This thesis is an inquiry into the nature of the high-level, rhetorical structure of unrestricted natural language texts, computational means to enable its derivation, and two applications (in automatic summarization and natural language generation) that follow from the ability to build such structures automatically. The thesis proposes a first-order formalization of the high-level, rhetorical structure of text. The formalization assumes that text can be sequenced into elementary units; that discourse relations hold between textual units of various sizes; that some textual units are more important to the writer's purpose than others; and that trees are a good approximation of the abstract structure of text. The formalization also introduces a linguistically motivated compositionality criterion, which is shown to hold for the text structures that are valid. The thesis proposes, analyzes theoretically, and compares empirically four algorithms for determining the valid text structures of ...

Semantic Role Labeling (SRL) has proved to be a valuable tool for performing automatic analysis of natural language texts. Currently however, most systems rely on a large training set, which is manually annotated, an effort that needs to be repeated whenever different languages or a different set of semantic roles is used in a certain application. A possible solution for this problem is semi-supervised learning, where a small set of training examples is automatically expanded using unlabeled texts. We present the Latent Words Language Model, which is a language model that learns word similarities from unlabeled texts. We use these similarities for different semi-supervised SRL methods as additional features or to automatically expand a small training set. We evaluate the methods on the PropBank dataset and find that for small training sizes our best performing system achieves an error reduction of 33.27 % F1-measure compared to a state-of-the-art supervised baseline. 1

What is social capital? In answering this question, the paper reports on new research which differentiates between the social interactive processes and social capital as the product of that process. Following a review of literature, structured as a social theory against which social capital might be understood, the paper then describes a study of a rural community, and reports on two analyses of data which contribute to answering the question, ‘What is the nature of the interactive productivity between the local networks in a community? ’ The paper concludes that social capital, for which a new definition is forwarded, can only ‘exist ’ if it is somehow able to be produced. This is the chief assumption of the paper. Social capital is a defined as an accumulation of the knowledge and identity resources drawn on by communities-of-common-purpose. If social capital originates in micro interactions which are in turn embedded in a macro social order, then these processes and connections should be observable. This paper makes an initial contribution to the establishment of such micro to macro links.

Retrieving information does not end with the identification of the top n relevant files. The identified text should then be presented to the user in a form that facilitates accessing the relevant information pieces. This work investigates the possibility of using lexical structures latent in the text to provide the reader with rich visual representations such as table-of-contents and topic index. The paper describes the approach for topic identification, reconstructing the hierarchical structure and the generation of sections' headings, as well as the back-end visualization system. 1 Introduction The cognitive process of reading expository text attempts to reconstruct the same framework of ideas the author had in mind when the text was originally written. The process is an exploratory one. The reader uses various devices as aids in this reconstruction, such as maps for orientation, or catalogs for objects being found in the process. If such devices are missing the process takes...

this paper were presented at City University of Hong Kong, the University of Melbourne, and the Australian National University. I would like to thank all those who participated in the discussions at those times, and I would also like to thank Sasha Aikenvald, Ruth Kempson, Steven Nicolle, Michael Pickering, Dan Sperber, and Elizabeth Traugott for sending me written comments. 2 and culture to be able to fill in the gaps in the perception of the utterance that are created by ambient noise. The inferential process becomes conscious when we are trying to understand an illegible word in a hand-written letter.

The primary aim of this paper is to address the methodological desiderata in obtaining a corpus of affected speech in human-machine interaction. We propose requirements that are to be met in order that a Wizard-of-Oz scenario designed to elicit affected speech could result in ecologically valid data. In addition, we report about the Wizard-of-Oz experiment conducted in the framework of the NIMITEK project. This project has the primary goal to investigate the role of emotions and intentions in human-machine dialogue. In the first phase of the experiment the focus was on prosody as a means to communicate emotional state. Keywords: human-machine interaction, affected speech, Wizard-of-Oz scenario, ecological validity. 1.

Abstract not found

The potentials of elearning are likely to be explored for many years to come in view of some of its obvious logistical benefits for learners and apparent financial incentives for institutions. One would hope that such exploration of teaching practice would have a parallel research program to investigate the nature and effects of dialogue and communities of enquiry in elearning in higher education. The common sense approach to this issue is to compare outcomes of elearning with those of classroom learning, but this approach has proven to be too simplistic to satisfy those attempting to understand the characteristics and potentials of elearning. As Garrison and Anderson (2003) put it, “Why would we expect to find significant differences if we do exactly the same thing [in the two modes of learning]…? ” (p. 6). They conceptualize the changes prompted by elearning as more radical than what can be captured through assessment of outcomes and comparisons with outcomes from classroom learning. At the core of the issue, in their view, is that education is about ideas not facts, and that elearning provides more than access to information; it affords opportunities for communication and interaction. But how does one assess how well learners are formulating ideas through