In this thesis, we describe extensive experiments on the classification of emotions from speech using acoustic features. This area of research has important applications in human computer interaction. We have thoroughly reviewed the current literature and present our results on some of the contemporary emotional speech databases. The principal focus is on creating a large set of acoustic features, descriptive of different emotional states and finding methods for selecting a subset of best performing features by using feature selection methods. In this thesis we have looked at several traditional feature selection methods and propose a novel scheme which employs a preferential Borda voting strategy for ranking features. The comparative results show that our proposed scheme can strike a balance between accurate but computationally intensive wrapper methods and less accurate but computationally less intensive filter methods for feature selection. By using the selected features, several schemes for extending the binary classifiers to multiclass classification are tested. Some of these classifiers form serial combinations of binary classifiers while others use a hierarchical structure to perform this task. We describe a new hierarchical classification scheme, which we call Data-Driven Dimensional Emotion Classification (3DEC), whose decision hierarchy is based on non-metric multidimensional scaling (NMDS) of the data. This method of creating a hierarchical structure for the classification of emotion classes gives significant improvements over other methods tested. The NMDS representation of emotional speech data can be interpreted in terms of the well-known valence-arousal model of emotion. We find that this model does not give a particularly good fit to the data: although the arousal dimension can be identified easily, valence is not well represented in the transformed data. From the recognition results on these two dimensions, we conclude that valence and arousal dimensions are not orthogonal to each other. In the last part of this thesis, we deal with the very difficult but important topic of improving the generalisation capabilities of speech emotion recognition (SER) systems over different speakers and recording environments. This topic has been generally overlooked in the current research in this area. First we try the traditional methods used in automatic speech recognition (ASR) systems for improving the generalisation of SER in intra– and inter–database emotion classification. These traditional methods do improve the average accuracy of the emotion classifier. In this thesis, we identify these differences in the training and test data, due to speakers and acoustic environments, as a covariate shift. This shift is minimised by using importance weighting algorithms from the emerging field of transfer learning to guide the learning algorithm towards that training data which gives better representation of testing data. Our results show that importance weighting algorithms can be used to minimise the differences between the training and testing data. We also test the effectiveness of importance weighting algorithms on inter–database and cross-lingual emotion recognition. From these results, we draw conclusions about the universal nature of emotions across different languages.

Over the previous few decades there has been an increased emphasis on the cultural aspects of English language teaching. However, in settings where English is used as a global lingua franca the cultural associations of the language are complex and the role culture plays in successful communication has yet to be extensively investigated. To conduct such a study it is necessary to explicate the role and nature of English in global contexts and particularly how English functions as a lingua franca (ELF). Furthermore, a theoretical understanding of the relationships between languages and cultures in intercultural communication is needed, which emphasises the fluid and dynamic nature of any connections. The thesis focuses on cultural awareness (CA) as an approach to equipping learners and users of English for the diversity of intercultural communication. However, it is suggested that CA has still not incorporated an understanding of the multifarious uses of English in global contexts where no clear cultural associations can be established. Thus, intercultural awareness (ICA) is offered as an alternative which addresses these needs. This results in the formulation of research questions which aim to explore how ICA can best be characterised in an expanding circle setting and the role it plays in intercultural communication. Furthermore, this research also aims to explicate the relationships between the English language and cultures in such an environment and how this reflects on language use and attitudes. The study was predominantly qualitative utilising approaches associated with ethnography with the aim of producing a rich description of the research participants and their environment. The fieldwork took place over a six month period in a Thai university and seven participants formed the core of this study. The main data sources were recordings of the participants engaged in intercultural communication and interviews with the participants. These were supplemented with a survey, diaries, observations and documents from the research site. The findings of the study suggest that in successful intercultural communication culturally based forms, practices and frames of reference are employed as emergent, dynamic and liminal resources in a manner that moves between individual, local, national and global references. Furthermore, the results also indicated that ICA was a valid construct in the context investigated for explaining the types of cultural knowledge and related skills needed by participants to take part in successful intercultural communication through English.

The aim of this study is to establish what insights the tracking of both input and output might contribute to corpus-based analyses of the acquisition of chunks among instructed adult second language learners. The automated analysis of electronic corpora of natural language has played a major role in revealing the prevalence of conventionalised word sequences in human communication, thereby challenging predominant atomistic conceptualisations of linguistic processing. While the mastery of nativelike phraseology would appear to be central to second language acquisition, reports of instructed adult learners have commonly highlighted their deficiencies in employing multi-word sequences as compared to native speaker norms. A problem with such studies has been their product-orientated focus, the majority tending to conflate the attempts of multiple learners at a range of chunks, variously specified, at a single point in time, the absence of information as to their corresponding instructional input making it impossible to compare their performance against their levels of exposure to the formulas in question. What appears to be missing is an examination of patterns of acquisition among a set of learners in respect of the same chunk, in relation to input, and over time. The aim of this exploratory investigation is to attempt to fill this gap among existing studies of second language chunk acquisition in instructed learning contexts by providing a window on both the processes and products involved. Drawing on an especially created 170 000 word longitudinal corpus composed of online classroom interaction, the study tracks the oral exposure to and use of a single internally complex word combination of 36 learners on an Open University beginners’ Spanish course. The study uncovers a multifaceted picture of classroom input and output in respect of the same sequence and reveals that, while there is a correlation between frequency of overall exposure and the learners’ propensity to attempt the chunk, this masks considerable variation in the form these attempts take for each individual over time. These findings underline the need to look beyond an amalgamated snapshot of learners’ use of chunks and consider individual differences in recalling and reproducing specific exemplars in relation to exposure to these sequences, while inviting further investigation into the factors that underpin such variation and continued enquiry into those aspects of input that might usefully contribute to this process.

This thesis investigates some of the practices by which interactants engage in responding to an informing in Greek talk-in-interaction. Using the analytical methodology of Conversation Analysis (CA), I investigate the ways in which responses to informings are typically constructed and how speakers recruit the assistance of their interlocutors in order to format their following action by examining the following particles: entaksi (mainly in the beginning of a turn), ne (=yes) with a questioning prosody, ela + name (sometimes incremented with the Greek particle re), bravo, etsi den ine (a form of tag question in Greek) and etsi in the final position of a turn. The analysis focuses on the sequential and social implications of these particles in interaction and suggests that there are certain resources interactants deploy in response to an informing, especially in turn-initial position, to indicate their stance towards the prior turn. Interlocutors deploy different practices in talk that serve the avoidance of conflicts, especially in the context of interactions between friends and intimates. The tokens under investigation are deployed by recipients of an informing to position themselves towards a prior turn but at the same time indicate the degree to which they accept the informing, absolute agreement or preliminary to a disagreement. As the first conversation analytic investigation of information receipts in Greek talk-in-interaction, this study attempts to illustrate the interactional significance of receipt tokens in the organization of talk and the accomplishment of actions in interaction.

Topic models are known to suffer from sparsity when applied to short text data. The problem is caused by a reduced number of observations available for a reliable inference (i.e.: the words in a document). A popular heuristic utilized to overcome this problem is to perform before training some form of document aggregation by context (e.g.: author, hashtag). We dedicated one part of this dissertation to modeling explicitly the implicit assumptions of the document aggregation heuristic and applying it to two well known model architectures: a mixture and an admixture. Our findings indicate that an admixture model benefits more from aggregation compared to a mixture model which rarely improved over its baseline (the standard mixture). We also find that the state of the art in short text data can be surpassed as long as every context is shared by a small number of documents. In the second part of the dissertation we develop a more general purpose topic model which can also be used when contextual information is not available. The proposed model is formulated around the observation that in normal text data, a classic topic model like an admixture works well because patterns of word co-occurrences arise across the documents. However, the possibility of such patterns to arise in a short text dataset is reduced. The model assumes every document is a bag of word co-occurrences, where each co-occurrence belongs to a latent topic. The documents are enhanced a priori with related co-occurrences from the other documents, such that the collection will have a greater chance of exhibiting word patterns. The proposed model performs well managing to surpass the state of the art and popular topic model baselines.

Scottish letters and court-records from the late 16th and early 17th centuries give access to a rich variety of conflicts, ranging from international disputes to everyday spats. This thesis investigates verbal offences reported by correspondents or recorded as legal evidence. Current models of (im)politeness (Culpeper, 2011a, Spencer-Oatey, e.g. 2005), which have rarely been tested on historical data, are synthesized with insights from historical pragmatics. The aims are to create qualitative reconstructions of how participants communicated their period- and situation-specific understandings of verbal offences, and how their expressed perceptions were shaped by private and public dimensions of different contexts. This thesis thus addresses three comparatively understudied aspects of (im)politeness research: historical impoliteness, Scottish (im)politeness, and the examination of private-public aspects of social interactions. Regarding the third point, a multi-dimensional framework is developed for systematic descriptions of private-public dimensions of conflict-situations, remodelling an existing pragmatic approach to news discourse (Landert and Jucker, 2011). Letters are drawn from the Breadalbane Collection, 1548-1583 (Dawson, 2004/2007) and James VI’s correspondence. Court-records are selected from editions of Justice Court papers and Kirk Session minutes. Case studies reveal that the vocabulary and discursive structure of conflict-narratives in letters is largely distinct from reported offences in court-records. Differences are presumably influenced by the genres’ contrasting contextual functions of more private versus more public conflict-settlement. However, the language of conflict-letters and court-records also shows shared moral and religious dimensions. Furthermore, verbal offences in the investigated letters and lawsuits refer primarily to collective identity-aspects of group-membership and social roles, while purely individual qualities appear to have been marginal. The perceived gravity of offence could be intensified by participants’ notions of private and public in multiple ways. Concerning comparisons within genres, the Scottish king’s epistolary language in conflicts shows similarities to that of Scottish upper-rank correspondents; nevertheless, it also has some distinct features reflecting James VI’s understanding of his royal status. Criminal and ecclesiastical court-records had largely different, yet semantically related, inventories of verbal offence terms.

No description available.

Researchers have studied voice onset time (VOT) in a number of languages but there is a scarcity of research on the acquisition of VOT of English, particularly by adult Saudi learners, and on the VOT of Saudi Arabic. The current study aims to fill these gaps. At the same time, we aimed to assess whether key claims of Flege's Speech Learning Model (SLM) were supported by this kind of data. 31 adult advanced Saudi learners of English and 60 monolinguals (30 native English and 30 Arabic monolinguals) participated in this study. The VOTs of the voiced and voiceless stops were measured followed by three different vowels, in both isolated word and word in sentence contexts. The results show that the learners produced English voiceless stops with aspiration closer to Arabic than to the higher native English VOT values, and voiced stops with pre-voicing, similar to Arabic, rather than with native English short-lag VOT values. Context had an effect in English but vowel did not, while the reverse was true for the learners and Arabic native speakers. Overall, learners' acquisition was modest despite their level and exposure, in that they overwhelmingly resembled Arabic rather than English native speakers. Several hypotheses based on SLM expectations were not confirmed in an unqualified way. However, support was found for learners' phonetic categories being ‘deflected’ away from both L1 and L2 categories. All three groups produced longer positive VOT for aspirated than unaspirated or voiced plosives. All exhibited VOT increasing across places of articulation, front to back for the voiceless stops, but only English native speakers showed this clearly for the voiced stops. Length of residence in UK and daily use of English did not seem to affect nativelikeness of learner VOT.

This thesis interrogates the construction of identity and self-presentation strategies in the discourse of the current President of the United States of America, Barack Hussein Obama. The study seeks to answer questions about how the President constructs the various identities evident in his discourse, what kinds of resources are drawn upon, and how the resulting identities contribute to gain the support of the audience and the progression of political discourse in general. The present study sheds light on the construction of the personal, relational and collective identities utilising a pluralistic mixed-method approach. It draws upon the tools provided by corpus linguistics alongside a more fine-grained, narrative-based critical discourse analysis. The qualitative analysis offers a methodological synergy based on the insights of research conducted in critical discourse analysis, sociolinguistics and narrative analysis. The study investigates not only the way identities are constructed and defended, but also their significance in shaping the professional image of the President as a caring and self-made leader. Moreover, the study examines the construction of attitudinal identity in Obama’s discourse, whether in reflecting upon his own attitude or in reference to the collective identity of the American people or the Democratic Party as a whole. The study concludes with a consideration of the potential significance of the present research, along with suggestions for future research explorations.

This thesis examines the vernacular psychology presented in Old Norse-Icelandic texts. It focuses on the concept 'hugr', generally rendered in English as ‘mind, soul, spirit’, and explores the conceptual relationships between emotion, cognition and the body. It argues that despite broad similarities, Old Norse-Icelandic and Old English vernacular psychology differ more than has previously been acknowledged. Furthermore, it shows that the psychology of Old Norse-Icelandic has less in common with its circumpolar neighbours than proposed by advocates of Old Norse-Icelandic shamanism. The thesis offers a fresh interpretation of Old Norse-Icelandic psychology which does not rely on cross-cultural evidence from other Germanic or circumpolar traditions. In particular, I demonstrate that emotion and cognition were not conceived of ‘hydraulically’ as was the case in Old English, and that 'hugr' was not thought to leave the body either in animal form or as a person’s breath. I show that Old Norse-Icelandic psychology differs from the Old English tradition, and argue that the Old English psychological model is a specific elaboration of the shared psychological inheritance of Germanic whose origins require further study. These differences between the two languages have implications for the study of psychological concepts in Proto-Germanic, as I argue that there are fewer semantic components which can be reliably reconstructed for the common ancestor of the North and West Germanic languages. As a whole, the thesis applies insights from cross-cultural linguistics and psychology in order to show how Old Norse-Icelandic psychological concepts differ not only from contemporary Germanic and circumpolar traditions but also from the Present Day English concepts used to describe them. The thesis comprises four chapters and conclusion. Chapter 1 introduces the field of study and presents the methodologies and sources used. It introduces the range of cross-cultural variety in psychological concepts, and places Old Norse-Icelandic 'hugr' and its Old English analogue 'mōd' in a typological perspective. Chapter 2 reviews previous approaches to early Germanic psychology and introduces the major strand of research that forms the background to this study: Lockett’s (2011) proposal that Old English vernacular psychology operated in terms of a ‘hydraulic model’, where the 'mōd' would literally boil and seethe within a person’s chest in response to strong emotions. Chapter 3 outlines the native Old Norse-Icelandic psychological model by examining indigenously produced vernacular texts. It looks first at the claims that 'hugr' could leave the body in animal form or as a person’s breath. It then describes the relationship between emotion, cognition and the body in Old Norse-Icelandic texts and contrasts this with the Old English system. Chapter 4 examines the foreign influences which could potentially account for the differences between the Old English and Old Norse-Icelandic systems. It looks first at the imported medical traditions which were known in medieval Scandinavia at the time Old Norse-Icelandic texts were being committed to writing. Next it considers the psychology of Christian tradition from the early Old Icelandic Homily Book to late-fourteenth-century devotional poetry. Finally, it examines the representation of emotion and the body in the translated Anglo-Norman and Old French texts produced at the court of Hákon Hákonarson and explores how this was transposed to native romances composed in Old Norse-Icelandic. The conclusion summarises the findings of the thesis and presents a proposal for the methodology of studying medieval psychological concepts with directions for further research.