arXiv subjects

Xiangyu Dong

Publications and source records attributed to Xiangyu Dong.

26 records · Page 2Linked to original sources

arXiv2021

Logic-Consistency Text Generation from Semantic Parses

Text generation from semantic parses is to generate textual descriptions for formal representation inputs such as logic forms and SQL queries. This is challenging due to two reasons: (1) the complex and intensive inner logic with the data scarcity constraint, (2) the lack of automatic evaluation metrics for logic consistency. To address these two challenges, this paper first proposes SNOWBALL, a framework for logic consistent text generation from semantic parses that employs an iterative training procedure by recursively augmenting the training set with quality control. Second, we propose a novel automatic metric, BLEC, for evaluating the logical consistency between the semantic parses and generated texts. The experimental results on two benchmark datasets, Logic2Text and Spider, demonstrate the SNOWBALL framework enhances the logic consistency on both BLEC and human evaluation. Furthermore, our statistical analysis reveals that BLEC is more logically consistent with human evaluation than general-purpose automatic metrics including BLEU, ROUGE and, BLEURT. Our data and code are available at https://github.com/Ciaranshu/relogic.

cs.CL↗

arXiv2021

Study to improve the performance of interferometer with ultra-cold atoms

Ultra-cold atoms provide ideal platforms for interferometry. The macroscopic matter-wave property of ultra-cold atoms leads to large coherent length and long coherent time, which enable high accuracy and sensitivity to measurement. Here, we review our efforts to improve the performance of the interferometer. We demonstrate a shortcut method for manipulating ultra-cold atoms in an optical lattice. Compared with traditional ones, this shortcut method can reduce manipulation time by up to three orders of magnitude. We construct a matter-wave Ramsey interferometer for trapped motional quantum states and significantly increase its coherence time by one order of magnitude with an echo technique based on this method. Efforts have also been made to enhance the resolution by multimode scheme. Application of a noise-resilient multi-component interferometer shows that increasing the number of paths could sharpen the peaks in the time-domain interference fringes, which leads to a resolution nearly twice compared with that of a conventional double-path two-mode interferometer. With the shortcut method mentioned above, improvement of the momentum resolution could also be fulfilled, which leads to atomic momentum patterns less than 0.6 $\hbar k_L$. To identify and remove systematic noises, we introduce the methods based on the principal component analysis (PCA) that reduce the noise in detection close to the $1/\sqrt{2}$ of the photon-shot noise and separate and identify or even eliminate noises. Furthermore, we give a proposal to measure precisely the local gravity acceleration within a few centimeters based on our study of ultracold atoms in precision measurements.

cond-mat.quant-gas↗

arXiv2021

Pairwise Learning for Name Disambiguation in Large-Scale Heterogeneous Academic Networks

Name disambiguation aims to identify unique authors with the same name. Existing name disambiguation methods always exploit author attributes to enhance disambiguation results. However, some discriminative author attributes (e.g., email and affiliation) may change because of graduation or job-hopping, which will result in the separation of the same author's papers in digital libraries. Although these attributes may change, an author's co-authors and research topics do not change frequently with time, which means that papers within a period have similar text and relation information in the academic network. Inspired by this idea, we introduce Multi-view Attention-based Pairwise Recurrent Neural Network (MA-PairRNN) to solve the name disambiguation problem. We divided papers into small blocks based on discriminative author attributes and blocks of the same author will be merged according to pairwise classification results of MA-PairRNN. MA-PairRNN combines heterogeneous graph embedding learning and pairwise similarity learning into a framework. In addition to attribute and structure information, MA-PairRNN also exploits semantic information by meta-path and generates node representation in an inductive way, which is scalable to large graphs. Furthermore, a semantic-level attention mechanism is adopted to fuse multiple meta-path based representations. A Pseudo-Siamese network consisting of two RNNs takes two paper sequences in publication time order as input and outputs their similarity. Results on two real-world datasets demonstrate that our framework has a significant and consistent improvement of performance on the name disambiguation task. It was also demonstrated that MA-PairRNN can perform well with a small amount of training data and have better generalization ability across different research areas.

cs.DL↗

arXiv2020

Did You Ask a Good Question? A Cross-Domain Question Intention Classification Benchmark for Text-to-SQL

Neural models have achieved significant results on the text-to-SQL task, in which most current work assumes all the input questions are legal and generates a SQL query for any input. However, in the real scenario, users can input any text that may not be able to be answered by a SQL query. In this work, we propose TriageSQL, the first cross-domain text-to-SQL question intention classification benchmark that requires models to distinguish four types of unanswerable questions from answerable questions. The baseline RoBERTa model achieves a 60% F1 score on the test set, demonstrating the need for further improvement on this task. Our dataset is available at https://github.com/chatc/TriageSQL.

cs.CL↗

arXiv2020

MedDialog: Two Large-scale Medical Dialogue Datasets

Medical dialogue systems are promising in assisting in telemedicine to increase access to healthcare services, improve the quality of patient care, and reduce medical costs. To facilitate the research and development of medical dialogue systems, we build two large-scale medical dialogue datasets: MedDialog-EN and MedDialog-CN. MedDialog-EN is an English dataset containing 0.3 million conversations between patients and doctors and 0.5 million utterances. MedDialog-CN is an Chinese dataset containing 1.1 million conversations and 4 million utterances. To our best knowledge, MedDialog-(EN,CN) are the largest medical dialogue datasets to date. The dataset is available at https://github.com/UCSD-AI4H/Medical-Dialogue-System

cs.LG↗

arXiv2020

Implementation of a double-path multimode interferometer using a spinor Bose-Einstein condensate

We realize a double-path multimode matter wave interferometer with spinor Bose-Einstein condensate and observe clear spatial interference fringes as well as a periodic change of the visibility in the time domain, which we refer to as the time domain interference and which is different from the traditional double-path interferometer. By changing the relative phase of the two paths, we find that the spatial fringes first lose coherence and recover. As the number of modes increases, the time domain interference signal with the narrower peaks is observed, which is beneficial to the improvement of the resolution of the phase measurement. We also investigated the influence of initial phase configuration and phase evolution rate between different modes in the two paths. With enhanced resolution, the sensitivity of interferometric measurements of physical observables can also be improved by properly assigning measurable quantities to the relative phase between two paths.

physics.atom-ph↗

arXiv2016

Atomic momentum patterns with narrower interval

We studied the atomic momentum distribution for a superposition of Bloch states spreading in the lowest band of an optical lattice after the action of the standing wave pulse. By designing the imposing pulse on this superposed state, an atomic momentum pattern appears with narrower interval between the adjacent peaks that can be far less than the double recoil momentum. The patterns with narrower interval come from the superposition of the action of the designed pulse on many Bloch states with quasi-momenta over the first Brillouin zone, where for each quasi-momentum there is an interference among several lowest bands. Our experimental result of narrow interval peaks is consistent with the theoretical simulation. The patterns of multi modes with different quasi-momenta are helpful for precise measurement and atomic manipulation.

cond-mat.quant-gas↗

arXiv2015

Excitation of atoms in an optical lattice driven by polychromatic amplitude modulation

We investigate the mutiphoton process between different Bloch states in an amplitude modulated optical lattice. In the experiment, we perform the modulation with more than one frequency components, which includes a high degree of freedom and provides a flexible way to coherently control quantum states. Based on the study of single frequency modulation, we investigate the collaborative effect of different frequency components in two aspects. Through double frequency modulations, the spectrums of excitation rates for different lattice depths are measured. Moreover, interference between two separated excitation paths is shown, emphasizing the influence of modulation phases when two modulation frequencies are commensurate. Finally, we demonstrate the application of the double frequency modulation to design a large-momentum-transfer beam splitter. The beam splitter is easy in practice and would not introduce phase shift between two arms.

cond-mat.quant-gas↗