No description available.

No description available.

Natural language understanding is a sub-field of natural language processing, which builds automated systems to understand natural language. It is such an ambitious task that it sometimes is referred to as an AI-complete problem, implying that its difficulty is equivalent to solving the central artificial intelligence problem -- making computers as intelligent as people. Despite its complexity, natural language understanding continues to be a fundamental problem in natural language processing in terms of its theoretical and empirical importance. In recent years, startling progress has been made at different levels of natural language processing tasks, which provides great opportunity for deeper natural language understanding. In this thesis, we focus on the task of semantic parsing, which maps a natural language sentence into a complete, formal meaning representation in a meaning representation language. We present two novel state-of-the-art learned syntax-based semantic parsers using statistical syntactic parsing techniques, motivated by the following two reasons. First, the syntax-based semantic parsing is theoretically well-founded in computational semantics. Second, adopting a syntax-based approach allows us to directly leverage the enormous progress made in statistical syntactic parsing. The first semantic parser, Scissor, adopts an integrated syntactic-semantic parsing approach, in which a statistical syntactic parser is augmented with semantic parameters to produce a semantically-augmented parse tree (SAPT). This integrated approach allows both syntactic and semantic information to be available during parsing time to obtain an accurate combined syntactic-semantic analysis. The performance of Scissor is further improved by using discriminative reranking for incorporating non-local features. The second semantic parser, SynSem, exploits an existing syntactic parser to produce disambiguated parse trees that drive the compositional semantic interpretation. This pipeline approach allows semantic parsing to conveniently leverage the most recent progress in statistical syntactic parsing. We report experimental results on two real applications: an interpreter for coaching instructions in robotic soccer and a natural-language database interface, showing that the improvement of Scissor and SynSem over other systems is mainly on long sentences, where the knowledge of syntax given in the form of annotated SAPTs or syntactic parses from an existing parser helps semantic composition. SynSem also significantly improves results with limited training data, and is shown to be robust to syntactic errors. / text

No description available.

No description available.

This thesis is primarily concerned with language production. In particular it investigates two issues: First, it explores some of the processing mechanisms underlying the production of different syntactic structures and word orders. Second, it explores the production of different syntactic structures and word order from a developmental perspective. These two issues are investigated experimentally and from a cross-linguistic point of view. First, a description is given of the possible word order permutations that Catalan allows and under which circumstances these word orders are produced. This is extended with a corpus analysis of spoken Catalan. The aim of this study is twofold: on the one hand, it aims to present the different positions where subjects and complements of the verb can appear in a sentence. On the other, it aims to compare the use of passivization between spoken and newspaper text in Catalan. Second, my experimental work in language production in four languages is presented. These languages include English, Brazilian Portuguese, Catalan and Spanish. The main aim of this study is to explore the effects of the non-linguistic factors of animacy and frequency upon the production of different word orders. The results of four experiments in the four languages mentioned yield evidence that these non-linguistic factors affect the on-line processing of language production. In the four languages, participants tend to prefer to produce syntactic structures which allow animate entities to be realised as the sentential subject, even if this means producing a passive structure rather than a (usually preferred) active structure. I have also found evidence that in some languages (e.g. Catalan and Spanish) animate/frequent entities appear at initial sentence position in the grammatical category of object (in dislocated active constructions). These results are explained on the light of some of the models of language production (e.g. Bock 1987a; Bock and Levelt 1994). Third, further cross-linguistic experiments in three languages (English, Catalan and Spanish) are presented. There I show that one particular contextual factor, discourse salience, can also affect the realisation of different syntactic structures during production. Entities which have been made more salient by the preceding context are more likely to appear as sentential subjects or in early sentential positions than entities which have also been introduced in previous discourse but are less salient. I suggest that these effects can be explained using the same mechanisms that explain other non-linguistic factors (e.g. animacy). The results also suggest that in the absence of context, animacy is a strong determinant of syntactic structure and word order, whereas in context, discourse salience may largely override animacy effects. Finally, these results suggest that from a processing point of view, the Given/New partition is not enough to account for the information structure of a sentence, but a more fine-grained distinction is need, in keeping with some recent pragmatic theories (e.g. Prince 1981, 1992; Sgall et al. 1986). Finally, I investigate the production of different word orders from a developmental point of view. In particular I examine the relationship between age and the production of different word orders by Catalan children, ranging from 4;11 to 11;11 years. The results of an experiment run with these children show that a dislocated active is a construction already consolidated at age 5. In contrast, the passive clause is a construction still not fully acquired at age 11. These results seem to suggest that for Catalan children, a dislocated active is a syntactic structure that is available earlier than the passive structure. Conversely, the placement of a patient in subject position and the creation of a verbal passive voice occurs later than simple word order permutation. Finally, a comparison between these results and existing results from English children shows that there are cross-linguistic differences on the age of production of passive clauses: while English children already produce passives at age 5, Catalan children start producing passives at age 11. I suggest some possible explanations for the cross-linguistic differences in the production of different syntactic structures. Overall, the main aim for this study is to gain insight into the production of different syntactic structures and word orders from a psycholinguistic and developmental point of view.

Thesis (Ed.D.)--Boston University / PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you. / This study investigated the concept of syntactic awareness as a reading strategy for complex sentence comprehension and the relationship between instruction in syntactic awareness and improved reading comprehension among mainstream high school students. When given the insight that sentences are important and when given simple rules to map syntactic structures onto thematic roles, with discussion and practice in simplification and restructuring of complex sentences, 91h and lth grade students demonstrated that they could significantly improve their reading comprehension abilities. The results support the concept that many high school students default to simple reading heuristics which work well on canonical sentence types, but which can fail with complex content-area texts. Sixty-three eleventh grade and forty-seven ninth grade mainstream students participated in a study in which one group of ninth and one group of eleventh graders, the experimental groups, were given a sentence comprehension strategy to help them when confronted with text that they block on, in many cases typical texts that are part of the high school curriculum. The control groups were given regular instruction in reading comprehension skills and strategies. Eleventh grade students who were enrolled in a SAT Preparation elective class were pre-tested and post-tested using both Real SAT tests from The College Board and the Stanford Diagnostic Reading Test (SDRT). A third group, the maturation group, was pre-tested and post-tested using the SDRT to control for normal academic growth. Ninth grade students who were enrolled in a reading elective course were pre-tested and post-tested using the SDRT. For the Eleventh Grade Experiment results were significant at the .01 level for the SDRT and at the .01 level for the SAT. The difference between the maturation group and the intervention group was significant at the .01 level but was not significant between the control and maturation group. The Ninth Grade Experiment results were significant at the .05 level for the SDRT. / 2031-01-02

No description available.

No description available.

This thesis takes as its starting point the proposal in Kayne (1994) that all syntactic structures are underlyingly spec-head-complement, and that they are right-branching. I will investigate this proposal taking data from English degree constructions, namely result clauses and comparatives. A comparison will be made between these constructions and English VPs, on which the majority of the phrase structure debate in the literature has been based. The evidence for left-branching and for right-branching in VPs will be considered, and similar evidence sought for degree constructions. We will see that VPs have a mostly right-branching structure, although left-branching structures are required in restricted circumstances. Also reason and manner adjuncts are argued to be right-adjoined to the VP node, a conclusion that is re-inforced by considering the constituency of VP adjuncts and some PP sequences noted by Jackendoff (1973). In degree constructions too, we argue that both left-branching and right-branching structures are necessary. My conclusion will be that Kayne's proposal is too strong, even though it is ideal from the perspective of a minimalist approach to syntax.