published_or_final_version / English Studies / Master / Master of Arts

Natural Language Processing (NLP) develops computational approaches to processing language data. Supervised machine learning has become the dominant methodology of modern NLP. The performance of a supervised NLP system crucially depends on the amount of data available for training. In the standard supervised framework, if a sequence of words was not encountered in the training set, the system can only guess at its label at test time. The cost of producing labeled training examples is a bottleneck for current NLP technology. On the other hand, a vast quantity of unlabeled data is freely available. This dissertation proposes effective, efficient, versatile methodologies for 1) extracting useful information from very large (potentially web-scale) volumes of unlabeled data and 2) combining such information with standard supervised machine learning for NLP. We demonstrate novel ways to exploit unlabeled data, we scale these approaches to make use of all the text on the web, and we show improvements on a variety of challenging NLP tasks. This combination of learning from both labeled and unlabeled data is often referred to as semi-supervised learning. Although lacking manually-provided labels, the statistics of unlabeled patterns can often distinguish the correct label for an ambiguous test instance. In the first part of this dissertation, we propose to use the counts of unlabeled patterns as features in supervised classifiers, with these classifiers trained on varying amounts of labeled data. We propose a general approach for integrating information from multiple, overlapping sequences of context for lexical disambiguation problems. We also show how standard machine learning algorithms can be modified to incorporate a particular kind of prior knowledge: knowledge of effective weightings for count-based features. We also evaluate performance within and across domains for two generation and two analysis tasks, assessing the impact of combining web-scale counts with conventional features. In the second part of this dissertation, rather than using the aggregate statistics as features, we propose to use them to generate labeled training examples. By automatically labeling a large number of examples, we can train powerful discriminative models, leveraging fine-grained features of input words.

No description available.

It is widely accepted that Old English personal pronouns often turn up in ‘special’ positions, i.e. positions in which functionally equivalent nominals rarely, if ever, appear. Leading theories of Old English syntax (e.g. van Kemenade 1987, Pintzuk 1991, 1996, Hulk & van Kemenade 1997, Kroch & Taylor 1997) account for the syntax of specially placed pronouns in different ways, but all treat special placement as a freely available option. Focusing on pronominal objects of prepositions in particular, this thesis shows, firstly, that current theories fail to account for the variety of special positions in which these pronouns appear and argues that at least three special positions must be recognised. The central concern of this thesis, however, is whether special placement is the freely available option that leading theories assume. Drawing on evidence from a number of descriptive studies of the syntax of pronominal objects of prepositions (e.g. Wende 1915, Taylor 2008, Alcorn 2009), statistical evidence is presented to show that, in a number of contexts, the probability of special placement is either too high or else too low to be plausibly ascribed to free variation. The thesis explores the linguistic basis of each of the statistically significant parameters identified, finding answers in some cases and intriguing puzzles in others.

This thesis focuses on the issue of how language users refer to an entity during discourse production, by investigating representations and processes that underlie the choice between pronouns and repeated noun phrases. Past research has shown that the use of pronouns (relative to more explicit expressions) is affected by the referent’s salience in the prior linguistic context, but much less is known about how non-linguistic context affects the referent’s salience and the choice of expression. Recent research has suggested that the referent’s non-linguistic salience has no effect on the choice of pronouns and names (Arnold & Griffin, 2007). One of the major findings of the research reported in this thesis is that the referent's salience in the visual context plays an important role in the form of reference: Pronouns were less frequent (relative to repeated noun phrases) when the competitor was present than absent in the visual context. My second major finding is that similarity-based interference affects the choice of referring expressions. Pronouns are less frequent when discourse entities are similar in terms of their inherent conceptual properties as well as extrinsic properties, suggesting that the more similar the competitor to the referent, the stronger the interference, reducing pronoun usage. My third major finding is that contrary to many linguistic theories that assume that speakers choose referring expressions that are optimally helpful for their addressee (Ariel, 1990; Clark & Marshall, 1981; Givón, 1983), speakers do not choose expressions by adopting the addressee's discourse model: Pronouns are more frequent when the referent is salient to the speaker, not to the addressee. I argue that the explicitness of referring expressions is affected by the degree of conceptual access that is needed to initiate production processes: The more conceptual access is needed, the more elaborate expressions tend to be produced.

This essay outlines the problems associated with the history and current absence of a gender-neutral third-person singular personal pronoun in the English language. The combination of the social and grammatical consequences of this language gap results in pronoun choices that are either politically incorrect or verbose. Experts’ attempts to fill this language gap have failed to take root on any widespread basis; but, interestingly, middle school children in Baltimore, Maryland created and started using “yo” as their own gender-neutral third-person singular personal pronoun. Stotko and Troyer’s (2007) study on this development sheds some light on exactly how students use “yo” as a third-person pronoun and proposes some theories regarding the origin of this change in language. This spontaneously produced gender-neutral pronoun has gained as much recognition as many gender-neutral third-person singular personal pronouns proposed by linguistic experts, perhaps as a result of children’s unique understanding of and ability to create language. This recent development indicates that common English speakers will likely spontaneously generate a solution to the current pronoun gap, although this will probably take some time to occur.

No description available.

Syftet med denna studie har varit att undersöka hur förskolepersonal uppfattar en användning av det könsneutrala pronomenet hen inom förskolan. Våra frågeställningar är: ”Hur uppfattar förskolepersonalen möjligheten att använda hen som könsneutralt pronomen inom förskolan?” samt ”Anser förskolepersonalen att det finns några användningsområden där det könsneutrala pronomenet hen kan ingå?”. I denna undersökning användes en kvalitativ undersökningsmetod där vi utförde 11 kvalitativa forskningsintervjuer med förskolepersonal från fem förskolor. Den inspelade empirin dokumenterades med hjälp av diktafon som sedan transkriberades och analyserades. I studien framkom hur informanterna uppfattade genus och kön på förskolan. Vi tolkade att den uppfattningen speglade informanternas inställning till ordet hen. Undersökningen visade på i hur stor utsträckning förskolepersonalen använde sig av pronomen som hon/han. Vi tolkade att informanterna har utgått ifrån ett biologiskt och heteronormativt perspektiv när de talat om kön och könsidentitet. De uttryckte att de inte gjorde skillnad på "flickor" och "pojkar" och ansåg att de arbetade utifrån ett individperspektiv men istället påvisades motsatsen. Vår slutsats är att informanterna hade svårt att se användningsområden med ett könsneutralt pronomen. / The purpose of this study is to investigate how preschool employees perceive a usage of the gender-neutral pronoun hen within Swedish preschools. Our questions are: ”How do the preschool employees in this study perceive the possibility of using hen as a gender-neutral pronoun within the preschool environment?” and ”Do they think that there are any fields of usage of hen that can be applied in this environment?” This study was carried out with a qualitative method of investigation and 11 qualitative research interviews with preschool employees from five different preschools. The recorded empirical information was documented with a dictaphone whereupon the data was transcribed and analyzed. The study revealed how the informants perceived gender and sex in the preschool environment. We interpreted this perception as a reflection of the attitudes to the word hen among the informants. The results showed to what extent the preschool employees used pronouns like hon/han. We interpreted that the perspective among the informants when discussing gender and gender identity was biological and heteronormative. The informants claimed that they did not make any difference between girls and boys and they felt that they were working from an individual perspective. However, on the contrary, we proved the opposite. Our conclusion is that the informants found it difficult to see the range of application of the gender-neutral pronoun.

This thesis investigated how a third-born female child acquired the deictic meaning of feminine third-person pronouns in English. The child began producing feminine third-person pronouns at 24 months of age and made few production errors. In contrast, she made systematic comprehension errors between 24 and 36 months of age and did not master the correct comprehension until 40 months of age. Analysis of the child's person errors indicated that she held the proper name interpretation that the feminine third-person pronoun her referred to herself. In production, however, the child rarely called herself with feminine third-person pronouns because she had already mastered the correct use of first-person pronouns in self-reference. The issues of why the child made systematic person errors for such a long period of time and how she corrected the errors are discussed with regard to Oshima-Takane's (1985, 1998) pronoun-learning model.

This thesis investigates how children learn the first and the second person pronouns in English. In the first phase two cross sectional studies, which examined production and comprehension in children between 16 months and 36 months of ages, were conducted to determine what types of hypotheses children entertain about the semantic rules of the pronouns. In the second phase an intervention experiment was conducted to determine whether children benefit from observing speech not addressed to them for discovering the correct rules. This hypothesis was evaluated by comparing the effects of two different intervention programs: One providing children with opportunities to observe the shifting reference of personal pronouns in speech addressed to others and the other not providing such opportunities. The results suggest that even children under two years old can learn the correct rules of personal pronouns from speech not addressed to them.