arXiv subjects

Fei Huang

Publications and source records attributed to Fei Huang.

At least 505 records · Page 28Linked to original sources

arXiv2020

AIN: Fast and Accurate Sequence Labeling with Approximate Inference Network

The linear-chain Conditional Random Field (CRF) model is one of the most widely-used neural sequence labeling approaches. Exact probabilistic inference algorithms such as the forward-backward and Viterbi algorithms are typically applied in training and prediction stages of the CRF model. However, these algorithms require sequential computation that makes parallelization impossible. In this paper, we propose to employ a parallelizable approximate variational inference algorithm for the CRF model. Based on this algorithm, we design an approximate inference network that can be connected with the encoder of the neural CRF model to form an end-to-end network, which is amenable to parallelization for faster training and prediction. The empirical results show that our proposed approaches achieve a 12.7-fold improvement in decoding speed with long sentences and a competitive accuracy compared with the traditional CRF approach.

cs.CL↗

arXiv2020

PALM: Pre-training an Autoencoding&Autoregressive Language Model for Context-conditioned Generation

Self-supervised pre-training, such as BERT, MASS and BART, has emerged as a powerful technique for natural language understanding and generation. Existing pre-training techniques employ autoencoding and/or autoregressive objectives to train Transformer-based models by recovering original word tokens from corrupted text with some masked tokens. The training goals of existing techniques are often inconsistent with the goals of many language generation tasks, such as generative question answering and conversational response generation, for producing new text given context. This work presents PALM with a novel scheme that jointly pre-trains an autoencoding and autoregressive language model on a large unlabeled corpus, specifically designed for generating new text conditioned on context. The new scheme alleviates the mismatch introduced by the existing denoising scheme between pre-training and fine-tuning where generation is more than reconstructing original text. An extensive set of experiments show that PALM achieves new state-of-the-art results on a variety of language generation benchmarks covering generative question answering (Rank 1 on the official MARCO leaderboard), abstractive summarization on CNN/DailyMail as well as Gigaword, question generation on SQuAD, and conversational response generation on Cornell Movie Dialogues.

cs.CL↗

arXiv2020

Photoproduction $γp \to K^{*+} Λ$ in a Reggeized model

The high-precision differential cross-section data for the reaction $γp \to K^{*+}Λ$ are reanalyzed within a Regge-inspired effective Lagrangian approach. The model adopts Regge phenomenology to constrain the $t$-channel contributions from the $κ$, $K$, and $K^*$ exchanges. A minimal number of resonances in the $s$ channel are introduced in constructing the reaction amplitudes in order to describe the data. It is shown that the differential cross-section data for $γp \to K^{*+}Λ$ can be satisfactorily described by introducing the only $N(2060){5/2}^-$ resonance in the $s$ channel, which is quite different from our earlier work performed in an effective Lagrangian approach [A. C. Wang {\it et al.}, Phys. Rev. C 96, 035206 (2017)], where the amplitudes are computed by evaluating Feynman diagrams and it is found that introducing at least one additional resonance apart from the $N(2060){5/2}^-$ is indispensable for reproducing the data. The roles of individual contributions from meson and baryon exchanges on the angular distributions are found to be highly model dependent. The extracted mass of $N(2060){5/2}^-$ turns out to be well determined, independent of how the $t$-channel amplitudes are constructed, whereas the width does not.

nucl-th↗

arXiv2020

Deciphering the Archaeological Record: Cosmological Imprints of Non-Minimal Dark Sectors

Many proposals for physics beyond the Standard Model give rise to a dark sector containing many degrees of freedom. In this work, we explore the cosmological implications of the non-trivial dynamics which may arise within such dark sectors, focusing on decay processes which take place entirely among the dark constituents. First, we demonstrate that such decays can leave dramatic imprints on the resulting dark-matter phase-space distribution. In particular, this distribution need not be thermal -- it can even be multi-modal, exhibiting a non-trivial pattern of peaks and troughs as a function of momentum. We then proceed to show how these features can induce modifications to the matter power spectrum. Finally, we assess the extent to which one can approach the archaeological "inverse" problem of deciphering the properties of an underlying dark sector from the matter power spectrum. Indeed, one of the main results of this paper is a remarkably simple conjectured analytic expression which permits the reconstruction of many of the important features of the dark-matter phase-space distribution directly from the matter power spectrum. Our results therefore provide an interesting toolbox of methods for learning about, and potentially constraining, the features of non-minimal dark sectors and their dynamics in the early universe.

astro-ph.CO↗

arXiv2020

Freeze-in Dark Matter from Secret Neutrino Interactions

We investigate a simplified freeze-in dark-matter model in which the dark matter only interacts with the standard-model neutrinos via a light scalar. The extremely small coupling for the freeze-in mechanism is naturally realized in several neutrino-portal scenarios with the secret neutrino interactions. We study possible evolution history of the dark sector: the dark sector would undergo pure freeze-in production if the interactions between the dark-sector particles are negligible, while thermal equilibrium within the dark sector could occur if the reannihilation of the dark matter and the scalar mediator is rapid enough. We investigate the relic abundance in the freeze-in and dark freeze-out regimes, calculate evolution of the dark temperature, and study its phenomenological aspects on BBN and CMB constraints, the indirect-detection signature, as well as the potential to solve the small scale structure problem.

hep-ph↗

arXiv2020

Structure-Level Knowledge Distillation For Multilingual Sequence Labeling

Multilingual sequence labeling is a task of predicting label sequences using a single unified model for multiple languages. Compared with relying on multiple monolingual models, using a multilingual model has the benefit of a smaller model size, easier in online serving, and generalizability to low-resource languages. However, current multilingual models still underperform individual monolingual models significantly due to model capacity limitations. In this paper, we propose to reduce the gap between monolingual models and the unified multilingual model by distilling the structural knowledge of several monolingual models (teachers) to the unified multilingual model (student). We propose two novel KD methods based on structure-level information: (1) approximately minimizes the distance between the student's and the teachers' structure level probability distributions, (2) aggregates the structure-level knowledge to local distributions and minimizes the distance between two local probability distributions. Our experiments on 4 multilingual tasks with 25 datasets show that our approaches outperform several strong baselines and have stronger zero-shot generalizability than both the baseline model and teacher models.

cs.CL↗

arXiv2020

Analysis of the data on spin density matrix elements for $γp \to K^{*+}Λ$

In our previous work [Phys. Rev. C {\bf 96}, 035206 (2017)], the high-precision differential cross-section data for $γp \to K^{*+}Λ$ reported by the CLAS Collaboration has been analyzed within an effective Lagrangian approach. It was found that apart from the $t$-channel $K$, $K^*$, and $κ$ exchanges, the $u$-channel $Λ$, $Σ$, and $Σ^*$ exchanges, the $s$-channel $N$ exchange, and the interaction current, one needs to introduce at least two nucleon resonances in the $s$ channel in constructing the reaction amplitudes to describe the cross-section data. One of the needed resonances is $N(2060)5/2^-$, and the other one could be one of the $N(2000)5/2^+$, $N(2040)3/2^+$, $N(2100)1/2^+$, $N(2120)3/2^-$, and $N(2190)7/2^-$ resonances. In this paper, we further include in our analysis the data on spin density matrix elements for $K^*$ meson reported recently by the CLAS Collaboration, with the purpose being to impose further constraints on extracting the resonance contents and to gain a better understanding of the reaction mechanism. It turns out that with the new data on spin density matrix elements taken into account, only the set with the $N(2060)5/2^-$ and $N(2000)5/2^+$ resonances among those five possible solutions extracted from the analysis of the differential cross-section data can satisfactorily describe the data on both the differential cross sections and the spin density matrix elements. Further analysis shows that this reaction is dominated by the $t$-channel $K$ exchange and $s$-channel $N(2060)5/2^-$ and $N(2000)5/2^+$ exchanges.

nucl-th↗

arXiv2020

Nucleon and $Δ$ resonances in $γp \to K^+ Σ^0(1385)$ photoproduction

The photoproduction of $γp \to K^+ Σ^0(1385)$ is investigated based on an effective Lagrangian approach using the tree-level Born approximation, with the purpose of understanding the reaction mechanisms and resonance contents and their associated parameters in this reaction. In addition to the $t$-channel $K$ and $K^\ast(892)$ exchanges, $s$-channel nucleon ($N$) exchange, $u$-channel $Λ$ exchange, and generalized contact term, the exchanges of a minimum number of $N$ and $Δ$ resonances in the $s$ channel are taken into account in constructing the reaction amplitudes to describe the experimental data. It is found that the most recent differential cross-section data from the CLAS Collaboration can be well reproduced by including one of the $N(1895){1/2}^-$, $Δ(1900){1/2}^-$, and $Δ(1930){5/2}^-$ resonances. The reaction mechanisms of $γp \to K^+ Σ^0(1385)$ are discussed in detail, and the predictions of the beam and target asymmetries for this reaction are given. The cross sections of $γp \to K^0 Σ^+(1385)$ are shown to be able to further constrain the theoretical models and pin down the resonance contents for $γp \to K^+ Σ^0(1385)$.

hep-ph↗

arXiv2020

Selected strong decays of pentaquark State $P_c(4312)$ in a chiral constituent quark model

The newly confirmed pentaquark state $P_c(4312)$ has been treated as a weakly bound $(Σ_c\bar{D})$ state by a well-established chiral constituent quark model and by a dynamical calculation on quark degrees of freedom where the quark exchange effect is accounted for. The obtained mass $4308$ MeV agrees with data. In this work, the selected strong decays of the $P_c(4312)$ state are studied with the obtained wave function. It is shown that the width of the $Λ_c\bar{D}^*$ decay is overwhelmed and the branching ratios of the $p\,η_c$ and $p\,J/ψ$ decays are both less than 1 percentage.

hep-ph↗

arXiv2020

CoTK: An Open-Source Toolkit for Fast Development and Fair Evaluation of Text Generation

In text generation evaluation, many practical issues, such as inconsistent experimental settings and metric implementations, are often ignored but lead to unfair evaluation and untenable conclusions. We present CoTK, an open-source toolkit aiming to support fast development and fair evaluation of text generation. In model development, CoTK helps handle the cumbersome issues, such as data processing, metric implementation, and reproduction. It standardizes the development steps and reduces human errors which may lead to inconsistent experimental settings. In model evaluation, CoTK provides implementation for many commonly used metrics and benchmark models across different experimental settings. As a unique feature, CoTK can signify when and which metric cannot be fairly compared. We demonstrate that it is convenient to use CoTK for model development and evaluation, particularly across different experimental settings.

cs.CL↗

arXiv2020

A Knowledge-Enhanced Pretraining Model for Commonsense Story Generation

Story generation, namely generating a reasonable story from a leading context, is an important but challenging task. In spite of the success in modeling fluency and local coherence, existing neural language generation models (e.g., GPT-2) still suffer from repetition, logic conflicts, and lack of long-range coherence in generated stories. We conjecture that this is because of the difficulty of associating relevant commonsense knowledge, understanding the causal relationships, and planning entities and events with proper temporal order. In this paper, we devise a knowledge-enhanced pretraining model for commonsense story generation. We propose to utilize commonsense knowledge from external knowledge bases to generate reasonable stories. To further capture the causal and temporal dependencies between the sentences in a reasonable story, we employ multi-task learning which combines a discriminative objective to distinguish true and fake stories during fine-tuning. Automatic and manual evaluation shows that our model can generate more reasonable stories than state-of-the-art baselines, particularly in terms of logic and global coherence.

cs.CL↗

arXiv2019

Event Ticket Price Prediction with Deep Neural Network on Spatial-Temporal Sparse Data

Event ticket price prediction is important to marketing strategy for any sports team or musical ensemble. An accurate prediction model can help the marketing team to make promotion plan more effectively and efficiently. However, given all the historical transaction records, it is challenging to predict the sale price of the remaining seats at any future timestamp, not only because that the sale price is relevant to a lot of features (seat locations, date-to-event of the transaction, event date, team performance, etc.), but also because of the temporal and spatial sparsity in the dataset. For a game/concert, the ticket selling price of one seat is only observable once at the time of sale. Furthermore, some seats may not even be purchased (therefore no record available). In fact, data sparsity is commonly encountered in many prediction problems. Here, we propose a bi-level optimizing deep neural network to address the curse of spatio-temporal sparsity. Specifically, we introduce coarsening and refining layers, and design a bi-level loss function to integrate different level of loss for better prediction accuracy. Our model can discover the interrelations among ticket sale price, seat locations, selling time, event information, etc. Experiments show that our proposed model outperforms other benchmark methods in real-world ticket selling price prediction.

cs.LG↗

arXiv2019

Knowing What, How and Why: A Near Complete Solution for Aspect-based Sentiment Analysis

Target-based sentiment analysis or aspect-based sentiment analysis (ABSA) refers to addressing various sentiment analysis tasks at a fine-grained level, which includes but is not limited to aspect extraction, aspect sentiment classification, and opinion extraction. There exist many solvers of the above individual subtasks or a combination of two subtasks, and they can work together to tell a complete story, i.e. the discussed aspect, the sentiment on it, and the cause of the sentiment. However, no previous ABSA research tried to provide a complete solution in one shot. In this paper, we introduce a new subtask under ABSA, named aspect sentiment triplet extraction (ASTE). Particularly, a solver of this task needs to extract triplets (What, How, Why) from the inputs, which show WHAT the targeted aspects are, HOW their sentiment polarities are and WHY they have such polarities (i.e. opinion reasons). For instance, one triplet from "Waiters are very friendly and the pasta is simply average" could be ('Waiters', positive, 'friendly'). We propose a two-stage framework to address this task. The first stage predicts what, how and why in a unified model, and then the second stage pairs up the predicted what (how) and why from the first stage to output triplets. In the experiments, our framework has set a benchmark performance in this novel triplet extraction task. Meanwhile, it outperforms a few strong baselines adapted from state-of-the-art related methods.

cs.CL↗

arXiv2019

The Inflaton Portal to a Highly decoupled EeV Dark Matter Particle

We explore the possibility that the dark-matter relic abundance is generated in a context where the inflaton is the only mediator between the visible and the hidden sectors of our universe. Due to the relatively large mass of the inflaton field, such a portal leads to an extremely feeble interaction between the dark and the visible sectors suggesting that the dark sector cannot reach any thermal equilibrium with the visible sector. After the two sectors are populated by the decay of the inflaton, a heavy dark-matter particle thermally decouples within the dark sector. Later, a lighter dark particle, whose decay width is naturally suppressed by the inflaton propagator, decays into the visible sector after it dominates the energy density of universe. This process dilutes the dark-matter relic density by injecting entropy in the visible sector. We show that an inflaton mass of $\mathcal{O}(10^{13})$ GeV together with couplings of order one are fully compatible with a dark-matter relic abundance $Ωh^2\sim 0.1$. As a general feature of the model, the entropy dilution mechanism is accompanied by a period of early matter domination, which modifies the amount of e-folds of inflation necessary to accommodate Planck data. Moreover, the coupling of the inflaton to the dark and visible sectors brings loop contributions to the inflationary potential which can destabilize the inflation trajectory. Considering all these complementary constraints, we show that, in the context of a plateau-inflation scenario such as the $α$-attractor model, the inflaton can constitute a viable mediator between the visible sector and a $\sim 10$ EeV dark-matter candidate. Furthermore, we show that improved constraints on the tensor-to-scalar ratio and spectral index could potentially rule out dark-matter scenarios of this sort in the future.

hep-ph↗

arXiv2019

ARAML: A Stable Adversarial Training Framework for Text Generation

Most of the existing generative adversarial networks (GAN) for text generation suffer from the instability of reinforcement learning training algorithms such as policy gradient, leading to unstable performance. To tackle this problem, we propose a novel framework called Adversarial Reward Augmented Maximum Likelihood (ARAML). During adversarial training, the discriminator assigns rewards to samples which are acquired from a stationary distribution near the data rather than the generator's distribution. The generator is optimized with maximum likelihood estimation augmented by the discriminator's rewards instead of policy gradient. Experiments show that our model can outperform state-of-the-art text GANs with a more stable training process.

cs.CL↗

arXiv2019

Unsupervised Multi-modal Neural Machine Translation

Unsupervised neural machine translation (UNMT) has recently achieved remarkable results with only large monolingual corpora in each language. However, the uncertainty of associating target with source sentences makes UNMT theoretically an ill-posed problem. This work investigates the possibility of utilizing images for disambiguation to improve the performance of UNMT. Our assumption is intuitively based on the invariant property of image, i.e., the description of the same visual content by different languages should be approximately similar. We propose an unsupervised multi-modal machine translation (UMNMT) framework based on the language translation cycle consistency loss conditional on the image, targeting to learn the bidirectional multi-modal translation simultaneously. Through an alternate training between multi-modal and uni-modal, our inference model can translate with or without the image. On the widely used Multi30K dataset, the experimental results of our approach are significantly better than those of the text-only UNMT on the 2016 test dataset.

cs.CV↗

arXiv2019

$d^*(2380)$ and its partners in a diquark model

The purpose of the present study was to explore the possibility of accommodating the $d^*(2380)$ and its flavor SU(3) partners in a diquark model. Proposing that $d^*(2380)$ is composed of three vector diquarks, its mass is calculated by use of an effective Hamiltonian approach and its decay width is estimated by considering the effects of quark tunneling from one diquark to the others and the decays of the subsequent two-baryon bound state. Both the obtained mass and decay width of $d^*(2380)$ are in agreement with the experimental data, with the unexpected narrow decay width being naturally explained by the large tunneling suppression of a quark between a pair of diquarks. The masses and decay widths of the flavor SU(3) partners of $d^*(2380)$ are also predicated within the same diquark scenario.

nucl-th↗

arXiv2018

Search for $W^{\prime}$ signal in single top quark production at LHC

The heavy charged gauge bosons were proposed in the theories beyond standard model. We explore the discovery potential for $W'\to t\bar{b}$ with top quark semi-leptonic decay at the LHC. We concentrate on the new physics signal search with the deviation from the standard model prediction if the resonance peak of $W'$ can not be observed directly. The events of signal with two jets plus one charged lepton and missing energy are simulated together with the dominant standard model backgrounds. In this paper, it is found that suitable cuts on the kinematic observables can effectively suppress the standard model backgrounds, so that it is possible to search for $W'$ signal at the LHC with its mass less than 6.6 TeV.

hep-ph↗