Tomoko Matsui

Bridging and Relevance

While it has long been taken for granted that context or background information plays a crucial role in reference assignment, there have been very few serious attempts to investigate exactly how they are used. This study provides an answer to the question through an extensive analysis of cases of bridging. The book demonstrates that when encountering a referring expression, the hearer is able to choose a set of contextual assumptions intended by the speaker in a principled way, out of all the assumptions possibly available to him. It claims more specifically that the use of context, as well as the assignment of referent, is governed by a single pragmatic principle, namely, the principle of relevance (Sperber a Wilson 1986/1995), which is also a single principle governing overall utterance interpretation. The explanatory power of the criterion based on the principle of relevance is tested against the two major, current alternatives — truth-based criteria and coherence-based criteria — using data elicited in a battery of referent assignment questionnaires. The results show clearly that the relevance-based criterion has more predictive power to handle a wider range of examples than any other existing criterion. As such, this work adds to the growing body of evidence supporting the insights of relevance theory. The work has been awarded the 2001 Ichikawa Award for the best achievement in English Linguistics by a young scholar in Japan.

Semantics and pragmatics of a Japanese discourse marker dakara (so/in other words): a unitary account

Abstract In this paper, I discuss the information encoded by the Japanese discourse connective dakara, and claim that the fundamental function of the connective is to mark the utterance that follows it as an interpretive representation of another (set of) utterance(s)/assumption(s): for instance, as a reformulation or a contextual implication of another (set of) utterance(s)/assumption(s). Unlike most existing accounts of dakara which see the connective as being polysemous, the present approach provides a unitary account, where the notion of interpretive resemblance [Relevance (1986)] plays a crucial role. Furthermore, using, as a point of departure, Blakemore’s analyses of English reformulation markers such as in other words, as well as her work on English discourse connectives such as so and therefore, I will argue that dakara is better analyzed as encoding procedural information instructing the hearer to identify a (set of) representation(s) which interpretively resembles the utterance introduced by the connective. By guiding the construction of appropriate higher-level representations of the utterance, the connective contributes to the explicit side of communication.

Understanding of speaker certainty and false-belief reasoning: a comparison of Japanese and German preschoolers

It has been repeatedly shown that when asked to identify a protagonists false belief on the basis of his false statement, English-speaking 3-year-olds dismiss the statement and fail to attribute to him a false belief. In the present studies, we tested 3-year-old Japanese children in a similar task, using false statements accompanied by grammaticalized particles of speaker (un)certainty, as in everyday Japanese utterances. The Japanese children were directly compared with same-aged German children, whose native language does not have grammaticalized epistemic concepts. Japanese children profited from the explicit statement of the protagonists false belief when it was marked with the attitude of certainty in a way that German children did not - presumably because Japanese but not German children must process such marking routinely in their daily discourse. These results are discussed in the broader context of linguistic and theory of mind development.

Robust Speech Recognition Using Generalized Distillation Framework.

In this paper, we propose a noise robust speech recognition system built using generalized distillation framework. It is assumed that during training, in addition to the training data, some kind of ”privileged” information is available and can be used to guide the training process. This allows to obtain a system which at test time outperforms those built on regular training data alone. In the case of noisy speech recognition task, the privileged information is obtained from a model, called ”teacher”, trained on clean speech only. The regular model, called ”student”, is trained on noisy utterances and uses teacher’s output for the corresponding clean utterances. Thus, for this framework a parallel clean/noisy speech data are required. We experimented on the Aurora2 database which provides such kind of data. Our system uses hybrid DNN-HMM acoustic model where neural networks provide HMM state probabilities during decoding. The teacher DNN is trained on the clean data, while the student DNN is trained using multi-condition (various SNRs) data. The student DNN loss function combines the targets obtained from forced alignment of the training data and the outputs of the teacher DNN when fed with the corresponding clean features. Experimental results clearly show that distillation framework is effective and allows to achieve significant reduction in the word error rate.

Experimental Pragmatics: Towards Testing Relevance-Based Predictions about Anaphoric Bridging Inferences

Critics of relevance theory have charged that relevance theory provides no testable predictions, hence is unfalsifiable. This paper is an attempt to identify some testable relevance-theoretic predictions about anaphoric bridging inferences, and to show possible ways of testing these predictions. A relevance-based model of utterance interpretation is compared with Levinsons GCI model, and their contrasting views on how to divide explicitly communicated content and implicitly communicated content of utterance are discussed. Moreover, predictions following each theory about derivation of bridging inferences are compared before possible ways to test these predictions are suggested.

Voice Liveness Detection for Speaker Verification based on a Tandem Single/Double-channel Pop Noise Detector.

This paper presents an algorithm for detecting spoofing attacks against automatic speaker verification (ASV) systems. While such systems now have performances comparable to those of other biometric modalities, spoofing techniques used against them have progressed drastically. Several techniques can be used to generate spoofing materials (e.g., speech synthesis and voice conversion techniques), and detecting them only on the basis of differences at an acoustic speaker modeling level is a challenging task. Moreover, differences between “live” and artificially generated material are expected to gradually decrease in the near future due to advances in synthesis technologies. A previously proposed “voice liveness” detection framework aimed at validating whether speech signals were generated by a person or artificially created uses elementary algorithms to detect pop noise. Detection is taken as evidence of liveness. A more advanced detection algorithm has now been developed that combines singleand double-channel pop noise detection. Experiments demonstrated that this tandem algorithm detects pop noise more effectively: the detection error rate was up to 80% less that those achieved with the elementary algorithms.

The role of prosody and context in sarcasm comprehension: Behavioral and fMRI evidence.

A hearers perception of an utterance as sarcastic depends on integration of the heard statement, the discourse context, and the prosody of the utterance, as well as evaluation of the incongruity among these aspects. The effect of prosody in sarcasm comprehension is evident in everyday conversation, but little is known about its underlying mechanism or neural substrates. To elucidate the neural underpinnings of sarcasm comprehension in the auditory modality, we conducted a functional MRI experiment with 21 adult participants. The participants were provided with a short vignette in which a child had done either a good or bad deed, about which a parent made a positive comment. The participants were required to judge the degree of the sarcasm in the parents positive comment (praise), which was accompanied by either positive or negative affective prosody. The behavioral data revealed that an incongruent combination of utterance and the context (i.e., the parents positive comment on a bad deed by the child) induced perception of sarcasm. There was a significant interaction between context and prosody: sarcasm perception was enhanced when positive prosody was used in the context of a bad deed or, vice versa, when negative prosody was used in the context of a good deed. The corresponding interaction effect was observed in the rostro-ventral portion of the left inferior frontal gyrus corresponding to Brodmanns Area (BA) 47. Negative prosody incongruent with a positive utterance (praise) activated the bilateral insula extending to the right inferior frontal gyrus, anterior cingulate cortex, and brainstem. Our findings provide evidence that the left inferior frontal gyrus, particularly BA 47, is involved in integration of discourse context and utterance with affective prosody in the comprehension of sarcasm.

Japanese preschoolers’ evaluation of circular and non-circular arguments

Abstract Observational and experimental data have revealed that preschoolers possess some argumentation skills, both in the production and the evaluation of arguments. However, these skills might have been fostered by the particular cultural context of Western middle- and upper-classes families, to which most children studied belong. Some data suggests that children in other cultures possess at least some of these skills, but no experimental data had been gathered in Eastern cultures. These cultures are supposed to frown on argumentation, and might thus be less conducive to the early development of argumentation skills. We test the emergence of argument evaluation skills in Japanese 5-year-olds by presenting them with a choice between endorsing a strong, perceptual argument, and a weak, circular argument. A first experiment revealed a trend in the direction of the strong argument. A second experiment that addresses some methodological concerns of the first demonstrates a significant tendency to follow the strong argument. These results are similar to those previously gathered in two other cultures (Swiss and Maya), and suggest that some basic argumentation skills are early developing across cultures.

On the Role of Context in Relevance-Based Accessibility Ranking of Candidate Referents

It is now a standard view that candidate referents are ranked according to their accessibility, based on either or both of the discourse structure and organisation of general knowledge. Various models of reference resolution based on such a view, including centering theory, have been ardently pursued and tested empirically, and the ones which are computationally workable have been given preferences in the past. In this paper, I propose an alternative view on accessibility ranking based on the consideration of relevance. Although it has not been tested computational• •, certain advantages over existing discourse-based approaches are demonstrated. It is suggested that one possible start to test the present approach computationally is to focus on the use of discourse connectives which constrain contextual assumptions by directing the way an utterance is likely to achieve relevance.

Children's Understanding of Certainty and Evidentiality: Advantage of Grammaticalized Forms over Lexical Alternatives

In verbal communication, the hearer takes advantage of the linguistic expressions of certainty and evidentiality to assess how committed the speaker might be to the truth of the informational content of the utterance. Little is known, however, about the precise developmental mechanism of this ability. In this chapter, we approach the question by elucidating factors that are likely to constrain young childrens understanding of linguistically encoded certainty and evidentiality, including the types of linguistic form of these expressions, namely, grammaticalized or lexical forms.