arXiv subjects

Grace Yang

Publications and source records attributed to Grace Yang.

13 recordsLinked to original sources

arXiv2024

Comparative Analysis of Extrinsic Factors for NER in French

Named entity recognition (NER) is a crucial task that aims to identify structured information, which is often replete with complex, technical terms and a high degree of variability. Accurate and reliable NER can facilitate the extraction and analysis of important information. However, NER for other than English is challenging due to limited data availability, as the high expertise, time, and expenses are required to annotate its data. In this paper, by using the limited data, we explore various factors including model structure, corpus annotation scheme and data augmentation techniques to improve the performance of a NER model for French. Our experiments demonstrate that these approaches can significantly improve the model's F1 score from original CRF score of 62.41 to 79.39. Our findings suggest that considering different extrinsic factors and combining these techniques is a promising approach for improving NER performance where the size of data is limited.

cs.CL↗

arXiv2023

Is GPT Powerful Enough to Analyze the Emotions of Memes?

Large Language Models (LLMs), representing a significant achievement in artificial intelligence (AI) research, have demonstrated their ability in a multitude of tasks. This project aims to explore the capabilities of GPT-3.5, a leading example of LLMs, in processing the sentiment analysis of Internet memes. Memes, which include both verbal and visual aspects, act as a powerful yet complex tool for expressing ideas and sentiments, demanding an understanding of societal norms and cultural contexts. Notably, the detection and moderation of hateful memes pose a significant challenge due to their implicit offensive nature. This project investigates GPT's proficiency in such subjective tasks, revealing its strengths and potential limitations. The tasks include the classification of meme sentiment, determination of humor type, and detection of implicit hate in memes. The performance evaluation, using datasets from SemEval-2020 Task 8 and Facebook hateful memes, offers a comparative understanding of GPT responses against human annotations. Despite GPT's remarkable progress, our findings underscore the challenges faced by these models in handling subjective tasks, which are rooted in their inherent limitations including contextual understanding, interpretation of implicit meanings, and data biases. This research contributes to the broader discourse on the applicability of AI in handling complex, context-dependent tasks, and offers valuable insights for future advancements.

cs.CL↗

arXiv2023

SN 2022oqm: A Bright and Multi-peaked Calcium-rich Transient

We present the photometric and spectroscopic evolution of SN 2022oqm, a nearby multi-peaked hydrogen- and helium-weak calcium-rich transient (CaRT). SN 2022oqm was detected 13.1 kpc from its host galaxy, the face-on spiral galaxy NGC 5875. Extensive spectroscopic coverage reveals an early hot (T >= 40,000 K) continuum and carbon features observed $\sim$1~day after discovery, SN Ic-like photospheric-phase spectra, and strong forbidden calcium emission starting 38 days after discovery. SN 2022oqm has a relatively high peak luminosity (MB = -17 mag) for (CaRTs), making it an outlier in the population. We determine that three power sources are necessary to explain the light curve (LC), with each corresponding to a distinct peak. The first peak is powered by an expanding blackbody with a power law luminosity, suggesting shock cooling by circumstellar material (CSM). Subsequent LC evolution is powered by a double radioactive decay model, consistent with two sources of photons diffusing through optically thick ejecta. From the LC, we derive an ejecta mass and 56Ni mass of ~0.6 solar masses and ~0.09 solar masses. Spectroscopic modeling suggests 0.6 solar masses of ejecta, and with well-mixed Fe-peak elements throughout. We discuss several physical origins for SN 2022oqm and find either a surprisingly massive white dwarf progenitor or a peculiar stripped envelope model could explain SN 2022oqm. A stripped envelope explosion inside a dense, hydrogen- and helium-poor CSM, akin to SNe Icn, but with a large 56Ni mass and small CSM mass could explain SN 2022oqm. Alternatively, helium detonation on an unexpectedly massive white dwarf could also explain SN 2022oqm.

astro-ph.HE↗

arXiv2023

Type II-P Supernova Progenitor Star Initial Masses and SN 2020jfo: Direct Detection, Light Curve Properties, Nebular Spectroscopy, and Local Environment

We present optical, ultraviolet, and infrared data of the type II supernova (SN II) 2020jfo at 14.5 Mpc. This wealth of multiwavelength data allows to compare different metrics commonly used to estimate progenitor masses of SN II for the same object. Using its early light curve, we infer SN 2020jfo had a progenitor radius of $\approx$700 $R_{\odot}$, consistent with red supergiants of initial mass $M_{\rm ZAMS}=$11-13 $M_{\odot}$. The decline in its late-time light curve is best fit by a ${}^{56}$Ni mass of 0.018$\pm$0.007 $M_{\odot}$ consistent with that ejected from SN II-P with $\approx$13 $M_{\odot}$ initial mass stars. Early spectra and photometry do not exhibit signs of interaction with circumstellar matter, implying that SN 2020jfo experienced weak mass loss within the final years prior to explosion. Our spectra at $>$250 days are best fit by models from 12 $M_{\odot}$ initial mass stars. We analyzed integral field unit spectroscopy of the stellar population near SN 2020jfo, finding its massive star population had a zero age main sequence mass of 9.7$\substack{+2.5\\-1.3} M_{\odot}$. We identify a single counterpart in pre-explosion imaging and find it has an initial mass of at most $7.2\substack{+1.2\\-0.6} M_{\odot}$. We conclude that the inconsistency between this mass and indirect mass indicators from SN 2020jfo itself is most likely caused by extinction with $A_{V}=2$-3 mag due to matter around the progenitor star, which lowered its observed optical luminosity. As SN 2020jfo did not exhibit extinction at this level or evidence for interaction with circumstellar matter between 1.6-450 days from explosion, we conclude that this material was likely confined within $\approx$3000 $R_{\odot}$ from the progenitor star.

astro-ph.SR↗

arXiv2023

An Asymmetric Double-Degenerate Type Ia Supernova Explosion with a Surviving Companion Star

We present nebular spectroscopy of SN 2020hvf, a Type Ia supernova (SN Ia) with an early bump in its light curve. SN 2020hvf shares many spectroscopic and photometric similarities to the carbon-rich high-luminosity "03fg-like" SNe Ia. At $>$240 days after peak brightness, we detect unambiguous emission from [Ca II] $\lambda\lambda$7291, 7324 which is never observed in normal-SNe Ia and only seen in peculiar subclasses. SN 2020hvf displays "saw-tooth" emission profiles near 7300 A that cannot be explained with single symmetric velocity components of [Fe II], [Ni II], and [Ca II], indicating an asymmetric explosion. The broad [Ca II] emission is best modeled by two velocity components offset by 1,220 km s$^{-1}$, which could be caused by ejecta associated with each star in the progenitor system, separated by their orbital velocity. For the first time in a SN Ia, we identify narrow (${\rm FWHM} = 180\pm40$ km s$^{-1}$) [Ca II] emission, which we associate with a wind from a surviving, puffed-up companion star. Few published spectra have sufficient resolution and signal-to-noise ratio necessary to detect similar narrow [Ca II] emission, however, we have detected similar line profiles in other 03fg-like SNe Ia. The extremely narrow velocity width of [Ca II] has only otherwise been observed in SNe Iax at late times. Since this event likely had a double-degenerate "super-Chandrasekhar" mass progenitor system, we suggest that a single white dwarf (WD) was fully disrupted and a wind from a surviving companion WD is producing the observed narrow emission. It is unclear if this unique progenitor and explosion scenario can explain the diversity of 03fg-like SNe Ia, potentially indicating that multiple progenitor channels contribute to this subclass.

astro-ph.HE↗

arXiv2023

Observational properties of a bright type Iax SN 2018cni and a faint type Iax SN 2020kyg

We present the optical photometric and spectroscopic analysis of two type Iax SNe 2018cni and 2020kyg. SN 2018cni is a bright type Iax SN (M$_{V,peak}$ = $-$17.81$\pm$0.21 mag) whereas SN 2020kyg (M$_{V,peak}$ = $-$14.52$\pm$0.21 mag) is a faint one. We derive $^{56}$Ni mass of 0.07 and 0.002 M${_\odot}$, ejecta mass of 0.48 and 0.14 M${_\odot}$ for SNe 2018cni and 2020kyg, respectively. A combined study of the bright and faint type Iax SNe in $R/r$- band reveals that the brighter objects tend to have a longer rise time. However, the correlation between the peak luminosity and decline rate shows that bright and faint type Iax SNe exhibit distinct behaviour. Comparison with standard deflagration models suggests that SN 2018cni is consistent with the deflagration of a CO white dwarf whereas the properties of SN 2020kyg can be better explained by the deflagration of a hybrid CONe white dwarf. The spectral features of both the SNe point to the presence of similar chemical species but with different mass fractions. Our spectral modelling indicates stratification at the outer layers and mixed inner ejecta for both the SNe.

astro-ph.HE↗

arXiv2023

Flight of the Bumblebee: the Early Excess Flux of Type Ia Supernova 2023bee revealed by $TESS$, $Swift$ and Young Supernova Experiment Observations

We present high-cadence ultraviolet through near-infrared observations of the Type Ia supernova (SN Ia) 2023bee in NGC~2708 ($D = 32 \pm 3$ Mpc), finding excess flux in the first days after explosion relative to the expected power-law rise from an expanding fireball. This deviation from typical behavior for SNe Ia is particularly obvious in our 10-minute cadence $TESS$ light curve and $Swift$ UV data. Compared to a few other normal SNe Ia with detected early excess flux, the excess flux in SN 2023bee is redder in the UV and less luminous. We present optical spectra of SN 2023bee, including two spectra during the period where the flux excess is dominant. At this time, the spectra are similar to those of other SNe Ia but with weaker Si II, C II and Ca II absorption lines, perhaps because the excess flux creates a stronger continuum. We compare the data to several theoretical models that have been proposed to explain the early flux excess in SNe Ia. Interaction with either a nearby companion star or close-in circumstellar material is expected to produce a faster evolution than seen in the data. Radioactive material in the outer layers of the ejecta, either from a double detonation explosion or simply an explosion with a $^{56}$Ni clump near the surface, can not fully reproduce the evolution either, likely due to the sensitivity of early UV observable to the treatment of the outer part of ejecta in simulation. We conclude that no current model can adequately explain the full set of observations. We find that a relatively large fraction of nearby, bright SNe Ia with high-cadence observations have some amount of excess flux within a few days of explosion. Considering potential asymmetric emission, the physical cause of this excess flux may be ubiquitous in normal SNe Ia.

astro-ph.HE↗

arXiv2022

Language Model Classifier Aligns Better with Physician Word Sensitivity than XGBoost on Readmission Prediction

Traditional evaluation metrics for classification in natural language processing such as accuracy and area under the curve fail to differentiate between models with different predictive behaviors despite their similar performance metrics. We introduce sensitivity score, a metric that scrutinizes models' behaviors at the vocabulary level to provide insights into disparities in their decision-making logic. We assess the sensitivity score on a set of representative words in the test set using two classifiers trained for hospital readmission classification with similar performance statistics. Our experiments compare the decision-making logic of clinicians and classifiers based on rank correlations of sensitivity scores. The results indicate that the language model's sensitivity score aligns better with the professionals than the xgboost classifier on tf-idf embeddings, which suggests that xgboost uses some spurious features. Overall, this metric offers a novel perspective on assessing models' robustness by quantifying their discrepancy with professional opinions. Our code is available on GitHub (https://github.com/nyuolab/Model_Sensitivity).

cs.CL↗

arXiv2022

Weak Mass Loss from the Red Supergiant Progenitor of the Type II SN 2021yja

We present high-cadence optical, ultraviolet (UV), and near-infrared data of the nearby ($D\approx23$ Mpc) Type II supernova (SN) 2021yja. Many Type II SNe show signs of interaction with circumstellar material (CSM) during the first few days after explosion, implying that their red supergiant (RSG) progenitors experience episodic or eruptive mass loss. However, because it is difficult to discover SNe early, the diversity of CSM configurations in RSGs has not been fully mapped. SN 2021yja, first detected within ${\approx}5.4$ hours of explosion, shows some signatures of CSM interaction (high UV luminosity, radio and x-ray emission) but without the narrow emission lines or early light curve peak that can accompany CSM. Here we analyze the densely sampled early light curve and spectral series of this nearby SN to infer the properties of its progenitor and CSM. We find that the most likely progenitor was an RSG with an extended envelope, encompassed by low-density CSM. We also present archival Hubble Space Telescope imaging of the host galaxy of SN 2021yja, which allows us to place a stringent upper limit of ${\lesssim}9\ M_\odot$ on the progenitor mass. However, this is in tension with some aspects of the SN evolution, which point to a more massive progenitor. Our analysis highlights the need to consider progenitor structure when making inferences about CSM properties, and that a comprehensive view of CSM tracers should be made to give a fuller view of the last years of RSG evolution.

astro-ph.HE↗

arXiv2022

The Renovated Thacher Observatory and First Science Results

Located on the campus of the Thacher School in Southern California, the Thacher Observatory has a legacy of astronomy research and education that dates back to the late 1950's. In 2016, the observatory was fully renovated with upgrades including a new 0.7-m telescope, a research grade camera, and a slit dome with full automation capabilities. The low-elevation site is bordered by the Los Padres National Forest and therefore affords dark to very dark skies allowing for accurate and precise photometric observations. We present a characterization of the site including sky brightness, weather, and seeing, and we demonstrate the on-sky performance of the facility. Our primary research programs are based around our multi-band photometric capabilities and include photometric monitoring of variable sources, a nearby supernova search and followup program, a quick response transient followup effort, and exoplanet and eclipsing binary light curves. Select results from these programs are included in this work which highlight the broad range of science available to an automated observatory with a moderately sized telescope.

astro-ph.IM↗

arXiv2017

With the Brexit referendum Britain has spoken or has it

The Brexit referendum on the United Kingdom membership of the European Union took place on 23 June 2016. On the basis of a 52-48 split, statements such as: the majority of the UK chose to leave the EU, or the British people have voted to leave the European Union, or the will of the British people is to leave the EU, have pervaded political discourse and newspaper articles. In a recent paper, the authors present scientific arguments to contend that all those statements are without basis. They point out the fundamental issues with the Brexit referendum, which lie in how the data was summarized, ignoring fundamental statistical principles, and in the lack of terms of reference fixing a threshold for the result and for the minimum turnout. If carefully planned, the actual division of the United Kingdom would have been properly captured and the result---either way, for Leave or Remain---would have been readily accepted. In the present paper, the authors summarize the discussion and add some context in non-technical terms.

physics.soc-ph↗

arXiv2016

What did the 2016 Brexit referendum data really say?

The Brexit referendum took place in the UK in June, 2016. The unweighted percentage of leavers over the whole population was 51.9%. In this paper, first, we demonstrate that a 52%-48% split represents only a difference that is not sufficiently different from a 50-50 split to claim a majority for either side. Second, and most important, on this basis of the unweighted percentage, statement like: The country voted to leave the EU, were made. When a statement about a population is made based on a subset of it (the turnout rate for Brexit was only 72% and therefore 37% of the eligible population voted Leave), it comes with an element of uncertainty that should not be ignored. The unweighted average disregards, not only between-region heterogeneity but also within-region variability. Our analysis, controlling for both, finds that the split of the Brexit is of negligible material significance and do not indicate majority for either side.

stat.AP↗

arXiv2012

Empirical Likelihood for Right Censored Lifetime Data

This paper considers the empirical likelihood (EL) construction of confidence intervals for a linear functional based on right censored lifetime data. Many of the results in literature show that log EL has a limiting scaled chi-square distribution, where the scale parameter is a function of the unknown asymptotic variance. The scale parameter has to be estimated for the construction. Additional estimation would reduce the coverage accuracy for the parameter. This diminishes a main advantage of the EL method for censored data. By utilizing certain influence functions in an estimating equation, it is shown that under very general conditions, log EL converges weakly to a standard chi-square distribution and thereby eliminates the need for estimating the scale parameter. Moreover, a special way of employing influence functions eases the otherwise very demanding computations of the EL method. Our approach yields smaller asymptotic variance of the influence function than those comparable ones considered by Wang and Jing (2001) and Qin and Zhao (2007). Thus it is not surprising that confidence intervals using influence functions give a better coverage accuracy as demonstrated by simulations.

math.ST↗