arXiv subjects

Bingbing Zhang

Publications and source records attributed to Bingbing Zhang.

22 records · Page 2Linked to original sources

arXiv2019

LIDA: Lightweight Interactive Dialogue Annotator

Dialogue systems have the potential to change how people interact with machines but are highly dependent on the quality of the data used to train them. It is therefore important to develop good dialogue annotation tools which can improve the speed and quality of dialogue data annotation. With this in mind, we introduce LIDA, an annotation tool designed specifically for conversation data. As far as we know, LIDA is the first dialogue annotation system that handles the entire dialogue annotation pipeline from raw text, as may be the output of transcription services, to structured conversation data. Furthermore it supports the integration of arbitrary machine learning models as annotation recommenders and also has a dedicated interface to resolve inter-annotator disagreements such as after crowdsourcing annotations for a dataset. LIDA is fully open source, documented and publicly available [ https://github.com/Wluper/lida ]

cs.CL↗

arXiv2018

Evolutionary Data Measures: Understanding the Difficulty of Text Classification Tasks

Classification tasks are usually analysed and improved through new model architectures or hyperparameter optimisation but the underlying properties of datasets are discovered on an ad-hoc basis as errors occur. However, understanding the properties of the data is crucial in perfecting models. In this paper we analyse exactly which characteristics of a dataset best determine how difficult that dataset is for the task of text classification. We then propose an intuitive measure of difficulty for text classification datasets which is simple and fast to calculate. We show that this measure generalises to unseen data by comparing it to state-of-the-art datasets and results. This measure can be used to analyse the precise source of errors in a dataset and allows fast estimation of how difficult a dataset is to learn. We searched for this measure by training 12 classical and neural network based models on 78 real-world datasets, then use a genetic algorithm to discover the best measure of difficulty. Our difficulty-calculating code ( https://github.com/Wluper/edm ) and datasets ( http://data.wluper.com ) are publicly available.

cs.CL↗

arXiv2018

Giant photoinduced lattice distortion in oxygen-vacancy ordered SrCoO2.5 thin films

Despite of the tremendous efforts spent on the oxygen vacancy migration in determining the property optimization of oxygen-vacancy enrichment transition metal oxides, few has focused on their dynamic behaviors non-equilibrium states. In this work, we performed multi-timescale ultrafast X-ray diffraction measurements by using picosecond synchrotron X-ray pulses and femtosecond table-top X-ray pulses to monitor the structural dynamics in the oxygen-vacancy ordered SrCoO2.5 thin films. A giant photoinduced strain (Δc/c > 1%) was observed, whose distinct correlation with the pump photon energy indicates a non-thermal origin of the photoinduced strain. The sub-picosecond resolution X-ray diffraction reveals the formation and propagation of the coherent acoustic phonons inside the film. We also simulate the effect of photoexcited electron-hole pairs and the resulting lattice changes using the Density Function Theory method to obtain further insight on the microscopic mechanism of the measured photostriction effect. Comparable photostrictive responses and the strong dependence on excitation wavelength are predicted, revealing a bonding to anti-bonding charge transfer or high spin to intermediate spin crossover induced lattice expansion in the oxygen-vacancy films.

cond-mat.mtrl-sci↗

arXiv2018

Underwater Source Localization Using TDOA and FDOA Measurements with Unknown Propagation Speed and Sensor Parameter Errors

Underwater source localization problems are complicated and challenging: a) the sound propagation speed is often unknown and the unpredictable ocean current might lead to the uncertainties of sensor parameters (i.e. position and velocity); b) the underwater acoustic signal travels much slower than the radio one in terrestrial environments, thus resulting in a significantly severe Doppler effect; c) energy-efficient techniques are urgently required and hence in favour of the design with a low computational complexity. Considering these issues, we propose a simple and efficient underwater source localization approach based on time difference of arrival (TDOA) and frequency difference of arrival (FDOA) measurements, which copes with unknown propagation speed and sensor parameter errors. The proposed method mitigates the impact of the Doppler effect for accurately inferring the source parameters (i.e. position and velocity). The Cramer-Rao lower bounds (CRLBs) for this kind of localization are derived and, moreover, the analytical study shows that our method can yield the performance that is very close to the CRLB, particularly under small noise. The numerical results not only confirm the above conclusions but also show that our method outperforms other competing approaches.

eess.SP↗