No description available.

No description available.

No description available.

The probabilistic (or quantitative) modal μ-calculus is a fixed-point logic designed for expressing properties of probabilistic labeled transition systems (PLTS’s). Two semantics have been studied for this logic, both assigning to every process state a value in the interval [0, 1] representing the probability that the property expressed by the formula holds at the state. One semantics is denotational and the other is a game semantics, specified in terms of two-player stochastic games. The two semantics have been proved to coincide on all finite PLTS’s. A first contribution of the thesis is to extend this coincidence result to arbitrary PLTS’s. A shortcoming of the probabilistic μ-calculus is the lack of expressiveness required to encode other important temporal logics for PLTS’s such as Probabilistic Computation Tree Logic (PCTL). To address this limitation, we extend the logic with a new pair of operators: independent product and coproduct, and we show that the resulting logic can encode the qualitative fragment of PCTL. Moreover, a further extension of the logic, with the operation of truncated sum and its dual, is expressive enough to encode full PCTL. A major contribution of the thesis is the definition of appropriate game semantics for these extended probabilistic μ-calculi. This relies on the definition of a new class of games, called tree games, which generalize standard 2-player stochastic games. In tree games, a play can be split into concurrent subplays which continue their evolution independently. Surprisingly, this simple device supports the encoding of the whole class of imperfect-information games known as Blackwell games. Moreover, interesting open problems in game theory, such as qualitative determinacy for 2-player stochastic parity games, can be reformulated as determinacy problems for suitable classes of tree games. Our main technical result about tree games is a proof of determinacy for 2-player stochastic metaparity games, which is the class of tree games that we use to give game semantics to the extended probabilistic μ-calculi. In order to cope with measure-theoretic technicalities, the proof is carried out in ZFC set theory extended with Martin’s Axiom at the first uncountable cardinal (MAℵ1). The final result of the thesis shows that the game semantics of the extended logics coincides with the denotational semantics, for arbitrary PLTS’s. However, in contrast to the earlier coincidence result, which is proved in ZFC, the proof of coincidence for the extended calculi is once again carried out in ZFC +MAℵ1.

Distributional models of semantics have proven themselves invaluable both in cognitive modelling of semantic phenomena and also in practical applications. For example, they have been used to model judgments of semantic similarity (McDonald, 2000) and association (Denhire and Lemaire, 2004; Griffiths et al., 2007) and have been shown to achieve human level performance on synonymy tests (Landuaer and Dumais, 1997; Griffiths et al., 2007) such as those included in the Test of English as Foreign Language (TOEFL). This ability has been put to practical use in automatic thesaurus extraction (Grefenstette, 1994). However, while there has been a considerable amount of research directed at the most effective ways of constructing representations for individual words, the representation of larger constructions, e.g., phrases and sentences, has received relatively little attention. In this thesis we examine this issue of how to compose meanings within distributional models of semantics to form representations of multi-word structures. Natural language data typically consists of such complex structures, rather than just individual isolated words. Thus, a model of composition, in which individual word meanings are combined into phrases and phrases combine to form sentences, is of central importance in modelling this data. Commonly, however, distributional representations are combined in terms of addition (Landuaer and Dumais, 1997; Foltz et al., 1998), without any empirical evaluation of alternative choices. Constructing effective distributional representations of phrases and sentences requires that we have both a theoretical foundation to direct the development of models of composition and also a means of empirically evaluating those models. The approach we take is to first consider the general properties of semantic composition and from that basis define a comprehensive framework in which to consider the composition of distributional representations. The framework subsumes existing proposals, such as addition and tensor products, but also allows us to define novel composition functions. We then show that the effectiveness of these models can be evaluated on three empirical tasks. The first of these tasks involves modelling similarity judgements for short phrases gathered in human experiments. Distributional representations of individual words are commonly evaluated on tasks based on their ability to model semantic similarity relations, e.g., synonymy or priming. Thus, it seems appropriate to evaluate phrase representations in a similar manner. We then apply compositional models to language modelling, demonstrating that the issue of composition has practical consequences, and also providing an evaluation based on large amounts of natural data. In our third task, we use these language models in an analysis of reading times from an eye-movement study. This allows us to investigate the relationship between the composition of distributional representations and the processes involved in comprehending phrases and sentences. We find that these tasks do indeed allow us to evaluate and differentiate the proposed composition functions and that the results show a reasonable consistency across tasks. In particular, a simple multiplicative model is best for a semantic space based on word co-occurrence, whereas an additive model is better for the topic based model we consider. More generally, employing compositional models to construct representations of multi-word structures typically yields improvements in performance over non-compositonal models, which only represent individual words.

The thesis presents a way to apply process algebra to computational linguistics. We are interested in how contexts can affect or contribute to language understanding and model the phenomena as a system of communicating processes to study the interaction between them in detail. For this purpose, we turn to the pie-calculus and investigate how communicating processes may be defined. While investigating the computational grounds of communication and concurrency,we devise a graphical representation for processes to capture the structure of interaction between them. Then, we develop a logic, combinatory intuitionistic linear logic with equality relation, to specify communicating processes logically. The development enables us to study Situation Semantics with process algebra. We construct semantic objects employed in Situation Semantics in the pi-calculus and then represent them in the logic. Through the construction,we also relate Situation Semantics with the research on the information flow, Channel Theory, by conceiving of linear logic as a theory of the information flow. To show how sentences can be parsed as the result of interactions between processes, we present a concurrent chart parser encoded in the pi-calculus. We also explain how a semantic representation can be generated as a process by the parser. We conclude the thesis by comparing the framework with other approaches.

No description available.

Many systems of Knowledge Representation exist, but none were designed specifically for general purpose large scale natural language processing. This thesis introduces a set of metrics to evaluate the suitability of representations for this purpose, derived from an analysis of the problems such processing introduces. These metrics address three broad categories of question: Is the representation sufficiently expressive to perform its task? What implications has its design on the architecture of the system using it? What inefficiencies are intrinsic to its design? An evaluation of existing Knowledge Representation systems reveals that none of them satisfies the needs of general purpose large scale natural language processing. To remedy this lack, this thesis develops a new representation: SemNet. SemNet benefits not only from the detailed requirements analysis but also from insights gained from its use as the core representation of the large scale general purpose system LOLITA (Large-scale Object-based Linguistic Interactor, Translator, and Analyser). The mapping process between Natural language and representation is presented in detail, showing that the representation achieves its goals in practice.

Much has been written on the Afrikaans verb system and verb categories. This, however, has been unconnected research which does not define the complete picture of the Afrikaans verb system. In research reports various categories of verbs are referred to comprehensively, while only five to six primary categories such as transitive, intransitive, copulative, prepositional and reflexive verbs, and to a lesser degree, "assistant" verbs, are referred to in some textbooks. In addition, research reports indicate that conventional identical verbs are sometimes named differently which raises the question whether the same verb can simultaneously belong to different verb categories. Sometimes the categorization results from a syntactical approach, while the naming of Afrikaans main verbs often result from a semantic perspective. In this study two different approaches of categorization are researched in order to establish which approach should apply as primary principle. The relevant approaches to categorization taken under scrutiny are the traditional or classic approach according tot which the limits of categories are relatively limited, and the prototype approach which provides for typical as well as peripheral categorical examples. The study aims to highlight the great number of Afrikaans verb categories and to place these in a semantic-syntactic framework. Two relevant models of categorization are taken as a basis to systemize the Afrikaans verb system. The method employed is to present an alphabetical exposition of the verb categories and then to establish within which semantic-cognitive image schemata such verbs are found. The part which metaphorical transfer, as principle of classification plays in the process of categorization is reviewed and illustrated. In addition, the semantic and syntactic characteristics of specific main verb categories are highlighted. From this study the following can be concluded: ▪ Afrikaans main verbs may simultaneously belong to different categories. Different 'categories' of main verbs are sometimes a single category with different names. ▪ Categorization from a syntactic perspective is systematic and simplistic and should be used as a point of departure. Verbs are firstly transitive, intransitive, copulative, prepositional, reflexive or "assistant" verbs, and may then be simultaneously connected to different semantic categories such as communicative verbs, verbs of weather conditions or process verbs. ▪ Categorization based only on semantic grounds provides a magnitude of categories which can be reduced by means of syntactic categorization. ▪ The traditional and prototype approaches to categorization do not exclude but complement each other and should be used interdependently as a categorical approach, the semantic subordinate to the syntactic. / Prof. A.E. Coetzee

In this thesis I investigate the relationship between the semantics of a verb and its selectional restrictions, which determine how many and what kind of arguments it must occur with in a clause. For most verbs, these restrictions are predictable from the semantics of the verb, but there are pairs of verbs with very similar semantics that differ in their argument restrictions. For example, both ask and wonder can take questions as their complements (John asked/wondered what time it was), but of the two, only ask can take a noun phrase complement with a question-like interpretation (John asked/*wondered the time). Similarly, while both eat and devour are verbs of consumption, the object can be omitted with eat but not devour (John ate/*devoured yesterday). Due to these and similar examples, many linguists have claimed that selectional restrictions are to some extent arbitrary and unpredictable from the semantics, and therefore must be learned as part of our knowledge of the relevant verbs. In this thesis I argue that these differences are not arbitrary; they recur across languages, and they can be predicted on the basis of lexical semantics, meaning they do not need to be learned on a word-by-word basis. In order for selectional features to be eliminated from the grammar, and replaced with semantic generalizations, two things must be shown. First, it must be demonstrated that the elements being selected for can be defined in terms of their semantics, rather than their syntactic properties. If not, the selectional properties could not be considered to be fully predictable based on the semantics of the selecting and selected items. Second, it must be shown that the selectional restrictions of a predicate are predictable from components of the selecting predicate’s meaning. In other words, the semantics of both the selected and the selecting elements must be accounted for. I focus mainly on the semantics of selected elements in Chapter 2, and on selecting elements in Chapters 3 and 4. Chapter 2 provides a brief review of the literature on selectional features, and argues that the elements being selected need not be defined in terms of their syntactic category and features. Instead, what are selected for are the semantic properties of the selected items. While the relationship between syntactic and semantic categories and properties is often systematic, it is not always, which can make it difficult in certain cases to determine the semantic basis for predicting what elements will be selected. Specifically, I argue that what appears to be selection for clausal categories (CPs or TPs) is in fact selection for propositional entities (including questions, assertions, facts, and so on); apparent selection for bare verb phrases (vPs) is selection for eventualities (events or states); and apparent selection for nominals (DPs) is selection for objects or things. Only properties of the nearest semantic entity (i.e., excluding elements embedded therein) can be selected for. In this way, I account for the selectional asymmetries between clausal and nominal complements noted by Bruening (2009) and Bruening et al. (2018): predicates selecting clausal complements can only select for (semantic) properties of the upper portion of the clause (in the CP domain), not for the lower portion (the vP domain), while predicates taking nominal complements can select for any properties of the nominal rather than being restricted to the upper portion. Since all syntactic properties of items are encoded as features, on a syntactic account it is expected that all features should be involved in selectional restrictions, contrary to fact; the semantic approach taken here allows for a principled explanation of what can and cannot be selected for. In Chapters 3 and 4 I turn to the lexical semantics of selecting elements, showing that these too are involved in determining selectional restrictions. I start in Chapter 3 by looking at c-selection (i.e., syntactic selection), specifically the case of eat versus devour. As mentioned above, their selectional properties of these two verbs differ in that the complement of eat is optional, while that of devour is obligatory, despite the two verbs having similar meanings. I show that this is due to the aspectual properties of these verbs: devour denotes an event where the complement necessarily undergoes a complete scalar change (i.e., it must be fully devoured by the end of the event), which means that the complement must be syntactically realized (Rappaport Hovav and Levin 2001; Rappaport Hovav 2008). Eat, on the other hand, does not entail a complete change of state in its complement, and so the complement is optional. I show that the correlation between scalar change entailments and obligatory argument realization holds for a wider group of verbs as well. Thus, the c-selectional properties of eat, devour, and similar verbs need not be stipulated in their lexical entries. In Chapter 4 I turn to the selection of complements headed by a particular lexical item, as with rely, which requires a PP complement headed by on, a phenomenon commonly referred to as l-selection. I show that the sets of verbs and prepositions involved in l-selection, and the observed verb-preposition combinations, are not fully random but can instead be (partially) predicted based on the thematic properties of the items in question. Furthermore, I show that there are different kinds of l-selecting predicates, and one kind is systematically present in satellite-framed languages (like English) and absent in verb-framed languages (like French), based on the Framing Typology of Talmy (1985, 1991, 2000). I account for this difference by analyzing l-selection as an instance of complex predicate formation, and showing that a certain kind of complex predicate (exemplified by rely on) is possible in satellite-framed languages but not in verb-framed languages. Thus, I show that the features that get selected for are semantic features, and that the problematic cases of eat versus devour and l-selection have semantic correlates, and need not be stipulated in the lexicon. While this leaves many instances of selectional features unaccounted for, it provides proposals for some components of lexical semantics that are relevant to selection, and demonstrates that a research program directed toward eliminating the remaining cases is plausibly viable.