Cross-Lingual spoken language understanding from unaligned data using discriminative classification models and machine translation

This paper investigates several approaches to bootstrapping a new spoken language understanding (SLU) component in a target language given a large dataset of semantically-annotated utterances in some other source language. The aim is to reduce the cost associated with porting a spoken dialogue system from one language to another by minimising the amount of data required in the target language. Since word-level semantic annotations are costly, Semantic Tuple Classifiers (STCs) are used in conjunction with statistical machine translation models both of which are trained from unaligned data to further reduce development time. The paper presents experiments in which a French SLU component in the tourist information domain is bootstrapped from English data. Results show that training STCs on automatically translated data produced the best performance for predicting the utterance's dialogue act type, however individual slot/value pairs are best predicted by training STCs on the source language and using them to decode translated utterances. © 2010 ISCA.

Continuous asr for flexible incremental dialogue

Spoken dialogue systems provide a convenient way for users to interact with a machine using only speech. However, they often rely on a rigid turn taking regime in which a voice activity detection (VAD) module is used to determine when the user is speaking and decide when is an appropriate time for the system to respond. This paper investigates replacing the VAD and discrete utterance recogniser of a conventional turn-taking system with a continuously operating recogniser that is always listening, and using the recogniser 1-best path to guide turn taking. In this way, a flexible framework for incremental dialogue management is possible. Experimental results show that it is possible to remove the VAD component and successfully use the recogniser best path to identify user speech, with more robustness to noise, potentially smaller latency times, and a reduction in overall recognition error rate compared to using the conventional approach. © 2013 IEEE.

Efficient decoding with generative score-spaces using the expectation semiring

State-of-the-art speech recognisers are usually based on hidden Markov models (HMMs). They model a hidden symbol sequence with a Markov process, with the observations independent given that sequence. These assumptions yield efficient algorithms, but limit the power of the model. An alternative model that allows a wide range of features, including word- and phone-level features, is a log-linear model. To handle, for example, word-level variable-length features, the original feature vectors must be segmented into words. Thus, decoding must find the optimal combination of segmentation of the utterance into words and word sequence. Features must therefore be extracted for each possible segment of audio. For many types of features, this becomes slow. In this paper, long-span features are derived from the likelihoods of word HMMs. Derivatives of the log-likelihoods, which break the Markov assumption, are appended. Previously, decoding with this model took cubic time in the length of the sequence, and longer for higher-order derivatives. This paper shows how to decode in quadratic time. © 2013 IEEE.

In Praise of Folly: International Administration and the Corruption of Humanity

Bain, William, 'In Praise of Folly: International Administration and the Corruption of Humanity', International Affairs, (2006) 82(3) pp.525-538 RAE2008

Lektura twórcza w edukacji humanistycznej

Wydział Filologii Polskiej i Klasycznej UAM: Instytut Filologii Polskiej

Idiom processing in aphasic patients

The aim of the present paper is to provide insight into the issue of idiom comprehension in pa- tients who are in the process of recovery from the syndrome of aphasia. Research in figurative language comprehension has seen a robust development in the recent decades. However, it has not been until quite recently that psycholinguists began to delve into the aspect of metaphorical language comprehension in brain damaged populations. It was observed that even though the ability to produce and understand language is recovered in the majority of patients with head trauma, the impairment of some aspects of comprehension may protract. The understanding of idioms, metaphors, similes and proverbs, due to their specific, non-literal character, has been evi- denced to pose a serious problem to aphasic patients, as they fail to decipher the figurative mean- ing of the utterance, and, instead, tend to process the message literally (Papagno et al. 2004). In the present study, three patients who suffered from aphasic disorder were tested for com- prehension of idioms by means of two multiple choice tasks. The obtained results corroborated the hypothesis that patients who are in the process of recovery from aphasia encounter various pitfalls in the comprehension of idiomatic language. Predominantly, they exhibit an inclination to choose the erroneous, literal paraphrases of the presented idioms over their correct, idiomatic counterparts. The present paper aims at accounting for the reasons underlying such a tendency.

When some is actually all: Scalar inferences in face-threatening contexts

Accounts of the scalar inference from 'some X-ed' to 'not all X-ed' are central to the debate between contemporary theories of conversational pragmatics. An important contribution to this debate is to identify contexts that decrease the endorsement rate of the inference. We suggest that the inference is endorsed less often in face-threatening contexts, i.e., when X implies a loss of face for the listener. This claim is successfully tested in Experiment 1. Experiment 2 rules out a possible confound between face-threatening contexts and lower-bound contexts. Experiment 3 shows that whilst saying 'some X-ed' when one knew for a fact that all X-ed is always perceived as an underinformative utterance, it is also seen as a nice and polite thing to do when X threatens the face of the listener. These findings are considered from the perspective of Relevance Theory as well as that of the Generalized Conversational Inference approach. (C) 2009 Elsevier B.V. All rights reserved.

Kawamoto’s "Dojoji Temple": Puppetry, Gesture and Myth

Dojoji Temple ( Dōjōji, 1976) is a short puppet animation directed by Kihachirō Kawamoto. Influenced by Bunraku (Japanese puppet plays), emaki (painted scroll), Noh theatre and Japanese myth, Dojoji Temple tells of a woman’s unrequited love for a young priest. Heartbroken, she then transforms into a sea serpent and goes after the priest for revenge. While Kawamoto’s animation is rich with Japanese aesthetics and tragedy, his animation is peopled by puppets who do not speak. Limited and restrained though the puppets may be, their animated gestures speak volumes of powerful emotions. For our article, we will select several scenes from the animation, and interpret their actions so that we can further understand the mythical world of Dojoji Temple and the essential being of puppetry. Our gesture analysis will take into account cinematographic compositions, sound and bodily attires, among other elements.

They’re not racist…’ Prejudice denial, mitigation and suppression in dialogue

Social scientific work on the suppression, mitigation or denial of prejudiced attitudes has tended to focus on the strategic self-presentation and self-monitoring undertaken by individual social actors on their own behalf. In this paper, we argue that existing perspectives might usefully be extended to incorporate three additional considerations. First, that social actors may, on some occasions, act to defend not only themselves, but also others from charges of prejudice. Second, that over the course of any social encounter, interactants may take joint responsibility for policing conversation and for correcting and suppressing the articulation of prejudiced talk. Third, that a focus on the dialogic character of conversation affords an appreciation of the ways in which the status of any particular utterance, action or event as 'racist' or 'prejudiced' may constitute a social accomplishment. Finally, we note the logical corollary of these observations - that in everyday life, the occurrence of 'racist discourse' is likely to represent a collaborative accomplishment, the responsibility for which is shared jointly between the person of the speaker and those other co-present individuals who occasion, reinforce or simply fail to suppress it.

'Regina v John Terry': The Discursive Construction of an Alleged Racist Event

This article explores the conformation in discourse of a verbal exchange and its subsequent mediatised and legal ramifications. The event concerns an allegedly racist insult directed by high profile English professional footballer John Terry towards another player, Anton Ferdinand, during a televised match in October 2011. The substance of Terry’s utterance, which included the noun phrase ‘fucking black cunt’, was found by a Chief Magistrate not to be a racist insult, although the fact that these actual words were framed within the utterance was not in dispute. The upshot of this ruling was that Terry was acquitted of a racially aggravated public order offence. A subsequent investigation by the regulatory commission of the English Football Association (FA) ruled, almost a year after the event, that Terry was guilty of racially abusing Ferdinand. Terry was banned for four matches and fined £220,000.

It is our contention that this event, played out in legal rulings, social media and print and broadcast media, constitutes a complex web of linguistic structures and strategies in discourse, and as such lends itself well to analysis with a broad range of tools from pragmatics, discourse analysis and cognitive linguistics. Amongst other things, such an analysis can help explain the seemingly anomalous - even contradictory - position adopted in the legal ruling with regard to the speech act status of ‘fucking black cunt’; namely, that the racist content of the utterance was not contested but that the speaker was found not to have issued a racist insult. Over its course, the article addresses this broader issue by making reference to the systemic-functional interpersonal function of language, particularly to the concepts of modality, polarity and modalisation. It also draws on models of verbal irony from linguistic pragmatics, notably from the theory of irony as echoic mention (c.f. Sperber and Wilson, 1981; Wilson and Sperber, 1992). Furthermore, the article makes use of the cognitive-linguistic framework, Text World Theory (c.f. Gavins, 2007; Werth, 1999) to examine the discourse positions occupied by key actors and adapts, from cognitive poetics, the theory of mind-modelling (c.f. Stockwell, 2009) to explore the conceptual means through which these actors discursively negotiate the event.

It is argued that the pragmatic and cognitive strategies that frame the entire incident go a long way towards mitigating the impact of so ostensibly stark an act of racial abuse. Moreover, it is suggested here that the reconciliation of Terry’s action was a result of the confluence of strategies of discourse with relations of power as embodied by the media, the law and perceptions of nationhood embraced by contemporary football culture. It is further proposed that the outcome of this episode, where the FA was put in the spotlight, and where both the conflict and its key antagonists were ‘intranational’, was strongly impelled by the institution of English football and its governing body both to reproduce and maintain social, cultural and ethnic cohesion and to avoid any sense that the event featured a discernable ‘out-group’.

Ficção e encenação : o desenho como narrativa visual

Tese de doutoramento, Belas-Artes (Ciências da Arte), Universidade de Lisboa, Faculdade de Belas-Artes, 2014

A Tradução de referências à luz da skopostheorie – estudo de um caso

Polissema: Revista de Letras do ISCAP 2001/N.º 1 - Tradução

Correlates of reading readiness in kindegarten children

The gene,nil purpose (If this thesis was to examine certain variables considered to be associated with reading readiness in kindergarten children Twenty-four children from a parent-funded parochial school i.n 5t Catharines. Ontaj-io were exanlined in this study, The children.'s reading readiness level, measured by the Test of Ea1'1y Reading Ability (JERA:) and by teacher"s ratings was correlated with various cognitive variables, These variables consisted (If a rough index of intelligence as measured by the Peabody Picture Vocabulary Test - Revised CPPVT-Rl Form M, mean len.gth of utterance (MUn, the sum of relative,. subordinate and coordinate clauses, the !wmber of core statements children make when telling stories and a memory test. All but the memory test related to at least one measure of reading readiness, Although the memory test did not correlate significantly with reading readiness, the children appeaj~ed to be sensitive to the stimulus set size in terms of their study time, In addition. interi'uption in the interval between studying the stimuli and the ,reca1i test had a negathre effect on pe.rfornlance and set size had a substantial effect 011 recognition performance, The educational implications of these correlates of reading readiness are discussed as weH as the implications for future research.

Cognition sociale et traitement du contexte dans la schizophrénie : effets des stéréotypes

Un déficit de la cognition sociale incluant l’attribution d’états mentaux, la perception sociale et la perception des émotions est l’une des caractéristiques les plus handicapantes de la schizophrénie. Les résultats d’une étude par Hardy-Baylé et al. (2003) suggèrent que les difficultés des schizophrènes (SZ) à attribuer des intentions aux autres seraient dues à l’incapacité à utiliser l’information contextuelle. Certaines études (Ivanko & Pexman, 2003; Pexman & Olineck, 2002b) démontrent que des facteurs comme le degré d’incongruité entre le contexte et l’énoncé, l’intonation de la voix et les caractéristiques des personnages peuvent influencer la compréhension de l’intention ironique chez les sujets sains (CT). L’objectif de cette étude est de manipuler des informations contextuelles afin de déterminer si le type de métier du locuteur est un indice social facilitant la compréhension de l’intention du locuteur (théorie de l’esprit ; TdE) et de la perception sociale chez les patients SZ. Trente participants SZ appariés avec trente participants CT ont été recrutés. Ils ont été évalués sur la compréhension de l’ironie et sur la mémoire de travail. Les histoires étaient développées selon deux conditions de métier: un métier favorisant la compréhension de l’ironie (e.g. comédien) et un métier ne favorisant pas la compréhension de l’ironie (e.g. prêtre). Les résultats montrent que les patients SZ ont un trouble de TdE et ils ne semblent pas sensibles aux stéréotypes contrairement aux participants CT. Toutefois, les résultats indiquent que les participants SZ n’ont pas de déficit concernant la perception sociale.