arXiv subjects

David Weir

Publications and source records attributed to David Weir.

At least 19 recordsLinked to original sources

arXiv2024

Primordial acoustic turbulence: three-dimensional simulations and gravitational wave predictions

Gravitational waves (GWs) generated by a first-order phase transition at the electroweak scale are detectable by future space-based detectors like LISA. The lifetime of the resulting shock waves plays an important role in determining the intensity of the generated GWs. We have simulated decaying primordial acoustic turbulence in three dimensions and make a prediction for the universal shape of the energy spectrum by using its self-similar decay properties and the shape of individual shock waves. The shape for the spectrum is used to determine the time dependence of the fluid kinetic energy and the energy containing length scale at late times. The inertial range power law is found to be close to the classically predicted $k^{-2}$ and approaches it with increasing Reynolds number. The resulting model for the velocity spectrum and its decay in time is combined with the sound shell model assumptions about the correlations of the velocity field to compute the GW power spectrum for flows that decay in less than the Hubble time. The decay is found to bring about a convergence in the spectral amplitude and the peak power law that leads to a power law shallower than the $k^9$ of the stationary case.

gr-qc↗

arXiv2023

Automated Query Generation for Evidence Collection from Web Search Engines

It is widely accepted that so-called facts can be checked by searching for information on the Internet. This process requires a fact-checker to formulate a search query based on the fact and to present it to a search engine. Then, relevant and believable passages need to be identified in the search results before a decision is made. This process is carried out by sub-editors at many news and media organisations on a daily basis. Here, we ask the question as to whether it is possible to automate the first step, that of query generation. Can we automatically formulate search queries based on factual statements which are similar to those formulated by human experts? Here, we consider similarity both in terms of textual similarity and with respect to relevant documents being returned by a search engine. First, we introduce a moderate-sized evidence collection dataset which includes 390 factual statements together with associated human-generated search queries and search results. Then, we investigate generating queries using a number of rule-based and automatic text generation methods based on pre-trained large language models (LLMs). We show that these methods have different merits and propose a hybrid approach which has superior performance in practice.

cs.CL↗

arXiv2022

Towards Structure-aware Paraphrase Identification with Phrase Alignment Using Sentence Encoders

Previous works have demonstrated the effectiveness of utilising pre-trained sentence encoders based on their sentence representations for meaning comparison tasks. Though such representations are shown to capture hidden syntax structures, the direct similarity comparison between them exhibits weak sensitivity to word order and structural differences in given sentences. A single similarity score further makes the comparison process hard to interpret. Therefore, we here propose to combine sentence encoders with an alignment component by representing each sentence as a list of predicate-argument spans (where their span representations are derived from sentence encoders), and decomposing the sentence-level meaning comparison into the alignment between their spans for paraphrase identification tasks. Empirical results show that the alignment component brings in both improved performance and interpretability for various sentence encoders. After closer investigation, the proposed approach indicates increased sensitivity to structural difference and enhanced ability to distinguish non-paraphrases with high lexical overlap.

cs.CL↗

arXiv2021

Decay of acoustic turbulence in two dimensions and implications for cosmological gravitational waves

Gravitational waves from a phase transition associated with the generation of the masses of elementary particles are within the reach of future space-based detectors such as LISA. A key determinant of the resulting power spectrum, not previously studied, is the lifetime of the acoustic turbulence which follows. We study decaying acoustic turbulence using numerical simulations of a relativistic fluid in two dimensions. Working in the limit of non-relativistic bulk velocities, with an ultra-relativistic equation of state, we find that the energy spectrum evolves towards a self-similar broken power law, with a high-wavenumber behaviour of $k^{-2.08 \pm 0.08}$, cut off at very high $k$ by the inverse width of the shock waves. Our model for the decay of acoustic turbulence can be extended to three dimensions using the universality of the high-$k$ power law and the evolution laws for the kinetic energy and the integral length scale. It is used to build an estimate for the gravitational wave power spectrum resulting from a collection of shock waves, as might be found in the aftermath of a strong first order phase transition in the early universe. The power spectrum has a peak wavenumber set by the initial length scale of the acoustic waves, and a new secondary scale at a lower wavenumber set by the integral scale after a Hubble time. Between these scales a distinctive new power law appears. Our results allow more accurate predictions of the gravitational wave power spectrum for a wide range of early universe phase transition scenarios.

gr-qc↗

arXiv2021

Legacy of the First Workshop on Gravitational Wave Astrophysics for Early Career Scientists

Gravitational wave science is a dynamical, fast-expanding research field founded on results, tools and methodologies drawn from different research areas and communities. Early career scientists entering this field must learn and combine knowledge and techniques from a range of disciplines. The Workshop on Gravitational-Wave Astrophysics for Early Career Scientists (GWAECS), held virtually in May 2021, planted the seeds of an interdisciplinary, well-connected and all-inclusive community of early career scientists working on gravitational waves, able to exchange relevant information and ideas, build a healthy professional and international environment, share and learn valuable skills, and ensure that ongoing research efforts are perpetuated and expanded in order to attain the main scientific goals envisioned by the whole community. GWAECS was the first event unifying early career scientists belonging to different communities, historically associated with different large-scale gravitational wave experiments. It provided a broad perspective on the future of gravitational waves, offered training on soft and transferable skills and allowed ample time for informal discussions between early career scientists and well-known research experts. The essence of those activities is summarised and collected in the present document, which presents a recap of each session of the workshop and aims to provide all early career scientists with a long-lasting, useful reference which constitutes the legacy of all the ideas that circulated at GWAECS.

gr-qc↗

arXiv2021

Infrared physics of the $\mathrm{SU}(2)$ Georgi-Glashow crossover transition

We perform a lattice study of the phase transition in the $\mathrm{SU}(2)$ Georgi-Glashow model in three dimensions, where the gauge symmetry is broken to $\mathrm{U}(1)$ by the Higgs mechanism and a photon-like state appears. Due to condensation of monopoles the photon acquires a mass, which depends on the number density of the monopoles. We show that the monopole density can be renormalised on the lattice using gradient flow. Our preliminary results suggest that Polyakov's semiclassical relation between the photon mass and the renormalised monopole density is valid also at the nonperturbative level.

hep-lat↗

arXiv2021

The Effect of Mission Duration on LISA Science Objectives

The science objectives of the LISA mission have been defined under the implicit assumption of a 4 yr continuous data stream. Based on the performance of LISA Pathfinder, it is now expected that LISA will have a duty cycle of $\approx 0.75$, which would reduce the effective span of usable data to 3 yr. This paper reports the results of a study by the LISA Science Group, which was charged with assessing the additional science return of increasing the mission lifetime. We explore various observational scenarios to assess the impact of mission duration on the main science objectives of the mission. We find that the science investigations most affected by mission duration concern the search for seed black holes at cosmic dawn, as well as the study of stellar-origin black holes and of their formation channels via multi-band and multi-messenger observations. We conclude that an extension to 6 yr of mission operations is recommended.

astro-ph.IM↗

arXiv2021

Vacuum bubble collisions: from microphysics to gravitational waves

We comprehensively study the effects of bubble wall thickness and speed on the gravitational wave emission spectrum of collisions of two vacuum bubbles. We numerically simulate a large dynamical range, making use of symmetry to reduce the dimensionality. The high-frequency slope of the gravitational wave spectrum is shown to depend on the thickness of the bubble wall, becoming steeper for thick-wall bubbles, in agreement with recent fully 3+1 dimensional lattice simulations of many-bubble collisions. This dependence is present, even for highly relativistic bubble wall collisions. We use the reduced dimensionality as an opportunity to investigate dynamical phenomena which may underlie the observed differences in the gravitational wave spectra. These phenomena include `trapping', which occurs most for thin-wall bubbles, and oscillations behind the bubble wall, which occur for thick-wall bubbles.

astro-ph.CO↗

arXiv2021

Representing Syntax and Composition with Geometric Transformations

The exploitation of syntactic graphs (SyGs) as a word's context has been shown to be beneficial for distributional semantic models (DSMs), both at the level of individual word representations and in deriving phrasal representations via composition. However, notwithstanding the potential performance benefit, the syntactically-aware DSMs proposed to date have huge numbers of parameters (compared to conventional DSMs) and suffer from data sparsity. Furthermore, the encoding of the SyG links (i.e., the syntactic relations) has been largely limited to linear maps. The knowledge graphs' literature, on the other hand, has proposed light-weight models employing different geometric transformations (GTs) to encode edges in a knowledge graph (KG). Our work explores the possibility of adopting this family of models to encode SyGs. Furthermore, we investigate which GT better encodes syntactic relations, so that these representations can be used to enhance phrase-level composition via syntactic contextualisation.

cs.CL↗

arXiv2020

Data Augmentation for Hypernymy Detection

The automatic detection of hypernymy relationships represents a challenging problem in NLP. The successful application of state-of-the-art supervised approaches using distributed representations has generally been impeded by the limited availability of high quality training data. We have developed two novel data augmentation techniques which generate new training examples from existing ones. First, we combine the linguistic principles of hypernym transitivity and intersective modifier-noun composition to generate additional pairs of vectors, such as "small dog - dog" or "small dog - animal", for which a hypernymy relationship can be assumed. Second, we use generative adversarial networks (GANs) to generate pairs of vectors for which the hypernymy relation can also be assumed. We furthermore present two complementary strategies for extending an existing dataset by leveraging linguistic resources such as WordNet. Using an evaluation across 3 different datasets for hypernymy detection and 2 different vector spaces, we demonstrate that both of the proposed automatic data augmentation and dataset extension strategies substantially improve classifier performance.

cs.CL↗

arXiv2017

Improving Semantic Composition with Offset Inference

Count-based distributional semantic models suffer from sparsity due to unobserved but plausible co-occurrences in any text collection. This problem is amplified for models like Anchored Packed Trees (APTs), that take the grammatical type of a co-occurrence into account. We therefore introduce a novel form of distributional inference that exploits the rich type structure in APTs and infers missing data by the same mechanism that is used for semantic composition.

cs.CL↗

arXiv2017

One Representation per Word - Does it make Sense for Composition?

In this paper, we investigate whether an a priori disambiguation of word senses is strictly necessary or whether the meaning of a word in context can be disambiguated through composition alone. We evaluate the performance of off-the-shelf single-vector and multi-sense vector models on a benchmark phrase similarity task and a novel task for word-sense discrimination. We find that single-sense vector models perform as well or better than multi-sense vector models despite arguably less clean elementary representations. Our findings furthermore show that simple composition functions such as pointwise addition are able to recover sense specific information from a single-sense vector model remarkably well.

cs.CL↗

arXiv2016

Aligning Packed Dependency Trees: a theory of composition for distributional semantics

We present a new framework for compositional distributional semantics in which the distributional contexts of lexemes are expressed in terms of anchored packed dependency trees. We show that these structures have the potential to capture the full sentential contexts of a lexeme and provide a uniform basis for the composition of distributional knowledge in a way that captures both mutual disambiguation and generalization.

cs.CL↗

arXiv2016

Improving Sparse Word Representations with Distributional Inference for Semantic Composition

Distributional models are derived from co-occurrences in a corpus, where only a small proportion of all possible plausible co-occurrences will be observed. This results in a very sparse vector space, requiring a mechanism for inferring missing knowledge. Most methods face this challenge in ways that render the resulting word representations uninterpretable, with the consequence that semantic composition becomes hard to model. In this paper we explore an alternative which involves explicitly inferring unobserved co-occurrences using the distributional neighbourhood. We show that distributional inference improves sparse word representations on several word similarity benchmarks and demonstrate that our model is competitive with the state-of-the-art for adjective-noun, noun-noun and verb-object compositions while being fully interpretable.

cs.CL↗

arXiv2015

Gravitational waves from cosmological first order phase transitions

First order phase transitions in the early Universe generate gravitational waves, which may be observable in future space-based gravitational wave observatiories, e.g. the European eLISA satellite constellation. The gravitational waves provide an unprecedented direct view of the Universe at the time of their creation. We study the generation of the gravitational waves during a first order phase transition using large-scale simulations of a model consisting of relativistic fluid and an order parameter field. We observe that the dominant source of gravitational waves is the sound generated by the transition, resulting in considerably stronger radiation than earlier calculations have indicated.

hep-lat↗

arXiv1997

Encoding Frequency Information in Lexicalized Grammars

We address the issue of how to associate frequency information with lexicalized grammar formalisms, using Lexicalized Tree Adjoining Grammar as a representative framework. We consider systematically a number of alternative probabilistic frameworks, evaluating their adequacy from both a theoretical and empirical perspective using data from existing large treebanks. We also propose three orthogonal approaches for backing off probability estimates to cope with the large number of parameters involved.

cmp-lg↗

arXiv1995

Encoding Lexicalized Tree Adjoining Grammars with a Nonmonotonic Inheritance Hierarchy

This paper shows how DATR, a widely used formal language for lexical knowledge representation, can be used to define an LTAG lexicon as an inheritance hierarchy with internal lexical rules. A bottom-up featural encoding is used for LTAG trees and this allows lexical rules to be implemented as covariation constraints within feature structures. Such an approach eliminates the considerable redundancy otherwise associated with an LTAG lexicon.

cmp-lg↗

arXiv1995

D-Tree Grammars

DTG are designed to share some of the advantages of TAG while overcoming some of its limitations. DTG involve two composition operations called subsertion and sister-adjunction. The most distinctive feature of DTG is that, unlike TAG, there is complete uniformity in the way that the two DTG operations relate lexical items: subsertion always corresponds to complementation and sister-adjunction to modification. Furthermore, DTG, unlike TAG, can provide a uniform analysis for em wh-movement in English and Kashmiri, despite the fact that the em wh element in Kashmiri appears in sentence-second position, and not sentence-initial position as in English.

cmp-lg↗