51 research outputs found
Efficient, Compositional, Order-Sensitive n-gram Embeddings
No full text<p>This is the companion data for the paper, `"Efficient, Compositional, Order-Sensitive n-gram Embeddings, Adam Poliak, Pushpendre Rastogi, M. Patrick Martin, Benjamin Van Durme, EACL(2017).` For more details see https://www.cs.jhu.edu/~apoliak1/papers/ECO--EACL-2017.pdf</p>
<p> </p>
<p>@inproceedings{Poliak:2017EACL,<br>
Title = {Efficient, Compositional, Order-sensitive n-gram Embeddings},<br>
Author = {Poliak, Adam and Rastogi, Pushpendre and Martin, M. Patrick and Van Durme, Benjamin},<br>
booktitle = {Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics},<br>
Year = {2017},<br>
Publisher = {Association for Computational Linguistics},<br>
location = {Valencia, Spain}<br>
}</p>
<p>This data contains individual skip-embeddings created and the English Wikipedia data used to generate the embeddings.</p>
<p>dim100_c10.tar.gz is missing the skip-embeddings 3 positions to the right of a given word. They can be downloaded from http://www.cs.jhu.edu/~apoliak1/data/eco/cocoon.mincount~5.dim~100.window~3.dim_divide~10.embeds.gz</p>
A survey on Recognizing Textual Entailment as an NLP Evaluation
No full textRecognizing Textual Entailment (RTE) was proposed as a unified evaluation framework to compare semantic understanding of different NLP systems. In this survey paper, we provide an overview of different approaches for evaluating and understanding the reasoning capabilities of NLP systems. We then focus our discussion on RTE by highlighting prominent RTE datasets as well as advances in RTE dataset that focus on specific linguistic phenomena that can be used to evaluate NLP systems on a fine-grained level. We conclude by arguing that when evaluating NLP systems, the community should utilize newly introduced RTE datasets that focus on specific linguistic phenomena
Hand pose estimation during car driving
No full textOdhad pozice ruky hraje klíčovou roli v interakcích člověka s počítačem, navíc do jisté míry umožňuje analyzovat chování člověka. Řešení tohoto problému je netriviální kvůli komplikovaným variacím ruky způsobeným komplexní artikulací, překrýváním částí ruky i tvarové, velikostní a barevné nejednoznačnosti. V této práci se k této problematice stavíme komplexně návrhem několika různých metod jak pro detekci, tak pro odhad pozice ruky a ověřením jejich přesnosti. Porovnáváme metody odhadu z hloubkové mapy a z RGB obrazu a následně je aplikujeme na reálná data z trenažéru řízení. Navíc navrhujeme vlastní nástroj k anotaci RGB-D dat pomocí něhož jsme vytvořili testovací datové sady.Hand pose estimation plays a fundamental role in human computer interactions, moreover it allows us to analyze human behavior. The problem is nontrivial due to complicated hand variations caused by complex articulations, self-occlusions or shape, size and color ambiguities. We provide complex proposal of different detection and pose estimation methods and their evaluation. The comparison of deep map and RGB based pose estimation is provided and applied to real-life data from the driving simulator. Furthermore we design an annotation tool for RGB-D data which we used to produce a test dataset
Mobile Application Using Deep Convolutional Neural Networks
No full textTáto práca popisuje proces tvorby mobilnej aplikácie, ktorá využíva hlboké konvolučné neurónové siete. Proces začína predstavením hlavnej myšlienky, po ktorej nasleduje produktový a technický návrh, implementácia a vyhodnotenie. Práca taktiež skúma technické pozadie rozpoznávania obrazu, a vyberá najvhodnejšie možnosti pre účely aplikácie. Tie sú detekcia objektov a multi-label klasifikácia, ktoré sú obe implementované, vyhodnotené a porovnané. Výsledná aplikácia sa snaží priniesť hodnotu z užívateľského aj technického hľadiska.This thesis describes a process of creating a mobile application using deep convolutional neural networks. The process starts with proposal of the main idea, followed by product and technical design, implementation and evaluation. The thesis also explores the technical background of image recognition, and chooses the most suitable options for the purpose of the application. These are object detection and multi-label classification, which are both implemented, evaluated and compared. The resulting application tries to bring value from both user and technical point of view.
Mobile Application Using Deep Convolutional Neural Networks
No full textThis thesis describes a process of creating a mobile application using deep convolutional neural networks. The process starts with proposal of the main idea, followed by product and technical design, implementation and evaluation. The thesis also explores the technical background of image recognition, and chooses the most suitable options for the purpose of the application. These are object detection and multi-label classification, which are both implemented, evaluated and compared. The resulting application tries to bring value from both user and technical point of view.
Mobile Application Using Deep Convolutional Neural Networks
No full textThis thesis describes a process of creating a mobile application using deep convolutional neural networks. The process starts with proposal of the main idea, followed by product and technical design, implementation and evaluation. The thesis also explores the technical background of image recognition, and chooses the most suitable options for the purpose of the application. These are object detection and multi-label classification, which are both implemented, evaluated and compared. The resulting application tries to bring value from both user and technical point of view
REVISITING RECOGNIZING TEXTUAL ENTAILMENT FOR EVALUATING NATURAL LANGUAGE PROCESSING SYSTEMS
No full textRecognizing Textual Entailment (RTE) began as a unified framework to evaluate the reasoning capabilities of Natural Language Processing (NLP) models. In recent years, RTE has evolved in the NLP community into a task that researchers focus on developing models for. This thesis revisits the tradition of RTE as an evaluation framework for NLP models, especially in the era of deep learning.
Chapter 2 provides an overview of different approaches to evaluating NLP sys- tems, discusses prior RTE datasets, and argues why many of them do not serve as satisfactory tests to evaluate the reasoning capabilities of NLP systems. Chapter 3 presents a new large-scale diverse collection of RTE datasets (DNC) that tests how well NLP systems capture a range of semantic phenomena that are integral to un- derstanding human language. Chapter 4 demonstrates how the DNC can be used to evaluate reasoning capabilities of NLP models. Chapter 5 discusses the limits of RTE as an evaluation framework by illuminating how existing datasets contain biases that may enable crude modeling approaches to perform surprisingly well.
The remaining aspects of the thesis focus on issues raised in Chapter 5. Chapter 6 addresses issues in prior RTE datasets focused on paraphrasing and presents a high-quality test set that can be used to analyze how robust RTE systems are to paraphrases. Chapter 7 demonstrates how modeling approaches on biases, e.g. adversarial learning, can enable RTE models overcome biases discussed in Chapter 5. Chapter 8 applies these methods to the task of discovering emergency needs during disaster events
Laser beam shaping
No full textABSTRACT: In this thesis, the author has tried to research about several methods and techniques to transform a Gaussian input beam of a optical system into a flat-top (uniform) output beam. In practical part, approaches of geometrical and diffusers techniques are used. The model of diffuser will be experimentally confirmed during the experiment. In the final part, results of this experiment and simulation are discussed.RESUMEN: En esta tesis, el autor ha tratado de investigar varios métodos y técnicas para transformar un haz de rayos en la entrada del sistema óptico con distribución Gaussiana en un haz de rayos en salida con una distribución uniforme. En la parte práctica se usan enfoques de las técnicas geométricas y de difusores. El modelo del difusor será confirmado con un experimento. Al final los resultados del experimento y la simulación serán discutidosIngeniería de Telecomunicació
An Immersive Computational Text Analysis Course for Non-Computer Science Students at Barnard College
No full textQuantum-chemical study of C-H bond dissociation enthalpies of various small non-aromatic organic molecules
No full textAbstract: In this work, C-H bond dissociation enthalpies (BDE) and vertical ionization potentials (IP) for various hydrocarbons and ketones were calculated using four density functional approaches. Calculated BDEs and IPs were correlated with experimental data. The linearity of the corresponding dependences can be considered very good. Comparing two used functionals, B3LYP C-H BDE values are closer to experimental results than PBE0 values for both used basis sets. The 6-31G* basis set employed with both functionals, gives the C-H BDEs closer to the experimental values than the 6-311++G** basis set. Using the obtained linear dependences BDE exp = f (BDE calc ), the experimental values of C-H BDEs for some structurally related compounds can be estimated solely from calculations. As a descriptor of the C-H BDE, the IPs and 13 C NMR chemical shifts have been investigated using data obtained from the B3LYP/6-31G* calculations. There is a slight indication of linear correlation between IPs and C-H BDEs in the sets of simple alkanes and alkenes/ cycloalkenes. However, for cycloalkanes and aliphatic carbonyl compounds, no linear correlation was found. In the case of the 13 C NMR chemical shifts, the correlation with C-H BDEs can be found for the sets of alkanes and cycloalkanes, but for the other studied molecules, no trends were detected
- …
