Researches on dialogue empathy aim to endow an agent with the capacity of accurate understanding and proper responding for emotions. Existing models for empathetic dialogue generation focus on the emotion flow in one direction, that is, from the cont
ext to response. We argue that conducting an empathetic conversation is a bidirectional process, where empathy occurs when the emotions of two interlocutors could converge on the same point, i.e., reaching an emotion consensus. Besides, we also find that the empathetic dialogue corpus is extremely limited, which further restricts the model performance. To address the above issues, we propose a dual-generative model, Dual-Emp, to simultaneously construct the emotion consensus and utilize some external unpaired data. Specifically, our model integrates a forward dialogue model, a backward dialogue model, and a discrete latent variable representing the emotion consensus into a unified architecture. Then, to alleviate the constraint of paired data, we extract unpaired emotional data from open-domain conversations and employ Dual-Emp to produce pseudo paired empathetic samples, which is more efficient and low-cost than the human annotation. Automatic and human evaluations demonstrate that our method outperforms competitive baselines in producing coherent and empathetic responses.
Being able to reply with a related, fluent, and informative response is an indispensable requirement for building high-quality conversational agents. In order to generate better responses, some approaches have been proposed, such as feeding extra inf
ormation by collecting large-scale datasets with human annotations, designing neural conversational models (NCMs) with complex architecture and loss functions, or filtering out untrustworthy samples based on a dialogue attribute, e.g., Relatedness or Genericness. In this paper, we follow the third research branch and present a data filtering method for open-domain dialogues, which identifies untrustworthy samples from training data with a quality measure that linearly combines seven dialogue attributes. The attribute weights are obtained via Bayesian Optimization (BayesOpt) that aims to optimize an objective function for dialogue generation iteratively on the validation set. Then we score training samples with the quality measure, sort them in descending order, and filter out those at the bottom. Furthermore, to accelerate the filter-train-evaluate iterations involved in BayesOpt on large-scale datasets, we propose a training framework that integrates maximum likelihood estimation (MLE) and negative training method (NEG). The training method updates parameters of a trained NCMs on two small sets with newly maintained and removed samples, respectively. Specifically, MLE is applied to maximize the log-likelihood of newly maintained samples, while NEG is used to minimize the log-likelihood of newly removed ones. Experimental results on two datasets show that our method can effectively identify untrustworthy samples, and NCMs trained on the filtered datasets achieve better performance.
Open-domain dialogue generation in natural language processing (NLP) is by default a pure-language task, which aims to satisfy human need for daily communication on open-ended topics by producing related and informative responses. In this paper, we p
oint out that hidden images, named as visual impressions (VIs), can be explored from the text-only data to enhance dialogue understanding and help generate better responses. Besides, the semantic dependency between an dialogue post and its response is complicated, e.g., few word alignments and some topic transitions. Therefore, the visual impressions of them are not shared, and it is more reasonable to integrate the response visual impressions (RVIs) into the decoder, rather than the post visual impressions (PVIs). However, both the response and its RVIs are not given directly in the test process. To handle the above issues, we propose a framework to explicitly construct VIs based on pure-language dialogue datasets and utilize them for better dialogue understanding and generation. Specifically, we obtain a group of images (PVIs) for each post based on a pre-trained word-image mapping model. These PVIs are used in a co-attention encoder to get a post representation with both visual and textual information. Since the RVIs are not provided directly during testing, we design a cascade decoder that consists of two sub-decoders. The first sub-decoder predicts the content words in response, and applies the word-image mapping model to get those RVIs. Then, the second sub-decoder generates the response based on the post and RVIs. Experimental results on two open-domain dialogue datasets show that our proposed approach achieves superior performance over competitive baselines.
Generating some appealing questions in open-domain conversations is an effective way to improve human-machine interactions and lead the topic to a broader or deeper direction. To avoid dull or deviated questions, some researchers tried to utilize ans
wer, the future information, to guide question generation. However, they separate a post-question-answer (PQA) triple into two parts: post-question (PQ) and question-answer (QA) pairs, which may hurt the overall coherence. Besides, the QA relationship is modeled as a one-to-one mapping that is not reasonable in open-domain conversations. To tackle these problems, we propose a generative triple-wise model with hierarchical variations for open-domain conversational question generation (CQG). Latent variables in three hierarchies are used to represent the shared background of a triple and one-to-many semantic mappings in both PQ and QA pairs. Experimental results on a large-scale CQG dataset show that our method significantly improves the quality of questions in terms of fluency, coherence and diversity over competitive baselines.
The modulation of the electronic structure by an external magnetic field, which could further control the electronic transport behaviour of a system, is highly desired. Herein, an unconventional anomalous Hall effect (UAHE) was observed during magnet
ization process in the magnetic Weyl semimetal EuB6, resulting in an unconventional anomalous Hall-conductivity as high as ~1000 {Omega}-1 cm-1 and a Hall-angle up to ~10%. The system even only shows the UAHE, meaning that the anomalous Hall signal completely comes from the UAHE, with UAHE accounting for 100% and 87.5% of the AHE and the total Hall response, respectively. Theoretical calculations revealed that a largely enhanced Berry curvature was induced by the dynamic folding of the topological bands due to the spin-canting effect under external magnetic fields, which further produced the prominent UAHE even in a low-field magnetization process. These findings elucidate the connection between the non-collinear magnetism and the topological electronic state as well as reveal a novel manner to manipulate the transport behaviour of topological electrons.
The hydrostatic pressure is expected to be an effective knob to tune the magnetostructural phase transitions of hexagonal MMX alloy. In this study, magnetization measurements under hydrostatic pressure were performed on a MMX martensitic MnNi0.77Fe0.
23Ge alloy. The magnetostructural transition temperature can be efficiently tuned to lower temperatures by applying moderate pressures, with a giant shift rate of -151 K GPa-1. A temperature span of 30 K is obtained under the pressure, within which a large magnetic entropy change of -23 J kg-1 K-1 in a field change of 5 T is induced by the mechanical energy gain due to the large volume change. Meanwhile, a decoupling of structural and magnetic transitions is observed at low temperatures when the martensitic transition temperature is lower than the Curie temperature. These results show a multi-parameter tunable caloric effect that benefits the solid-state cooling.
The topological metal states in electronic systems have been extensively studied, but topological phonons were explored only in few examples so far. Here, we expose for the first time that the topological nodal gimbal phonons, type-I and type-II Weyl
phonons are simultaneously present in T-carbon, a recently realized new allotrope of carbon. At about 15.2 THz, we find that there exist three mutually intersecting nodal loops (named as nodal gimbal phonons) around {Gamma} point, and two pairs of type-I Weyl phonons on the boundary of Brillouin zone around each X point. In addition, there exist three pairs of type-II Weyl phonons at about 14.5 THz around each L point. It is shown that these exotic topological phonons are protected by corresponding symmetries, and lead to topologically nontrivial surface states. Our findings not only afford plenty of intriguing topological phonon states in a simple material like T-carbon but also provide a new platform to study novel properties of topological phonons, which would facilitate further both experimental and theoretical works in future.
A fundamental dichotomous classification for all physical systems is according to whether they are spinless or spinful. This is especially crucial for the study of symmetry-protected topological phases, as the two classes have distinct symmetry algeb
ra. As a prominent example, the spacetime inversion symmetry $PT$ satisfies $(PT)^2=pm 1$ for spinless/spinful systems, and each class features unique topological phases. Here, we reveal a possibility to switch the two fundamental classes via $mathbb{Z}_2$ projective representations. For $PT$ symmetry, this occurs when $P$ inverses the gauge transformation needed to recover the original $mathbb{Z}_2$ gauge connections under $P$. As a result, we can achieve topological phases originally unique for spinful systems in a spinless system, and vice versa. We explicitly demonstrate the claimed mechanism with several concrete models, such as Kramers degenerate bands and Kramers Majorana boundary modes in spinless systems, and real topological phases in spinful systems. Possible experimental realization of these models is discussed. Our work breaks a fundamental limitation on topological phases and opens an unprecedented possibility to realize intriguing topological phases in previously impossible systems.
Anomalous Hall effect (AHE) can be induced by intrinsic mechanism due to the band Berry phase and extrinsic one arising from the impurity scattering. The recently discovered magnetic Weyl semimetal Co3Sn2S2 exhibits a large intrinsic anomalous Hall c
onductivity (AHC) and a giant anomalous Hall angle (AHA). The predicted energy dependence of the AHC in this material exhibits a plateau at 1000 {Omega}-1 cm-1 and an energy width of 100 meV just below EF, thereby implying that the large intrinsic AHC will not significantly change against small-scale energy disturbances such as slight p-doping. Here, we successfully trigger the extrinsic contribution from alien-atom scattering in addition to the intrinsic one of the pristine material by introducing a small amount of Fe dopant to substitute Co in Co3Sn2S2. Our experimental results show that the AHC and AHA can be prominently enhanced up to 1850 {Omega}-1 cm-1 and 33%, respectively, owing to the synergistic contributions from the intrinsic and extrinsic mechanisms as distinguished by the TYJ model. In particular, the tuned AHA holds a record value in low fields among known magnetic materials. This study opens up a pathway to engineer giant AHE in magnetic Weyl semimetals, thereby potentially advancing the topological spintronics/Weyltronics.
We report a comprehensive neutron scattering study on the spin excitations in the magnetic Weyl semimetal Co$_3$Sn$_2$S$_2$ with quasi-two-dimensional structure. Both in-plane and out-of-plane dispersions of the spin waves are revealed in the ferroma
gnetic state, similarly dispersive but damped spin excitations persist into the paramagnetic state. The effective exchange interactions have been estimated by a semi-classical Heisenberg model to consistently reproduce the experimental $T_C$ and spin stiffness. However, a full spin wave gap below $E_g=2.3$ meV is observed at $T=4$ K, much larger than the estimated magnetic anisotropy energy ($sim0.6$ meV), while its temperature dependence indicates a significant contribution from the Weyl fermions. These results suggest that Co$_3$Sn$_2$S$_2$ is a three-dimensional correlated system with large spin stiffness, and the low-energy spin dynamics could interplay with the topological electron states.
