Jon Oberlander

ILEX: an architecture for a dynamic hypertext generation system

Generating text in a hypermedia environment places different demands on a text generation system than occurs in non-interactive environments. This paper describes some of these demands, then shows how the architecture of one text generation system, ILEX, has been shaped by them. The architecture is described in terms of the levels of linguistic representation used, and the processes which map between them. Particular attention is paid to the processes of content selection and text structuring.

Language with Character: A Stratified Corpus Comparison of Individual Differences in E-Mail Communication.

To what extent does the wording and syntactic form of peoples writing reflect their personalities? Using a bottom-up stratified corpus comparison, rather than the top-down content analysis techniques that have been used before, we examine a corpus of e-mail messages elicited from individuals of known personality, as measured by the Eysenck Personality Questionnaire–Revised (S. Eysenck, Eysenck, & Barrett, 1985). This method allowed us to isolate linguistic features associated with different personality types, via both word and part-of-speech n-gram analysis. We investigated the extent to which extraversion is associated with linguistic features involving positivity, sociability, complexity, and implicitness and neuroticism is associated with negativity, self-concern, emphasis, and implicitness. Numerous interesting features were uncovered. For instance, higher levels of extraversion involved a preference for adjectives, whereas lower levels of neuroticism involved a preference for adverbs. However, neither positivity nor negativity was as prominent as expected, and there was little evidence for implicitness.

Contrasting the cognitive effects of graphical and sentential logic teaching: Reasoning, representation and individual differences

Abstract Hyperproof is a computer program created by Barwise and Etchemendy (1994) for teaching logic using multimodal graphical and sententid representations. Elsewhere, we have proposed a theory of the cognitive impact of assigning information to different modalities. The theory predicts that Hyperproofs devices for graphical abstraction will play a pivotal role in determining learning outcomes. Here, the claims are tested by a controlled comparison of the effects of teaching undergraduate classes using Hyperproof and a traditional syntactic teaching method. The results indicate that there is significant transfer from the logic courses to a range of verbal reasoning problems. There are also significant interactions between theoretically motivated precourse aptitude measures and teaching methods, and these interactions influence post-course reasoning performance in transfer domains. As well as being theoretically significant, the results provide support for the important practical conclusion that indivi...

Conversation in the museum: experiments in dynamic hypermedia with the intelligent labelling explorer

We outline experience with the Intelligent Labelling Explorer, a dynamic hypertext system developed at the University of Edinburgh, in collaboration with the National Museums of Scotland. First, we indicate a number of ways in which labels on museum objects ought to be tuned to take into account types of visit, the interests of visitors, and their evolving knowledge during a visit. Secondly, we sketch the general architecture of our system, and then focus on the conversational effects which the system can create. We then briefly indicate future directions of research, before critically discussing the applicability (or otherwise) of the spatial metaphor to flexible hypertexts.

The language of emotion in short blog texts

Emotion is central to human interactions, and automatic detection could enhance our experience with technologies. We investigate the linguistic expression of fine-grained emotion in 50 and 200 word samples of real blog texts previously coded by expert and naive raters. Content analysis (LIWC) reveals angry authors use more affective language and negative affect words, and that joyful authors use more positive affect words. Additionally, a co-occurrence semantic space approach (LSA) was able to identify fear (which naive human emotion raters could not do). We relate our findings to human emotion perception and note potential computational applications.

Interpreting pronouns and connectives: Interactions among focusing, thematic roles and coherence relations

This paper investigates the relationship between focusing and coherence relations in pronoun comprehension. In their focusing model of pronoun comprehension, Stevenson, Crawley and Kleinman (1994) proposed a default focus on the thematic role associated with the consequences of a described event, a focus that may be modified by the attention-directing properties of a subsequent connective. In this paper we examine a second function of connectives: that of signalling the coherence relations between two clauses (e.g., a NARRATIVE relation or a RESULT relation). In three studies, we identified the coherence relations between sentence fragments ending in pronouns and participants’ continuations to the fragments. We then examined the relationship between the coherence relation, the preferred referent of the pronoun and the referent’s thematic role. The results of studies 1 and 2 showed that people aim to keep the focused entity, the coherence relation and the referent of the pronoun in alignment. Study 3 included the connective next, which enabled us to generate different predictions for the roles of focusing and coherence relations in pronoun resolution. The results favoured the focusing view. The preferred referent of the pronoun was the focused, first mentioned, individual, whereas the coherence relation was consistent with the thematic role of the pronominal referent. If the pronoun referred to an Agent, a NARRATIVE relation was preferred, if the pronoun referred to a Patient, a RESULT relation was preferred. Discussion of these and other results led to the following conclusions. First, pronoun resolution is primarily determined by focusing, either semantic or structural, although a range of other features, including coherence relations and verb semantics, may also act as pressures on pronoun resolution. Second, the consistent link we observed between thematic roles and coherence relations may provide a mapping between a represented entity and a represented event. Third, the connectives we used have three distinct functions: an attention directing function, a function for constraining the possible coherence relation between two events, and a function for interpreting a clause as having either a causal or a temporal structure.

The roles of haptic-ostensive referring expressions in cooperative, task-based human-robot dialogue

Generating referring expressions is a task that has received a great deal of attention in the natural-language generation community, with an increasing amount of recent effort targeted at the generation of multimodal referring expressions. However, most implemented systems tend to assume very little shared knowledge between the speaker and the hearer, and therefore must generate fully-elaborated linguistic references. Some systems do include a representation of the physical context or the dialogue context; however, other sources of contextual information are not normally used. Also, the generated references normally consist only of language and, possibly, deictic pointing gestures. When referring to objects in the context of a task-based interaction involving jointly manipulating objects, a much richer notion of context is available, which permits a wider range of referring options. In particular, when conversational partners cooperate on a mutual task in a shared environment, objects can be made accessible simply by manipulating them as part of the task. We demonstrate that such expressions are common in a corpus of human-human dialogues based on constructing virtual objects, and then describe how this type of reference can be incorporated into the output of a humanoid robot that engages in similar joint construction dialogues with a human partner.

Speaking the users' languages

The authors describe a system that generates descriptions of museum objects tailored to the user. The texts presented to adults, children, and experts differ in several ways, from the choice of words used to the complexity of the sentence forms. M-PIRO can currently generate text in three languages: English, Greek, and Italian. The grammar resources are language independent as much as possible. M-PIROs system architecture is significantly more modular than that of its predecessor ILEX. In particular, the linguistic resources, database, and user-modeling subsystems are now separate from the systems that perform the natural language generation and speech synthesis.

Large scale personality classification of bloggers

Personality is a fundamental component of an individuals affective behavior. Previous work on personality classification has emerged from disparate sources: Varieties of algorithms and feature-selection across spoken and written data have made comparison difficult. Here, we use a large corpus of blogs to compare classification feature selection; we also use these results to identify characteristic language information relating to personality. Using Support Vector Machines, the best accuracies range from 84.36% (openness to experience) to 70.51% (neuroticism). To achieve these results, the best performing features were a combination of: (1) stemmed bigrams; (2) no exclusion of stopwords (i.e. common words); and (3) the boolean, presence or absence of features noted, rather than their rate of use. We take these findings to suggest that both the structure of the text and the presence of common words are important. We also note that a common dictionary of words used for content analysis (LIWC) performs less well in this classification task, which we propose is due to their conceptual breadth. To get a better sense of how personality is expressed in the blogs, we explore the best performing features and discuss how these can provide a deeper understanding of personality language behavior online.

Abducing Temporal Discourse

We focus on the following question: given the causal and temporal relations between events in a knowledge base, what are the ways they can be described in extended text? We argue that we want to be able to generate laconic text, where certain temporal information remains implicit but pragmatically inferrable. An algorithm for generating laconic text is proposed, interleaving abduction and nonmonotonic deduction over a formal model of pragmatic implicature. We demonstrate that the nonmonotonicity ensures that the generation of laconic text is influenced by the preceding linguistic and extra-linguistic context.