arXiv subjects

Mark Johnson

Publications and source records attributed to Mark Johnson.

At least 55 records · Page 3Linked to original sources

arXiv2017

A Novel Neural Network Model for Joint POS Tagging and Graph-based Dependency Parsing

We present a novel neural network model that learns POS tagging and graph-based dependency parsing jointly. Our model uses bidirectional LSTMs to learn feature representations shared for both POS tagging and dependency parsing tasks, thus handling the feature-engineering problem. Our extensive experiments, on 19 languages from the Universal Dependencies project, show that our model outperforms the state-of-the-art neural network-based Stack-propagation model for joint POS tagging and transition-based dependency parsing, resulting in a new state of the art. Our code is open-source and available together with pre-trained models at: https://github.com/datquocnguyen/jPTDP

cs.CL↗

arXiv2017

Neighborhood Mixture Model for Knowledge Base Completion

Knowledge bases are useful resources for many natural language processing tasks, however, they are far from complete. In this paper, we define a novel entity representation as a mixture of its neighborhood in the knowledge base and apply this technique on TransE-a well-known embedding model for knowledge base completion. Experimental results show that the neighborhood information significantly helps to improve the results of the TransE model, leading to better performance than obtained by other state-of-the-art embedding models on three benchmark datasets for triple classification, entity prediction and relation prediction tasks.

cs.CL↗

arXiv2017

STransE: a novel embedding model of entities and relationships in knowledge bases

Knowledge bases of real-world facts about entities and their relationships are useful resources for a variety of natural language processing tasks. However, because knowledge bases are typically incomplete, it is useful to be able to perform link prediction or knowledge base completion, i.e., predict whether a relationship not in the knowledge base is likely to be true. This paper combines insights from several previous link prediction models into a new embedding model STransE that represents each entity as a low-dimensional vector, and each relation by two matrices and a translation vector. STransE is a simple combination of the SE and TransE models, but it obtains better link prediction performance on two benchmark datasets than previous embedding models. Thus, STransE can serve as a new baseline for the more complex models in the link prediction task.

cs.CL↗

arXiv2016

Search Personalization with Embeddings

Recent research has shown that the performance of search personalization depends on the richness of user profiles which normally represent the user's topical interests. In this paper, we propose a new embedding approach to learning user profiles, where users are embedded on a topical interest space. We then directly utilize the user profiles for search personalization. Experiments on query logs from a major commercial web search engine demonstrate that our embedding approach improves the performance of the search engine and also achieves better search performance than other strong baselines.

cs.IR↗

arXiv2016

An empirical study for Vietnamese dependency parsing

This paper presents an empirical comparison of different dependency parsers for Vietnamese, which has some unusual characteristics such as copula drop and verb serialization. Experimental results show that the neural network-based parsers perform significantly better than the traditional parsers. We report the highest parsing scores published to date for Vietnamese with the labeled attachment score (LAS) at 73.53% and the unlabeled attachment score (UAS) at 80.66%.

cs.CL↗

arXiv2016

SPICE: Semantic Propositional Image Caption Evaluation

There is considerable interest in the task of automatically generating image captions. However, evaluation is challenging. Existing automatic evaluation metrics are primarily sensitive to n-gram overlap, which is neither necessary nor sufficient for the task of simulating human judgment. We hypothesize that semantic propositional content is an important component of human caption evaluation, and propose a new automated caption evaluation metric defined over scene graphs coined SPICE. Extensive evaluations across a range of models and datasets indicate that SPICE captures human judgments over model-generated captions better than other automatic metrics (e.g., system-level correlation of 0.88 with human judgments on the MS COCO dataset, versus 0.43 for CIDEr and 0.53 for METEOR). Furthermore, SPICE can answer questions such as `which caption-generator best understands colors?' and `can caption-generators count?'

cs.CV↗

arXiv2016

N(N)LO event files: applications and prospects

In this contribution we comment on the use of NLO n-Tuples and consider their extension to NNLO. As an application of n-Tuples we present preliminary results of a strong coupling extraction from a measurement of the production of Z + 2, 3, 4 jets at the LHC at 7 TeV.

hep-ph↗

arXiv2014

High-throughput screening for modulators of cellular contractile force

When cellular contractile forces are central to pathophysiology, these forces comprise a logical target of therapy. Nevertheless, existing high-throughput screens are limited to upstream signaling intermediates with poorly defined relationship to such a physiological endpoint. Using cellular force as the target, here we screened libraries to identify novel drug candidates in the case of human airway smooth muscle cells in the context of asthma, and also in the case of Schlemm's canal endothelial cells in the context of glaucoma. This approach identified several drug candidates for both asthma and glaucoma. We attained rates of 1000 compounds per screening day, thus establishing a force-based cellular platform for high-throughput drug discovery.

physics.bio-ph↗

arXiv2014

Ubiquity of complete intersection liaison classes

In this paper, we provide constructions to enumerate large numbers of CI-liaison classes. To this end, we introduce a liaison invariant and prove several results concerning it, notably that it commutes with hypersurface sections. This theory is applied to the CI-liaison classes of ruled joins of projective schemes, yielding strong obstructions for such joins to lie in the same liaison class. A second construction arises from the actions of automorphisms on liaison classes, allowing the enumeration of many liaison classes of perfect ideals of codimension at least three.

math.AC↗

arXiv2014

Electronic and vibronic properties of a discotic liquid-crystal and its charge transfer complex

Discotic liquid crystalline (DLC) charge transfer (CT) complexes combine visible light absorption and rapid charge transfer characteristics within the CT complex, being favorable properties for photovoltaic (PV) applications. We present a detailed study of the electronic and vibrational properties of the prototypic 1:1 mixture of discotic 2,3,6,7,10,11-hexakishexyloxytriphenylene (HAT6) and 2,4,7-trinitro-9-fluorenone (TNF). It is shown that intermolecular charge transfer occurs in the groundstate of the complex: a charge delocalization of about 10-2 electron from the HAT6 core to TNF is deduced from both Raman and our previous NMR measurements (Reference 32), implying the presence of permanent dipoles at the donor-acceptor interface. A combined analysis of density functional theory calculations, resonant Raman and UV-VIS absorption measurements indicate that fast relaxation occurs in the UV region due to intramolecular vibronic coupling of HAT6 quinoidal modes with lower lying electronic states. Relatively slower relaxation in the visible region CT-band of the complex is also indicated, which likely involves motions of the TNF nitro groups. The fast quinoidal relaxation process in the hot UV band of HAT6 relates to pseudo-Jahn-Teller interactions in a single benzene unit, suggesting that the underlying vibronic coupling mechanism can be generic for polyaromatic hydrocarbons. Both the presence of CT dipoles and relatively slow relaxation processes in the CT band can be relevant concerning the design of DLC based organic PV systems.

cond-mat.soft↗

arXiv2009

Film Edge Nonlocal Spin Valves

Spintronics is a new paradigm for integrated digital electronics. Recently established as a niche for nonvolatile magnetic random access memory (MRAM), it offers new functionality while demonstrating low power and high speed performance. However, to reach high density spintronic technology must make a transition to the nanometer scale. Prototype devices are presently made using a planar geometry and have an area determined by the lithographic feature size, currently about 100 nm. Here we present a new nonplanar geometry in which one lateral dimension is given by a film thickness, the order of 10 nm. With this new approach, cell sizes can shrink by an order of magnitude. The geometry is demonstrated with a nonlocal spin valve, where we study devices with an injector/detector separation much less than the spin diffusion length.

cond-mat.mtrl-sci↗

arXiv2006

Hierarchical geometric frustration in La3Cu2VO9

The crystallographic structure and magnetic properties of the La3Cu2VO9 were investigated by powder neutron diffraction and magnetization measurements. The compound materializes geometric frustration at two spatial scales, within clusters and between clusters, and at different temperature scales. It is shown by exactly solving the hamiltonian spectrum that collective spins are formed on each clusters at low temperature before inter-clusters coupling operates.

cond-mat.str-el↗

arXiv2001

Joint and conditional estimation of tagging and parsing models

This paper compares two different ways of estimating statistical language models. Many statistical NLP tagging and parsing models are estimated by maximizing the (joint) likelihood of the fully-observed training data. However, since these applications only require the conditional probability distributions, these distributions can in principle be learnt by maximizing the conditional likelihood of the training data. Perhaps somewhat surprisingly, models estimated by maximizing the joint were superior to models estimated by maximizing the conditional, even though some of the latter models intuitively had access to ``more information''.

cs.CL↗

arXiv2000

Lexicalized Stochastic Modeling of Constraint-Based Grammars using Log-Linear Measures and EM Training

We present a new approach to stochastic modeling of constraint-based grammars that is based on log-linear models and uses EM for estimation from unannotated data. The techniques are applied to an LFG grammar for German. Evaluation on an exact match task yields 86% precision for an ambiguity rate of 5.4, and 90% precision on a subcat frame match for an ambiguity rate of 25. Experimental comparison to training from a parsebank shows a 10% gain from EM training. Also, a new class-based grammar lexicalization is presented, showing a 10% gain over unlexicalized models.

cs.CL↗

arXiv2000

Estimators for Stochastic ``Unification-Based'' Grammars

Log-linear models provide a statistically sound framework for Stochastic ``Unification-Based'' Grammars (SUBGs) and stochastic versions of other kinds of grammars. We describe two computationally-tractable ways of estimating the parameters of such grammars from a training corpus of syntactic analyses, and apply these to estimate a stochastic version of Lexical-Functional Grammar.

cs.CL↗

arXiv2000

Exploiting auxiliary distributions in stochastic unification-based grammars

This paper describes a method for estimating conditional probability distributions over the parses of ``unification-based'' grammars which can utilize auxiliary distributions that are estimated by other means. We show how this can be used to incorporate information about lexical selectional preferences gathered from other sources into Stochastic ``Unification-based'' Grammars (SUBGs). While we apply this estimator to a Stochastic Lexical-Functional Grammar, the method is general, and should be applicable to stochastic versions of HPSGs, categorial grammars and transformational grammars.

cs.CL↗

arXiv2000

Explaining away ambiguity: Learning verb selectional preference with Bayesian networks

This paper presents a Bayesian model for unsupervised learning of verb selectional preferences. For each verb the model creates a Bayesian network whose architecture is determined by the lexical hierarchy of Wordnet and whose parameters are estimated from a list of verb-object pairs found from a corpus. ``Explaining away'', a well-known property of Bayesian networks, helps the model deal in a natural fashion with word sense ambiguity in the training data. On a word sense disambiguation test our model performed better than other state of the art systems for unsupervised learning of selectional preferences. Computational complexity problems, ways of improving this approach and methods for implementing ``explaining away'' in other graphical frameworks are discussed.

cs.CL↗

arXiv2000

Compact non-left-recursive grammars using the selective left-corner transform and factoring

The left-corner transform removes left-recursion from (probabilistic) context-free grammars and unification grammars, permitting simple top-down parsing techniques to be used. Unfortunately the grammars produced by the standard left-corner transform are usually much larger than the original. The selective left-corner transform described in this paper produces a transformed grammar which simulates left-corner recognition of a user-specified set of the original productions, and top-down recognition of the others. Combined with two factorizations, it produces non-left-recursive grammars that are not much larger than the original.

cs.CL↗