Searcharxiv⌕ Search

arXiv subjects

Chang Wang

Publications and source records attributed to Chang Wang.

40 records · Page 3Linked to original sources

Practice in Synonym Extraction at Large Scale

Synonym extraction is an important task in natural language processing and often used as a submodule in query expansion, question answering and other applications. Automatic synonym extractor is highly preferred for large scale applications. Previous studies in synonym extraction are most limited to small scale datasets. In this paper, we build a large dataset with 3.4 million synonym/non-synonym pairs to capture the challenges in real world scenarios. We proposed (1) a new cost function to accommodate the unbalanced learning problem, and (2) a feature learning based deep neural network to model the complicated relationships in synonym pairs. We compare several different approaches based on SVMs and neural networks, and find out a novel feature learning based neural network outperforms the methods with hand-assigned features. Specifically, the best performance of our model surpasses the SVM baseline with a significant 97\% relative improvement.

cs.CL↗

Medical Synonym Extraction with Concept Space Models

In this paper, we present a novel approach for medical synonym extraction. We aim to integrate the term embedding with the medical domain knowledge for healthcare applications. One advantage of our method is that it is very scalable. Experiments on a dataset with more than 1M term pairs show that the proposed approach outperforms the baseline approaches by a large margin.

cs.CL↗

Dark Solitary Waves in a Class of Collisionally Inhomogeneous Bose-Einstein Condensates

We study the structure, stability, and dynamics of dark solitary waves in parabolically trapped, collisionally inhomogeneous Bose-Einstein condensates (BECs) with spatially periodic variations of the scattering length. This collisional inhomogeneity yields a nonlinear lattice, which we tune from a small-amplitude, approximately sinusoidal structure to a periodic sequence of densely spaced spikes. We start by investigating time-independent inhomogeneities, and we subsequently examine the dynamical response when one starts with a collisionally homogeneous BEC and then switches on an inhomogeneity either adiabatically or nonadiabatically. Using Bogoliubov-de Gennes linearization as well as direct numerical simulations of the Gross-Pitaevskii equation, we observe dark solitary waves, which can become unstable through oscillatory or exponential instabilities. We find a critical wavelength of the nonlinear lattice that is comparable to the healing length. Near this value, the fundamental eigenmode responsible for the stability of the dark solitary wave changes its direction of movement as a function of the strength of the nonlinearity. When it increases, it collides with other eigenmodes, leading to oscillatory instabilities; when it decreases, it collides with the origin and becomes imaginary, illustrating that the instability mechanism is fundamentally different in wide-well versus narrow-well lattices. When starting from a collisionally homogeneous setup and switching on inhomogeneities, we find that dark solitary waves are preserved generically for aligned lattices. We briefly examine the time scales for the onset of solitary-wave oscillations in this scenario.

nlin.PS↗

Cavity-meidated collisionless sympathetic cooling of molecules with atoms

Cooling a range of molecules to ultracold temperatures (<1 mK) is a difficult but important challenge in molecular physics and chemistry. Collective cavity cooling of molecules is a promising method that does not rely on molecular energy level and thus can be applied to all molecules in principle. However, the initial lack of cold molecules leads to the difficulty in its experimental implementation. We show that efficient collective sympathetic cooling of molecules to sub-mK temperatures using a large ensemble of atoms within a cavity is feasible. This approach is a new type of sympathetic cooling which does not rely on direct collisions between atoms and molecules, but utilizes thermalization via their mutual interaction with a cavity field. Two important mechanisms are identified. This include: (1) giant enhancement of cavity optical field from the efficient scattering of the pump light by the atoms; (2) cavity-mediated collective interaction between the atoms and the molecules. We show an optimal cavity detuning for maximizing cooling, which is dependent on the atom and molecule numbers. We determine a threshold for the molecular pump strength and show that it is independent of molecule number when the number of atoms is much greater than the molecules. This can be reduced by orders of magnitude when compared to cavity cooling of single molecular species only. Using this new sympathetic cavity cooling technique, cooling molecules to sub-mK within a high-Q cavity could be within reach of experimental demonstration.

physics.atom-ph↗