To achieve this, we propose taking a duration ratio for every pair of consecutive vowels in reference to duration contrast of the same vowel pair produced by native speakers. Furthermore, we propose a weighted mean of the ratios on a logarithmic scale in consideration of local importance within a word. In evaluation with English word utterances produced by Japanese learners, the introduction of the weighted mean of the ratio significantly improved the correlation coefficient with subjective scores.

Recent studies have introduced end-to-end TTS, which integrates the production of context and acoustic features in statistical parametric speech synthesis. As a result, a single neural network replaced laborious feature engineering with automated feature learning. However, little is known about what types of context information end-to-end TTS extracts from text input before synthesizing speech, and the previous knowledge about context features is barely utilized. In this work, we first point out the model similarity between end-to-end TTS and parametric TTS. Based on the similarity, we evaluate the quality of encoder outputs from an end-to-end TTS system against eight criteria that are derived from a standard set of context information used in parametric TTS. We conduct experiments using an evaluation procedure that has been newly developed in the machine learning literature for quantitative analysis of neural representations, while adapting it to the TTS domain. Experimental results show that the encoder outputs reflect both linguistic and phonetic contexts, such as vowel reduction at phoneme level, lexical stress at syllable level, and part-of-speech at word level, possibly due to the joint optimization of context and acoustic features.

We propose an efficient speech-recognition error-correction interface for Japanese text entry on smart-watches. Although the accuracy of automatic speech recognition (ASR) has significantly improved, an interface for text modification is still essential. Considering the strict limitation of a narrow display area and practical demand of text modification, the proposed interface arranges the N-best results of ASR and a list of morphemes consisting of the 1-best result to enable quick access to any word to be modified. Specifically, multiple screens of the N-best results are switched by horizontal flicks, and another extended screen listing a morpheme sequence of the 1-best result is scrolled by vertical flicks. The proposed interface was compared with a software keyboard and a speech-input-enabled input method editor (IME), which was a simple combination of speech input and software keyboard. The proposed interface outperformed the other two interfaces in terms of time required to complete specified sentences, subjective score using system usability scale (SUS), and perceived workload quantified using the NASA Task Load Index (NASA-TLX).

Quantitative analysis of gazes between a speaker and listeners was conducted from the viewpoint of mutual activities in floor apportionment, with the assumption that mutual gaze plays an important role in coordinating speech interaction. We conducted correlation analyses of the speaker's and listener's gazes in a three-party conversation, comparing native language (L1) and second language (L2) interaction in two types (free-flowing and goal-orient- ed). The analyses showed significant correlations between gazes from the current to the next speaker and those from the next to the current speaker during utterances preceding a speaker change in L1 conversation, suggesting that the participants were coordinating their speech turns with mutual gazes. In L2 conversation, however, such a correlation was found only in the goal-oriented type, suggesting that linguistic proficiency may affect the floor-apportionment function of mutual gazes, possibly because of the cognitive load of understanding/producing utterances.

We propose BubbleFlick, an effective interface for Japanese text entry on smartwatches. While various ideas have been proposed to provide easy and fast text entry for the Latin alphabet, Japanese text entry has additional challenges such as having more than fifty syllabary characters, or kana, to enter and the subsequent kana-kanji conversion, which translates a sequence of the syllabary characters into a standard expression with a mixture of kanji and kana characters. This paper focuses on interfaces for entry of kana syllabary characters. We designed and prototyped three interfaces: 1) Japanese kana syllabary keyboard, 2) Dial&Flick interface, and 3) DualBelts interface. Through a comparative pilot study of the prototypes, we refined the most promising Dial&Flick interface into BubbleFlick. BubbleFlick provides the widest possible area for easy flick operations while also leaving an area for editing text. We conducted a 30-day consecutive user study on BubbleFlick in comparison with Google's latest Japanese text-entry method based on a numeric keypad. After thirty days, BubbleFlick showed a text-entry speed of over 35 characters per minute, which was comparable to Google's numeric-keypad-based method for novice participants. Through the user study, BubbleFlick showed a lower error rate and gave us informative hints for further improvement.

Recursive autoencoders (RAEs) for compositionality of a vector space model were applied to utterance intent classification of a smartphone-based Japanese-language spoken dialogue system.Though the RAEs express a nonlinear operation on the vectors of child nodes, the operation is considered to be different intrinsically depending on types of child nodes.To relax the difference, a data-driven untying of autoencoders (AEs) is proposed.The experimental result of the utterance intent classification showed an improved accuracy with the proposed method compared with the basic tied RAE and untied RAE based on a manual rule.

The importance of conversation in second languages during international collaboration continues to increase, as does the risk of miscommunication caused by the differences in the linguistic proficiency of participants. This study provides a basis for monitoring each participant's status by studying the effects of linguistic proficiency on such communicative activities as utterances and gazes in second language conversations. In such conversations, we found that gaze produces different functions from those in native language conversations and that partners' linguistic proficiency may also affect gaze functions.

In this article we present a technique for mining transliteration pairs using a set of simple features derived from a many-to-many bilingual forced-alignment at the grapheme level to classify candidate transliteration word pairs as correct transliterations or not. We use a nonparametric Bayesian method for the alignment process, as this process rewards the reuse of parameters, resulting in compact models that align in a consistent manner and tend not to over-fit. Our approach uses the generative model resulting from aligning the training data to force-align the test data. We rely on the simple assumption that correct transliteration pairs would be well modeled and generated easily, whereas incorrect pairs---being more random in character---would be more costly to model and generate. Our generative model generates by concatenating bilingual grapheme sequence pairs. The many-to-many generation process is essential for handling many languages with non-Roman scripts, and it is hard to train well using a maximum likelihood techniques, as these tend to over-fit the data. Our approach works on the principle that generation using only grapheme sequence pairs that are in the model results in a high probability derivation, whereas if the model is forced to introduce a new parameter in order to explain part of the candidate pair, the derivation probability is substantially reduced and severely reduced if the new parameter corresponds to a sequence pair composed of a large number of graphemes. The features we extract from the alignment of the test data are not only based on the scores from the generative model, but also on the relative proportions of each sequence that are hard to generate. The features are used in conjunction with a support vector machine classifier trained on known positive examples together with synthetic negative examples to determine whether a candidate word pair is a correct transliteration pair. In our experiments, we used all data tracks from the 2010 Named-Entity Workshop (NEWS’10) and use the performance of the best system for each language pair as a reference point. Our results show that the new features we propose are powerfully predictive, enabling our approach to achieve levels of performance on this task that are comparable to the state of the art.

Eye gaze is an important means for controlling interaction and coordinating the participants' turns smoothly. We have studied how eye gaze correlates with spoken interaction and especially focused on the combined effect of the speech signal and gazing to predict turn taking possibilities. It is well known that mutual gaze is important in the coordination of turn taking in two-party dialogs, and in this article, we investigate whether this fact also holds for three-party conversations. In group interactions, it may be that different features are used for managing turn taking than in two-party dialogs. We collected casual conversational data and used an eye tracker to systematically observe a participant's gaze in the interactions. By studying the combined effect of speech and gaze on turn taking, we aimed to answer our main questions: How well can eye gaze help in predicting turn taking? What is the role of eye gaze when the speaker holds the turn? Is the role of eye gaze as important in three-party dialogs as in two-party dialogue? We used Support Vector Machines (SVMs) to classify turn taking events with respect to speech and gaze features, so as to estimate how well the features signal a change of the speaker or a continuation of the same speaker. The results confirm the earlier hypothesis that eye gaze significantly helps in predicting the partner's turn taking activity, and we also get supporting evidence for our hypothesis that the speaker is a prominent coordinator of the interaction space. Such a turn taking model could be used in interactive applications to improve the system's conversational performance.

This paper addresses unsupervised speaker clustering for multiparty conversations. Hierarchical clustering methods were mainly used in previous studies. However, these methods require many processes, such as distance calculation and cluster merging, when there are many utterances in conversation data. We propose a clustering method based on non-negative matrix factorization. The proposed method can perform fast and robust clustering by decomposing a matrix consisting of distances between models. We conducted speaker clustering experiments using a Bayesian information criterion based method, a method based on the likelihood ratio between Gaussian mixture models, and the proposed method. Experimental results showed that the proposed method achieves higher clustering accuracy than these conventional methods. Index Terms: unsupervised speaker clustering, non-negative matrix factorization, agglomerative hierarchical clustering, multi-party conversation

Abstract Spoken interactions are known for accurate timing and alignment between interlocutors: turn-taking and topic flow are managed in a manner that provides conversational fluency and smooth progress of the task. This paper studies the relation between the interlocutors’ eye-gaze and spoken utterances, and describes our experiments on turn alignment. We conducted classification experiments by Support Vector Machine on turn-taking using the features for dialogue act, eye-gaze, and speech prosody in conversation data. As a result, we demonstrated that eye-gaze features are important signals in turn management, and seem even more important than speech features when the intention of utterances is clear. Index Terms : eye-gaze, dialogue, interaction, speech analysis, turn-taking 1. Introduction The role of eye-gaze in fluent communication has long since been acknowledged ([2]; [7]). Previous research has established close relations between eye-gaze and conversational feedback ([3]), building trust and rapport, as well as focus of shared attention ([15]). Eye-gaze is also important in turn-taking signalling: usually the interlocutors signal their wish to give the turn by gazing up to the interlocutor, leaning back, and dropping in pitch and loudness, and the partner can, accordingly, start preparing to take the turn. There is evidence that lack of eye contact decreases turn-taking efficiency in video-conferencing ([16]), and that the coupling of speech and gaze streams in a word acquisition task can improve performance significantly ([11]). Several computational models of eye-gaze behaviour for artificial agents have also been designed. For instance, [9] describe an eye-gaze model for believable virtual humans, [13] demonstrate gaze modelling for conversational engage-ment, and [10] built an eye-gaze model to ground information in interactions with embodied conversational agents. Our research focuses on turn-taking and eye-gaze alignment in natural dialogues and especially on the role of eye-gaze as a means to coordinate and control turn-taking. In our previous work [5,6] we noticed that in multi-party dialogues the participants head movement was important in signalling turn-taking, maybe because of its greater visibility than eye-gaze. (This is in agreement with [12], who noticed that in virtual environments, head tracking seems sufficient when people turn their heads to look but if the person is not turning their head to look at an object, then eye-tracking is important to discern the gaze of a person.) The main objective in the current research is to explore the relation between eye-gaze and speech, in particular, how the annotated turn and dialogue features and automatically recognized speech properties affect in turn management. Methodologically our research relies on experimentation and observation: signal-level measurements and analysis of gaze and speech are combined with human-level observation of dialogue events (dialogue acts and turn-taking). We use our three-party dialogue data that is analysed with respect to the interlocutors’ speech, and annotated with dialogue acts, eye-gaze, and turn-taking features [6]. The experiments deal with the classification of turn-taking events using the analysed features and the results show that eye-gaze speech information significantly improves the accuracy compared with the classification with dialogue act information only. However, what is also interesting that the difference between gaze and speech features is not significant, i.e. eye-gaze and speech are important signals in turn management, but their effect is parallel rather than complementary. Moreover, eye-gaze seems to more important than speech when the intention of the utterance is clear. The paper is structured as follows. We first describe the research on turn-taking and the alignment of speech and gaze in Section 2. We then present our data and speech analysis in Section 3, and experimental results as well as discussion concerning their importance in Section 4. Section 5 draws conclusions and points to future research.

This paper describes a system that gives a mobile robot the ability to perform automatic speech recognition with simultaneous speakers. A microphone array is used along with a real-time implementation of geometric source separation (GSS) and a postfilter that gives a further reduction of interference from other sources. The postfllter is also used to estimate the reliability of spectral features and compute a missing feature mask. The mask is used in a missing feature theory-based speech recognition system to recognize the speech from simultaneous Japanese speakers in the context of a humanoid robot. Recognition rates are presented for three simultaneous speakers located at 2 m from the robot. The system was evaluated on a 200-word vocabulary at different azimuths between sources, ranging from 10deg to 90deg. Compared to the use of the microphone array source separation alone, we demonstrate an average reduction in relative recognition error rate of 24% with the postfllter and of 42% when the missing features approach is combined with the postfllter. We demonstrate the effectiveness of our multisource microphone array postfilter and the improvement it provides when used in conjunction with the missing features theory.

This paper investigates issues in preparing corpora for developing speech-to-speech translation (S2ST). It is impractical to create a broad-coverage parallel corpus only from dialog speech. An alternative approach is to have bilingual experts write conversational-style texts in the target domain, with translations. There is, however, a risk of losing fidelity to the actual utterances. This paper focuses on balancing a tradeoff between these two kinds of corpora through the analysis of two newly developed corpora in the travel domain: a bilingual parallel corpus with 420 K utterances and a collection of in-domain dialogs using actual S2ST systems. We found that the first corpus is effective for covering utterances in the second corpus if complimented with a small number of utterances taken from monolingual dialogs. We also found that characteristics of in-domain utterances become closer to those of the first corpus when more restrictive conditions and instructions to speakers are given. These results suggest the possibility of a bootstrap-style of development of corpora and S2ST systems, where an initial S2ST system is developed with parallel texts, and is then gradually improved with in-domain utterances collected by the system as restrictions are relaxed

In this paper, we describe the ATR multilingual speech-to-speech translation (S2ST) system, which is mainly focused on translation between English and Asian languages (Japanese and Chinese). There are three main modules of our S2ST system: large-vocabulary continuous speech recognition, machine text-to-text (T2T) translation, and text-to-speech synthesis. All of them are multilingual and are designed using state-of-the-art technologies developed at ATR. A corpus-based statistical machine learning framework forms the basis of our system design. We use a parallel multilingual database consisting of over 600 000 sentences that cover a broad range of travel-related conversations. Recent evaluation of the overall system showed that speech-to-speech translation quality is high, being at the level of a person having a Test of English for International Communication (TOEIC) score of 750 out of the perfect score of 990.

This paper experimentally compares two automatic evaluators, RED and BLEU, to determine how close the evaluation results of each automatic evaluator are to average evaluation results by human evaluators, following the ATR standard of MT evaluation. This paper gives several cautionary remarks intended to prevent MT developers from drawing misleading conclusions when using the automatic evaluators. In addition, this paper reports a way of using the automatic evaluators so that their results agree with those of human evaluators.

The main goal of the present paper is to propose new schemes for the overall evaluation of a speech translation system. These schemes are expected to support and improve the design of the target application system, and precisely determine its performance. Experiments are conducted on the Japanese-to-English speech translation system ATR-MATRIX, which was developed at ATR Interpreting Telecommunications Research Laboratories. In the proposed schemes, the system’s translations are compared with those of a native Japanese taking the Test of English for International Communication (TOEIC), which is used as a measure of one’s speech translation capability. Subjective and automatic comparisons are made and the results are compared. A regression analysis on the subjective results shows that the speech translation capability of ATR-MATRIX matches a Japanese person scoring around 500 on the TOEIC. The automatic comparisons also show promising results.

An automatic translation quality evaluation method is proposed. In the proposed method, a parallel corpus is used to query translation answer candidates. The translation output is evaluated by measuring the similarity between the translation output and translation answer candidates with DP matching. This method evaluates a language translation subsystem of the Japanese-to-English ATR-MATRIX speech translation system developed at ATR Interpreting Telecommunications Research Laboratories. Discriminant analysis is then carried out to examine the evaluation performance of the proposed method. Experimental results show the effectiveness of the proposed method. The discriminant ratio is 83.5% for 2-class discrimination between absolutely correct and less appropriate translations classified subjectively. Also discussed are issues of the proposed method when it is applied to speech translation systems which inevitably make recognition errors.

In this study, we investigated how well acoustic events extracted from a cross-spectral temporal measure could be used to classify the manner and voicing of consonants. In particular, we developed seven measures that look at the strength and time difference between various onsets and offsets of acoustic energy. Consistent with findings by Shannon et al. (1995), our classification results show that manner and voicing information can be determined from dynamic temporal cues.

ATR-MATRIX is a multi-lingual speech-to-speech translation system designed to facilitate communications between two parties of different languages engaged in a spontaneous conversation in a travel arrangement domain. In this paper, we propose a new evaluation method for speech translation systems. Our current focus is on measuring the robustness of a language translation sub-system, with quick calculation and low cost. Therefore, we calculate the difference between the translation output from transcription texts and the translation output from input speech by a dynamic programming method. We present the first trial experiment of this method applied to our Japanese-to-English speech translation system. We also provide related discussions on such points as error analysis and the relationship between the proposed method and translation quality evaluation manually done by humans.

A system has been developed to rapidly calculate images of parametric rate constants, without acquiring dynamic frame data for clinical positron emission tomography (PET). This method is based on the weighted-integration algorithms for the two- and three-compartment models, and hardware developments (real-time operation and a large cache memory system) in a PET scanner, Headtome-IV, which enables the acquisition of multiple sinograms with independent weight integration functions. Following the administration of the radiotracer, the scan is initiated to collect multiple time-weighted, integrated sinograms with three different weight functions. These sinograms are reconstructed and the images, with the arterial blood data, are inserted into the operational equations to provide parametric rate constant images. The implementation of this method has been checked in H(2)(15 )O and (18)F-fluorophenylalanine ((18)FPhe) studies based on a two-compartment model, and in a (18)F-fluorodeoxyglucose ((18)FDG) study based on the three-compartment model. A volunteer study, completed for each compound, yielded results consistent with those produced by existing nonlinear fitting methods. Thus, this system has been developed capable of generating rapidly quantitative, physiological images, without dynamic data acquisition, which will be of great advantage to PET in the clinical environment. This system would also be of great advantage in the new generation high-resolution PET tomography, which acquire data in a 3-D, septaless mode.

A text-to-speech system for Japanese was developed based on waveform splicing. A stored unit is a sequence of phonemes segmented at vowel-consonant boundaries. Four and eight phoneme groups are distinguished for the preceding and succeeding phonemic environment, respectively. An inventory of waveform segments including frequently used 1020 units was constructed based on a statistical analysis of a text database consisting of 20 million phonemes. Each stored unit has, on average, 2.5 waveform segments with different fundamental frequency (F/sub 0/) and phoneme duration. The F/sub 0/ and phoneme duration are modified by a pitch synchronous overlap add (PSOLA) method. A time window which has a flat portion at its center (Tukey window) was adopted in place of an ordinary Hanning window. A preference test indicated that the Tukey window gives better quality when the F/sub 0/ is lowered. The articulation score of an intelligibility test was 89.2%.>

The use of the public data network for facsimile communication has remarkable advantages with regard to high speed and reliable transmission and, in addition, it is well applicable to provide value added facilities for future service extension. The Research and Development Laboratories of Kokusai Denshin Denwa (KDD) Company, Ltd. have developed an experimental facsimile communication system on the packet switched data network (PSDN). The system is composed of packet switching exchanges (PSE), high speed digital facsimile terminal equipment, a facsimile communication processor (FCP), and facsimile packet assembly/ disassembly equipment (FAX-PAD). The facsimile equipment in the system adopts a two-dimensional modified READ code and has the interface protocol to the PSDN. The FCP provides the store-and-forward facilities to provide such services as delayed delivery, multiaddress delivery, and mailbox service. The FAX-PAD realizes the interworking between facsimile equipments on the public switched telephone network (PSTN) and those on PSDN. After various experiments including international field tests, it was confirmed that the system is useful for future advanced facsimile service.