arXiv subjects

Lin Pan

Publications and source records attributed to Lin Pan.

23 records · Page 2Linked to original sources

arXiv2019

Frustratingly Easy Natural Question Answering

Existing literature on Question Answering (QA) mostly focuses on algorithmic novelty, data augmentation, or increasingly large pre-trained language models like XLNet and RoBERTa. Additionally, a lot of systems on the QA leaderboards do not have associated research documentation in order to successfully replicate their experiments. In this paper, we outline these algorithmic components such as Attention-over-Attention, coupled with data augmentation and ensembling strategies that have shown to yield state-of-the-art results on benchmark datasets like SQuAD, even achieving super-human performance. Contrary to these prior results, when we evaluate on the recently proposed Natural Questions benchmark dataset, we find that an incredibly simple approach of transfer learning from BERT outperforms the previous state-of-the-art system trained on 4 million more examples than ours by 1.9 F1 points. Adding ensembling strategies further improves that number by 2.3 F1 points.

cs.CL↗

arXiv2019

Multi-Granular Text Encoding for Self-Explaining Categorization

Self-explaining text categorization requires a classifier to make a prediction along with supporting evidence. A popular type of evidence is sub-sequences extracted from the input text which are sufficient for the classifier to make the prediction. In this work, we define multi-granular ngrams as basic units for explanation, and organize all ngrams into a hierarchical structure, so that shorter ngrams can be reused while computing longer ngrams. We leverage a tree-structured LSTM to learn a context-independent representation for each unit via parameter sharing. Experiments on medical disease classification show that our model is more accurate, efficient and compact than BiLSTM and CNN baselines. More importantly, our model can extract intuitive multi-granular evidence to support its predictions.

cs.CL↗

arXiv2018

Coexisting surface states in the weak and crystalline topological insulator Bi_2TeI

The established diversity of electronic topology classes lends the opportunity to pair them into dual topological complexes. Bulk-surface correspondence then ensures the coexistence of a combination of boundary states that cannot be realized but only at the various surfaces of such a dual topological material. We show that the layered compound Bi_2TeI realizes a dual topological insulator. It exhibits band inversions at two time reversal symmetry points of the bulk band which classify it as a weak topological insulator with metallic states on its (010) 'side' surfaces. Additional mirror symmetry of the crystal structure concurrently classifies it as a topological crystalline insulator. Bi2TeI is therefore predicted to host a pair of Dirac cones protected by time reversal symmetry on its 'side' surfaces and three pairs of Dirac cones protected by mirror symmetry on its 'top' and 'bottom' (001) surfaces. We spectroscopically map the top cleaved surface of Bi_2TeI, and crystallographic step edges therein. We show the existence of both two dimensional surface states which are susceptible to mirror symmetry breaking, as well as one dimensional channels that reside along the step edges. Their mutual coexistence on the step edge where both facets join is facilitated by momentum and energy segregation. Our observations of a dual topological insulator make way to additional pairing of other dual topology classes with distinct surface manifestations coexisting at their boundaries.

cond-mat.mes-hall↗

arXiv2016

The role of ionized impurity scattering on the thermoelectric performances of rock salt AgPbmSnSe2+m

We report on the successful synthesis and on the properties of polycrystalline AgPbmSnSe2+m (m = ++++, 100, 50, 25) samples with a rock salt structure. Between 160 K and 400 K, the dominant scattering process of the carriers in this system changes from acoustic phonon scattering in PbSe to ionized impurity scattering in AgPbmSnSe2+m, which synergistically optimizes electrical and thermal transport properties. Thanks to the faint amount of AgSnSe2, the Seebeck coefficient is enhanced by boosting the scattering factor, the electric conductivity is improved by the increase of the concentration of holes coupled to a limited degradation of their mobility, and the total thermal conductivity is reduced by suppressing bipolar thermal conductivity. Therefore, ZT of AgPbmSnSe2+m (m = 50) reaches 1.3 at 889 K. The mechanism suggested in this study opens new paths to improve the thermoelectric performances of other families of materials.

cond-mat.mtrl-sci↗

arXiv2006

Geometric, electronic properties and the thermodynamics of pure and Al--doped Li clusters

The first--principles density functional molecular dynamics simulations have been carried out to investigate the geometric, the electronic, and the finite temperature properties of pure Li clusters (Li$_{10}$, Li$_{12}$) and Al--doped Li clusters (Li$_{10}$Al, Li$_{10}$Al$_2$). We find that addition of two Al impurities in Li$_{10}$ results in a substantial structural change, while the addition of one Al impurity causes a rearrangement of atoms. Introduction of Al--impurities in Li$_{10}$ establishes a polar bond between Li and nearby Al atom(s), leading to a multicentered bonding, which weakens the Li--Li metallic bonds in the system. These weakened Li--Li bonds lead to a premelting feature to occur at lower temperatures in Al--doped clusters. In Li$_{10}$Al$_2$, Al atoms also form a weak covalent bond, resulting into their dimer like behavior. This causes Al atoms not to `melt' till 800 K, in contrast to the Li atoms which show a complete diffusive behavior above 400 K. Thus, although one Al impurity in Li$_{10}$ cluster does not change its melting characteristics significantly, two impurities results in `surface melting' of Li atoms whose motions are confined around Al dimer.

cond-mat.mes-hall↗