ERA
https://era.ed.ac.uk:443
The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material.2021-05-08T10:17:23ZUnsupervised Learning of Relational Entailment Graphs from Text
https://hdl.handle.net/1842/37596
Unsupervised Learning of Relational Entailment Graphs from Text
Hosseini, Mohammad Javad
Recognizing textual entailment and paraphrasing is critical to many core natural language processing applications including question answering and semantic parsing. The surface form of a sentence that answers a question such as “Does Facebook own Instagram?” frequently does not directly correspond to the form of the question, but is rather a paraphrase or an expression such as “Facebook bought Instagram”, that entails the answer. Relational entailments (e.g., buys entails owns) are crucial for bridging the gap between queries and text resources. In this thesis, we describe different unsupervised approaches to construct relational entailment graphs, with typed relations (e.g., company buys company) as nodes and entailment as directed edges. The entailment graphs provide an explainable resource for downstream tasks such as question answering; however, the existing methods suffer from noise and sparsity inherent to the data.
We extract predicate-argument structures from large multiple-source news corpora using a fast Combinatory Categorial Grammar parser. We compute entailment scores between relations based on the Distributional Inclusion Hypothesis which states that a word (relation) p entails another word (relation) q if and only if in any context that p can be used, q can be used in its place. The entailment scores are used to build local entailment graphs. We then build global entailment graphs by exploiting the dependencies between the entailment rules. Previous work has used transitivity constraints, but these constraints are intractable on large graphs. We instead propose a scalable method that learns globally consistent similarity scores based on new soft constraints that consider both the structures across typed entailment graphs and inside each graph. We show that our method significantly improves the entailment graphs.
Additionally, we show the duality of entailment graph induction with the task of link prediction. The link prediction task infers missing relations between entities in an incomplete knowledge graph and discovers new facts. We present a new method in which link prediction on the knowledge graph of assertions extracted from raw text is used to improve entailment graphs which are learned from the same text. The entailment graphs are in turn used to improve the link prediction task.
Finally, we define the contextual link prediction task that uses both the structure of the knowledge graph of assertions and their textual contexts. We fine-tune pre-trained language models with an unsupervised contextual link prediction objective. We augment the existing assertions with novel predictions of our model and use them to build higher quality entailment graphs. Similarly, we show that the entailment graphs improve the contextual link prediction task.
2021-07-31T00:00:00ZThe Stability of Hyperbolic PDEs in String Theory, Particle Physics and Cosmology
https://hdl.handle.net/1842/37595
The Stability of Hyperbolic PDEs in String Theory, Particle Physics and Cosmology
Wyatt, Zoe
In this thesis we study hyperbolic PDEs arising from general relativity and the standard model of particle physics. In particular we prove the asymptotic stability of special solutions of these PDEs against small initial data perturbations. The study of stability elucidates our understanding of whether such PDEs can provide mathematically reasonable models for physical phenomena in our universe.
In the first chapter, we prove the stability of a system of quasilinear wave equations satisfying the weak null condition. In particular, we prove that the Kaluza-Klein spacetime, the cartesian product of Minkowski spacetime with a circle, viewed as a solution to the vacuum Einstein equations, is stable to circle-independent perturbations. In the second chapter, we show that the Milne spacetime is a stable solution to the Einstein-Klein-Gordon equations. We upgrade a technique that uses the continuity equation complementary to L^2 estimates to control massive matter fields. In contrast to earlier applications of this idea we require a correction to the energy density to obtain sufficiently strong pointwise bounds. In the third chapter, we use the hyperboloidal foliation method to study an interesting PDE system relevant in particular physics. In particular we establish the stability of the ground state of the U(1) standard model of electroweak interactions. In particular, we investigate here the Dirac equation and consider a new energy functional for this field defined with respect to the hyperboloidal foliation of Minkowski spacetime. We provide a novel decay result for the Dirac equation which is uniform in the mass coefficient, and thus allows for the Dirac mass coefficient to be arbitrarily small. In the final chapter, we bring together ideas developed in the first three chapters and prove the stability, with respect to the evolution determined by the vacuum Einstein equations, of the Cartesian product of high-dimensional Minkowski space with a compact Riemannian manifold admitting nonzero parallel spinors. Such a product includes the example of special holonomy compactifications, which play a central role in string theory.
2020-09-01T00:00:00Z‘If it weren’t for YouTube, I wouldn’t be here’ Exploring Indigenous Perspectives on Online Learning
https://hdl.handle.net/1842/37594
‘If it weren’t for YouTube, I wouldn’t be here’ Exploring Indigenous Perspectives on Online Learning
Gniadek, Iwona
In this study, I explored the voices of Indigenous students and teachers, who, despite the Truth and Reconciliation Commission’s Calls to Action (2015), continue to be underrepresented in online learning at post-secondary institutions in Canada. This dissertation draws upon the narrative inquiry methodology to explore accounts of lived experiences through in-depth interviews and Sharing Circles. Data analysis included contextualisation and an inductive thematic analysis, which yielded seven themes: 1. Widening participation, 2. Building capacity, 3. Digital inequity, 4. Text-based dominance, 5. Impersonal environment, 6. Role and impact of peers, 7. Teacher presence. Overall, findings reveal that despite the potential and benefits of online learning for Indigenous students, barriers persist. Efforts must be increased to design and teach online courses in decolonizing ways. Nine tentative design recommendations are proposed.
2021-07-31T00:00:00ZThe Effect of Semantic Constraint on Lexical Access in Bilingual Word Recognition
https://hdl.handle.net/1842/37593
The Effect of Semantic Constraint on Lexical Access in Bilingual Word Recognition
Winther, Irene
The current study investigated how a constraining sentence context affects processing times in second language (L2) word identification. We used eye-tracking to look at whether the cognate facilitation effect, a cue of non-selectiveness in bilingual lexical access, is affected by the presence of a strong semantical sentence context. Norwegian-English bilinguals read sentences containing cognates or matched controls in sentences providing either a high constraining or a low constraining context. We found cognate facilitation effects for high constraining sentences for gaze durations, but none of the other eye-tracking measures. This supports a theory of bilingual non-selective lexical access, which can vary in degree based on different factors. We discuss our results in context of the BIA+ model (Dijkstra & van Heuven, 2002).
2017-10-13T00:00:00Z