id
stringlengths
33
34
title
stringlengths
6
218
authors
stringlengths
5
6.98k
categories
stringlengths
5
188
summary
stringlengths
34
3.26k
link_pdf
stringlengths
33
34
updated
timestamp[s]
published
timestamp[s]
embeddings
dict
http://arxiv.org/abs/2212.09721v2
KNIFE: Distilling Reasoning Knowledge From Free-Text Rationales
Aaron Chan, Zhiyuan Zeng, Wyatt Lake, Brihi Joshi, Hanjie Chen, Xiang Ren
cs.CL, cs.AI, cs.LG
Language models (LMs) have yielded impressive results on many language reasoning tasks, but their unexpected errors raise doubts about their reasoning abilities. In light of this, there is growing interest in finetuning/prompting LMs with both task instances and their associated free-text rationales (FTRs), which expla...
http://arxiv.org/pdf/2212.09721v2
2023-05-22T00:19:21
2022-12-19T18:49:09
{ "title": [ -0.021942139, -0.024642944, -0.043914795, -0.0715332, -0.012077332, -0.006668091, -0.022918701, 0.0075416565, 0.008804321, 0.010269165, -0.030532837, 0.006046295, -0.0068893433, -0.016616821, -0.029327393, -0.016052246, 0.0015096664, ...
http://arxiv.org/abs/2212.09710v1
Continual Learning for Instruction Following from Realtime Feedback
Alane Suhr, Yoav Artzi
cs.CL, cs.AI, cs.LG
We study the problem of continually training an instruction-following agent through feedback provided by users during collaborative interactions. During interaction, human users instruct an agent using natural language, and provide realtime binary feedback as they observe the agent's instruction execution. We cast lear...
http://arxiv.org/pdf/2212.09710v1
2022-12-19T18:39:43
2022-12-19T18:39:43
{ "title": [ 0.0051841736, -0.00013124943, -0.015716553, -0.06109619, 0.0016937256, -0.019805908, -0.017959595, -0.025100708, 0.0057525635, 0.034179688, -0.016708374, 0.0067329407, -0.015838623, -0.01965332, -0.037322998, 0.0068969727, 0.022979736, ...
http://arxiv.org/abs/2212.09702v3
On Event Individuation for Document-Level Information Extraction
William Gantt, Reno Kriz, Yunmo Chen, Siddharth Vashishtha, Aaron Steven White
cs.CL, cs.AI, cs.LG
As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of template filling has seen renewed interest as benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demand...
http://arxiv.org/pdf/2212.09702v3
2023-10-20T21:26:29
2022-12-19T18:30:36
{ "title": [ -0.014305115, -0.0025615692, -0.033599854, -0.08453369, 0.020751953, 0.010185242, -0.040283203, 0.0049743652, 0.011726379, 0.011253357, -0.023406982, 0.0061416626, -0.0015621185, -0.007129669, -0.026107788, -0.0030403137, -0.01612854, ...
http://arxiv.org/abs/2212.09701v1
Graph-based Semantical Extractive Text Analysis
Mina Samizadeh
cs.CL, cs.LG
In the past few decades, there has been an explosion in the amount of available data produced from various sources with different topics. The availability of this enormous data necessitates us to adopt effective computational tools to explore the data. This leads to an intense growing interest in the research community...
http://arxiv.org/pdf/2212.09701v1
2022-12-19T18:30:26
2022-12-19T18:30:26
{ "title": [ -0.043121338, -0.0070762634, -0.055023193, -0.04284668, -0.003435135, -0.001496315, -0.041412354, -0.014083862, 0.006954193, 0.027313232, -0.006313324, 0.019943237, -0.016830444, -0.011734009, -0.05429077, 0.009666443, -0.015029907, ...
http://arxiv.org/abs/2212.09700v2
Resolving Open-textured Rules with Templated Interpretive Arguments
John Licato, Logan Fields, Zaid Marji
cs.CL
Open-textured terms in written rules are typically settled through interpretive argumentation. Ongoing work has attempted to catalogue the schemes used in such interpretive argumentation. But how can the use of these schemes affect the way in which people actually use and reason over the proper interpretations of open-...
http://arxiv.org/pdf/2212.09700v2
2023-02-03T16:41:06
2022-12-19T18:30:09
{ "title": [ 0.0025920868, -0.009086609, -0.023651123, -0.10656738, -0.0072288513, 0.029464722, -0.0793457, 0.002023697, -0.030227661, 0.019454956, -0.018096924, 0.004611969, -0.015808105, -0.03152466, -0.007858276, -0.030288696, 0.014640808, -0....
http://arxiv.org/abs/2212.09699v3
SegAugment: Maximizing the Utility of Speech Translation Data with Segmentation-based Augmentations
Ioannis Tsiamas, José A. R. Fonollosa, Marta R. Costa-jussà
cs.CL, cs.SD, eess.AS
End-to-end Speech Translation is hindered by a lack of available data resources. While most of them are based on documents, a sentence-level version is available, which is however single and static, potentially impeding the usefulness of the data. We propose a new data augmentation strategy, SegAugment, to address this...
http://arxiv.org/pdf/2212.09699v3
2023-11-01T14:18:40
2022-12-19T18:29:31
{ "title": [ -0.021652222, -0.010643005, -0.02709961, -0.061523438, -0.01826477, 0.00630188, -0.043945312, 0.0138168335, -0.03704834, 0.046295166, -0.00856781, 0.033966064, 0.003107071, -0.01499939, -0.029891968, 0.018127441, 0.009407043, 0.03004...
http://arxiv.org/abs/2212.09689v1
Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor
Or Honovich, Thomas Scialom, Omer Levy, Timo Schick
cs.CL, cs.AI, cs.LG
Instruction tuning enables pretrained language models to perform new tasks from inference-time natural language descriptions. These approaches rely on vast amounts of human supervision in the form of crowdsourced datasets or user interactions. In this work, we introduce Unnatural Instructions: a large dataset of creati...
http://arxiv.org/pdf/2212.09689v1
2022-12-19T18:21:00
2022-12-19T18:21:00
{ "title": [ -0.047698975, -0.019241333, -0.03805542, -0.043518066, -0.016921997, -0.0011129379, -0.055023193, -0.00061893463, -0.009437561, 0.022201538, -0.015655518, 0.004146576, 0.005027771, -0.014350891, -0.040100098, 0.0067825317, -0.007133484, ...
http://arxiv.org/abs/2212.09686v2
A Natural Bias for Language Generation Models
Clara Meister, Wojciech Stokowiec, Tiago Pimentel, Lei Yu, Laura Rimell, Adhiguna Kuncoro
cs.CL
After just a few hundred training updates, a standard probabilistic model for language generation has likely not yet learnt many semantic or syntactic rules of natural language, making it difficult to estimate the probability distribution over next tokens. Yet around this point, these models have identified a simple, l...
http://arxiv.org/pdf/2212.09686v2
2023-06-23T05:59:16
2022-12-19T18:14:36
{ "title": [ -0.04711914, -0.02166748, -0.048583984, -0.049713135, -0.008041382, 0.0073776245, -0.07055664, 0.007434845, 0.0071487427, 0.038116455, -0.015655518, -0.006500244, 0.011413574, -0.012428284, -0.04788208, 0.013504028, -0.019058228, 0.0...
http://arxiv.org/abs/2212.09683v4
Human-in-the-loop Evaluation for Early Misinformation Detection: A Case Study of COVID-19 Treatments
Ethan Mendes, Yang Chen, Wei Xu, Alan Ritter
cs.CL
We present a human-in-the-loop evaluation framework for fact-checking novel misinformation claims and identifying social media messages that support them. Our approach extracts check-worthy claims, which are aggregated and ranked for review. Stance classifiers are then used to identify tweets supporting novel misinform...
http://arxiv.org/pdf/2212.09683v4
2023-07-03T07:04:45
2022-12-19T18:11:10
{ "title": [ 0.015098572, 0.002254486, -0.029373169, -0.04547119, 0.003112793, -0.02017212, 0.006565094, -0.0015306473, -0.027297974, 0.02961731, -0.013504028, 0.002910614, -0.0029716492, -0.040100098, -0.02053833, -0.015686035, -0.0090408325, 0....
http://arxiv.org/abs/2212.09682v1
Multilingual Sequence-to-Sequence Models for Hebrew NLP
Matan Eyal, Hila Noga, Roee Aharoni, Idan Szpektor, Reut Tsarfaty
cs.CL
Recent work attributes progress in NLP to large language models (LMs) with increased model size and large quantities of pretraining data. Despite this, current state-of-the-art LMs for Hebrew are both under-parameterized and under-trained compared to LMs in other languages. Additionally, previous work on pretrained Heb...
http://arxiv.org/pdf/2212.09682v1
2022-12-19T18:10:23
2022-12-19T18:10:23
{ "title": [ -0.010643005, -0.0060424805, -0.014572144, -0.095458984, 0.011833191, 0.034057617, -0.06451416, 0.023254395, 0.0021457672, 0.027709961, -0.007648468, 0.02192688, 0.02128601, -0.01889038, -0.01687622, 0.027542114, -0.016723633, 0.0407...
http://arxiv.org/abs/2212.09676v2
Words as Gatekeepers: Measuring Discipline-specific Terms and Meanings in Scholarly Publications
Li Lucy, Jesse Dodge, David Bamman, Katherine A. Keith
cs.CL, cs.DL
Scholarly text is often laden with jargon, or specialized language that can facilitate efficient in-group communication within fields but hinder understanding for out-groups. In this work, we develop and validate an interpretable approach for measuring scholarly jargon from text. Expanding the scope of prior work which...
http://arxiv.org/pdf/2212.09676v2
2023-05-23T02:24:54
2022-12-19T18:01:06
{ "title": [ -0.066223145, 0.011047363, -0.06951904, 0.008026123, 0.010871887, -0.022247314, -0.008651733, -0.041900635, -0.016952515, 0.02545166, -0.005203247, 0.040161133, -0.033996582, -0.0058174133, -0.04647827, -0.0018062592, -0.010627747, 0...
http://arxiv.org/abs/2212.09674v2
LR-Sum: Summarization for Less-Resourced Languages
Chester Palen-Michel, Constantine Lignos
cs.CL
This preprint describes work in progress on LR-Sum, a new permissively-licensed dataset created with the goal of enabling further research in automatic summarization for less-resourced languages. LR-Sum contains human-written summaries for 40 languages, many of which are less-resourced. We describe our process for extr...
http://arxiv.org/pdf/2212.09674v2
2023-10-26T19:50:29
2022-12-19T18:00:09
{ "title": [ -0.014785767, -0.026245117, -0.047851562, -0.11138916, 0.0032672882, 0.012107849, -0.06640625, 0.02255249, -0.008705139, 0.03945923, -0.027679443, 0.029541016, -0.004558563, 0.00035595894, -0.045959473, 0.007171631, 0.0019779205, 0.0...
http://arxiv.org/abs/2212.09670v1
StyleFlow: Disentangle Latent Representations via Normalizing Flow for Unsupervised Text Style Transfer
Kangchen Zhu, Zhiliang Tian, Ruifeng Luo, Xiaoguang Mao
cs.CL
Text style transfer aims to alter the style of a sentence while preserving its content. Due to the lack of parallel corpora, most recent work focuses on unsupervised methods and often uses cycle construction to train models. Since cycle construction helps to improve the style transfer ability of the model by rebuilding...
http://arxiv.org/pdf/2212.09670v1
2022-12-19T17:59:18
2022-12-19T17:59:18
{ "title": [ -0.030563354, -0.0027370453, -0.06500244, -0.046081543, -0.000600338, -0.0047569275, -0.03515625, -0.0006990433, 0.0067749023, 0.048553467, -0.016937256, 0.04852295, 0.021392822, -0.0000628829, -0.012893677, 0.028411865, 0.009803772, ...
http://arxiv.org/abs/2212.09667v2
Foveate, Attribute, and Rationalize: Towards Physically Safe and Trustworthy AI
Alex Mei, Sharon Levy, William Yang Wang
cs.CL, cs.AI, cs.LG
Users' physical safety is an increasing concern as the market for intelligent systems continues to grow, where unconstrained systems may recommend users dangerous actions that can lead to serious injury. Covertly unsafe text is an area of particular interest, as such text may arise from everyday scenarios and are chall...
http://arxiv.org/pdf/2212.09667v2
2023-05-19T05:19:18
2022-12-19T17:51:47
{ "title": [ -0.022338867, 0.005809784, -0.04638672, -0.054016113, 0.022338867, -0.017913818, -0.008132935, -0.022216797, -0.040161133, 0.034576416, -0.016113281, 0.022521973, 0.0054893494, -0.014167786, -0.022598267, -0.00843811, -0.0006980896, ...
http://arxiv.org/abs/2212.09666v1
MultiCoder: Multi-Programming-Lingual Pre-Training for Low-Resource Code Completion
Zi Gong, Yinpeng Guo, Pingyi Zhou, Cuiyun Gao, Yasheng Wang, Zenglin Xu
cs.CL
Code completion is a valuable topic in both academia and industry. Recently, large-scale mono-programming-lingual (MonoPL) pre-training models have been proposed to boost the performance of code completion. However, the code completion on low-resource programming languages (PL) is difficult for the data-driven paradigm...
http://arxiv.org/pdf/2212.09666v1
2022-12-19T17:50:05
2022-12-19T17:50:05
{ "title": [ -0.02267456, -0.034942627, -0.035705566, -0.089660645, -0.004486084, 0.00818634, -0.066223145, 0.020599365, -0.004512787, 0.042785645, -0.00014972687, 0.027542114, 0.0076828003, -0.023101807, -0.022949219, 0.009796143, -0.012062073, ...
http://arxiv.org/abs/2212.09663v3
Norm of Word Embedding Encodes Information Gain
Momose Oyama, Sho Yokoi, Hidetoshi Shimodaira
cs.CL
Distributed representations of words encode lexical semantic information, but what type of information is encoded and how? Focusing on the skip-gram with negative-sampling method, we found that the squared norm of static word embedding encodes the information gain conveyed by the word; the information gain is defined b...
http://arxiv.org/pdf/2212.09663v3
2023-11-02T16:01:32
2022-12-19T17:45:07
{ "title": [ -0.0037498474, -0.011566162, -0.022720337, 0.004257202, -0.06378174, -0.03967285, -0.020614624, -0.04071045, -0.016036987, 0.038604736, -0.04498291, 0.021881104, 0.027572632, -0.013595581, 0.011795044, 0.014457703, 0.019363403, 0.044...
http://arxiv.org/abs/2212.09662v2
MatCha: Enhancing Visual Language Pretraining with Math Reasoning and Chart Derendering
Fangyu Liu, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Yasemin Altun, Nigel Collier, Julian Martin Eisenschlos
cs.CL, cs.AI, cs.CV
Visual language data such as plots, charts, and infographics are ubiquitous in the human world. However, state-of-the-art vision-language models do not perform well on these data. We propose MatCha (Math reasoning and Chart derendering pretraining) to enhance visual language models' capabilities in jointly modeling cha...
http://arxiv.org/pdf/2212.09662v2
2023-05-23T18:21:27
2022-12-19T17:44:54
{ "title": [ -0.0053520203, -0.031829834, -0.043182373, -0.05218506, -0.019332886, -0.01828003, -0.033477783, 0.005332947, 0.017211914, 0.0063323975, -0.012145996, -0.00040102005, 0.004776001, -0.041137695, -0.00459671, -0.005771637, 0.0056037903, ...
http://arxiv.org/abs/2212.09660v2
The Decades Progress on Code-Switching Research in NLP: A Systematic Survey on Trends and Challenges
Genta Indra Winata, Alham Fikri Aji, Zheng-Xin Yong, Thamar Solorio
cs.CL
Code-Switching, a common phenomenon in written text and conversation, has been studied over decades by the natural language processing (NLP) research community. Initially, code-switching is intensively explored by leveraging linguistic theories and, currently, more machine-learning oriented approaches to develop models...
http://arxiv.org/pdf/2212.09660v2
2023-05-24T19:07:20
2022-12-19T17:42:07
{ "title": [ -0.028640747, -0.012016296, -0.021942139, -0.015304565, -0.0018892288, -0.0021076202, -0.008148193, -0.00869751, -0.008934021, 0.020248413, -0.009567261, 0.0124435425, 0.015716553, -0.027908325, -0.017028809, 0.0069122314, -0.027633667, ...
http://arxiv.org/abs/2212.09656v1
Visconde: Multi-document QA with GPT-3 and Neural Reranking
Jayr Pereira, Robson Fidalgo, Roberto Lotufo, Rodrigo Nogueira
cs.CL, cs.IR
This paper proposes a question-answering system that can answer questions whose supporting evidence is spread over multiple (potentially long) documents. The system, called Visconde, uses a three-step pipeline to perform the task: decompose, retrieve, and aggregate. The first step decomposes the question into simpler q...
http://arxiv.org/pdf/2212.09656v1
2022-12-19T17:39:07
2022-12-19T17:39:07
{ "title": [ -0.014793396, -0.0093307495, -0.021316528, -0.029296875, -0.04107666, -0.003993988, -0.035125732, -0.021102905, 0.009063721, 0.039001465, -0.048858643, 0.013290405, 0.007209778, 0.0013618469, -0.03894043, -0.012397766, -0.018814087, ...
http://arxiv.org/abs/2212.09651v4
Cross-Lingual Retrieval Augmented Prompt for Low-Resource Languages
Ercong Nie, Sheng Liang, Helmut Schmid, Hinrich Schütze
cs.CL
Multilingual Pretrained Language Models (MPLMs) have shown their strong multilinguality in recent empirical cross-lingual transfer studies. In this paper, we propose the Prompts Augmented by Retrieval Crosslingually (PARC) pipeline to improve the zero-shot performance on low-resource languages (LRLs) by augmenting the ...
http://arxiv.org/pdf/2212.09651v4
2023-07-10T22:27:15
2022-12-19T17:29:37
{ "title": [ -0.009414673, -0.015991211, -0.030075073, -0.04525757, -0.0074043274, -0.007129669, -0.041656494, -0.009376526, 0.018753052, 0.055908203, -0.01272583, -0.0041389465, 0.03753662, -0.014007568, -0.044189453, 0.0178833, -0.0028305054, 0...
http://arxiv.org/abs/2212.09648v4
NusaCrowd: Open Source Initiative for Indonesian NLP Resources
Samuel Cahyawijaya, Holy Lovenia, Alham Fikri Aji, Genta Indra Winata, Bryan Wilie, Rahmad Mahendra, Christian Wibisono, Ade Romadhony, Karissa Vincentio, Fajri Koto, Jennifer Santoso, David Moeljadi, Cahya Wirawan, Frederikus Hudi, Ivan Halim Parmonangan, Ika Alfina, Muhammad Satrio Wicaksono, Ilham Firdausi Putra, Sa...
cs.CL, cs.AI
We present NusaCrowd, a collaborative initiative to collect and unify existing resources for Indonesian languages, including opening access to previously non-public resources. Through this initiative, we have brought together 137 datasets and 118 standardized data loaders. The quality of the datasets has been assessed ...
http://arxiv.org/pdf/2212.09648v4
2023-07-21T14:44:45
2022-12-19T17:28:22
{ "title": [ 0.0072746277, -0.027893066, -0.04788208, -0.041381836, -0.029403687, -0.018600464, -0.03164673, 0.0017204285, -0.0146102905, 0.023010254, -0.030700684, 0.007205963, -0.00038814545, -0.028182983, -0.02519226, -0.021331787, -0.015602112, ...
http://arxiv.org/abs/2212.09631v2
Optimal Transport for Unsupervised Hallucination Detection in Neural Machine Translation
Nuno M. Guerreiro, Pierre Colombo, Pablo Piantanida, André F. T. Martins
cs.CL, cs.LG
Neural machine translation (NMT) has become the de-facto standard in real-world machine translation applications. However, NMT models can unpredictably produce severely pathological translations, known as hallucinations, that seriously undermine user trust. It becomes thus crucial to implement effective preventive stra...
http://arxiv.org/pdf/2212.09631v2
2023-05-19T16:36:12
2022-12-19T17:06:58
{ "title": [ 0.0009036064, 0.011703491, 0.020812988, -0.024993896, -0.009841919, 0.0016021729, -0.005012512, -0.037384033, 0.0058403015, 0.050567627, -0.0058250427, 0.0052833557, 0.024230957, -0.021209717, -0.03845215, 0.025024414, -0.007904053, ...
http://arxiv.org/abs/2212.09621v1
Wukong-Reader: Multi-modal Pre-training for Fine-grained Visual Document Understanding
Haoli Bai, Zhiguang Liu, Xiaojun Meng, Wentao Li, Shuang Liu, Nian Xie, Rongfu Zheng, Liangwei Wang, Lu Hou, Jiansheng Wei, Xin Jiang, Qun Liu
cs.CL, cs.CV
Unsupervised pre-training on millions of digital-born or scanned documents has shown promising advances in visual document understanding~(VDU). While various vision-language pre-training objectives are studied in existing solutions, the document textline, as an intrinsic granularity in VDU, has seldom been explored so ...
http://arxiv.org/pdf/2212.09621v1
2022-12-19T17:00:54
2022-12-19T17:00:54
{ "title": [ -0.01675415, -0.0014686584, -0.033447266, -0.05206299, -0.012969971, -0.0028896332, -0.057281494, -0.021362305, 0.028396606, 0.02810669, -0.035614014, -0.0053138733, 0.006324768, -0.0031147003, -0.058441162, -0.018188477, 0.013702393, ...
http://arxiv.org/abs/2212.09611v1
Optimizing Prompts for Text-to-Image Generation
Yaru Hao, Zewen Chi, Li Dong, Furu Wei
cs.CL, cs.CV
Well-designed prompts can guide text-to-image models to generate amazing images. However, the performant prompts are often model-specific and misaligned with user input. Instead of laborious human engineering, we propose prompt adaptation, a general framework that automatically adapts original user input to model-prefe...
http://arxiv.org/pdf/2212.09611v1
2022-12-19T16:50:41
2022-12-19T16:50:41
{ "title": [ -0.0074424744, 0.002614975, -0.016967773, -0.03137207, -0.00409317, -0.0009813309, -0.02557373, -0.03363037, 0.022766113, 0.04473877, -0.03842163, 0.024047852, 0.0015640259, -0.0049438477, -0.040374756, 0.0082473755, 0.007949829, 0.0...
http://arxiv.org/abs/2212.09603v2
Explanation Regeneration via Information Bottleneck
Qintong Li, Zhiyong Wu, Lingpeng Kong, Wei Bi
cs.CL
Explaining the black-box predictions of NLP models naturally and accurately is an important open problem in natural language generation. These free-text explanations are expected to contain sufficient and carefully-selected evidence to form supportive arguments for predictions. Due to the superior generative capacity o...
http://arxiv.org/pdf/2212.09603v2
2023-07-11T05:17:19
2022-12-19T16:41:19
{ "title": [ -0.032043457, -0.012321472, -0.035003662, -0.05895996, -0.025894165, 0.0013446808, -0.043884277, 0.004211426, -0.002872467, 0.018539429, -0.019424438, -0.0043678284, -0.008987427, -0.012672424, -0.02859497, 0.0066490173, 0.005001068, ...
http://arxiv.org/abs/2212.09597v8
Reasoning with Language Model Prompting: A Survey
Shuofei Qiao, Yixin Ou, Ningyu Zhang, Xiang Chen, Yunzhi Yao, Shumin Deng, Chuanqi Tan, Fei Huang, Huajun Chen
cs.CL, cs.AI, cs.CV, cs.IR, cs.LG
Reasoning, as an essential ability for complex problem-solving, can provide back-end support for various real-world applications, such as medical diagnosis, negotiation, etc. This paper provides a comprehensive survey of cutting-edge research on reasoning with language model prompting. We introduce research works with ...
http://arxiv.org/pdf/2212.09597v8
2023-09-18T10:47:13
2022-12-19T16:32:42
{ "title": [ -0.03744507, -0.03302002, -0.020645142, -0.08331299, -0.018569946, 0.028823853, -0.094055176, 0.009552002, 0.0032024384, 0.0014925003, -0.021194458, -0.006122589, -0.0143966675, -0.029205322, -0.019989014, 0.0003066063, -0.0010948181, ...
http://arxiv.org/abs/2212.09593v3
Unsupervised Summarization Re-ranking
Mathieu Ravaut, Shafiq Joty, Nancy Chen
cs.CL
With the rise of task-specific pre-training objectives, abstractive summarization models like PEGASUS offer appealing zero-shot performance on downstream summarization tasks. However, the performance of such unsupervised models still lags significantly behind their supervised counterparts. Similarly to the supervised s...
http://arxiv.org/pdf/2212.09593v3
2023-05-26T05:26:23
2022-12-19T16:29:26
{ "title": [ 0.01612854, 0.024642944, -0.023590088, -0.027786255, -0.017074585, -0.024124146, -0.050964355, -0.04348755, 0.02658081, 0.06896973, -0.048309326, 0.022598267, 0.03564453, 0.00667572, -0.026641846, 0.026245117, -0.0004720688, 0.013374...
http://arxiv.org/abs/2212.09588v2
Query Enhanced Knowledge-Intensive Conversation via Unsupervised Joint Modeling
Mingzhu Cai, Siqi Bao, Xin Tian, Huang He, Fan Wang, Hua Wu
cs.CL
In this paper, we propose an unsupervised query enhanced approach for knowledge-intensive conversations, namely QKConv. There are three modules in QKConv: a query generator, an off-the-shelf knowledge selector, and a response generator. QKConv is optimized through joint training, which produces the response by explorin...
http://arxiv.org/pdf/2212.09588v2
2023-05-26T11:02:13
2022-12-19T16:21:05
{ "title": [ -0.0050811768, 0.015960693, -0.057556152, -0.07531738, 0.01965332, -0.0047569275, -0.054016113, 0.007980347, 0.0011329651, 0.033966064, -0.007194519, 0.002023697, 0.0076141357, -0.0178833, -0.044952393, 0.0019779205, -0.0073013306, 0...
http://arxiv.org/abs/2212.09580v1
Independent Components of Word Embeddings Represent Semantic Features
Tomáš Musil, David Mareček
cs.CL
Independent Component Analysis (ICA) is an algorithm originally developed for finding separate sources in a mixed signal, such as a recording of multiple people in the same room speaking at the same time. It has also been used to find linguistic features in distributional representations. In this paper, we used ICA to ...
http://arxiv.org/pdf/2212.09580v1
2022-12-19T16:13:52
2022-12-19T16:13:52
{ "title": [ -0.0007286072, -0.0036334991, -0.04321289, -0.040283203, -0.013702393, -0.034240723, -0.032196045, -0.005378723, 0.0030078888, 0.024429321, -0.018676758, -0.014595032, 0.0019569397, -0.034057617, -0.019683838, 0.013587952, 0.007843018, ...
http://arxiv.org/abs/2212.09577v3
CiteBench: A benchmark for Scientific Citation Text Generation
Martin Funkquist, Ilia Kuznetsov, Yufang Hou, Iryna Gurevych
cs.CL
Science progresses by building upon the prior body of knowledge documented in scientific publications. The acceleration of research makes it hard to stay up-to-date with the recent developments and to summarize the ever-growing body of prior work. To address this, the task of citation text generation aims to produce ac...
http://arxiv.org/pdf/2212.09577v3
2023-11-03T19:55:56
2022-12-19T16:10:56
{ "title": [ -0.025878906, -0.0073890686, -0.038024902, -0.051849365, 0.010726929, 0.0048561096, -0.04071045, -0.022918701, 0.004886627, 0.044281006, 0.009361267, 0.019744873, 0.005558014, -0.012496948, -0.05911255, 0.017227173, -0.018661499, 0.0...
http://arxiv.org/abs/2212.09573v1
Privacy Adhering Machine Un-learning in NLP
Vinayshekhar Bannihatti Kumar, Rashmi Gangadharaiah, Dan Roth
cs.CL
Regulations introduced by General Data Protection Regulation (GDPR) in the EU or California Consumer Privacy Act (CCPA) in the US have included provisions on the \textit{right to be forgotten} that mandates industry applications to remove data related to an individual from their systems. In several real world industry ...
http://arxiv.org/pdf/2212.09573v1
2022-12-19T16:06:45
2022-12-19T16:06:45
{ "title": [ -0.00048828125, -0.002008438, -0.03668213, -0.018997192, -0.02217102, -0.027023315, -0.01008606, -0.01108551, -0.014457703, 0.03652954, -0.028503418, 0.006755829, 0.021514893, -0.0309906, 0.0009937286, 0.004447937, -0.016403198, 0.06...
http://arxiv.org/abs/2212.09563v1
Source-Free Domain Adaptation for Question Answering with Masked Self-training
M. Yin, B. Wang, Y. Dong, C. Ling
cs.CL
Most previous unsupervised domain adaptation (UDA) methods for question answering(QA) require access to source domain data while fine-tuning the model for the target domain. Source domain data may, however, contain sensitive information and may be restricted. In this study, we investigate a more challenging setting, so...
http://arxiv.org/pdf/2212.09563v1
2022-12-19T15:53:12
2022-12-19T15:53:12
{ "title": [ -0.008270264, 0.005554199, -0.039215088, -0.048461914, 0.0044288635, -0.010215759, -0.007698059, 0.024810791, -0.003019333, 0.032196045, 0.026535034, 0.0014133453, -0.011192322, -0.02848816, -0.04498291, 0.015792847, 0.0067710876, 0....
http://arxiv.org/abs/2212.09561v5
Large Language Models are Better Reasoners with Self-Verification
Yixuan Weng, Minjun Zhu, Fei Xia, Bin Li, Shizhu He, Shengping Liu, Bin Sun, Kang Liu, Jun Zhao
cs.AI, cs.CL
Recently, with the chain of thought (CoT) prompting, large language models (LLMs), e.g., GPT-3, have shown strong reasoning ability in several natural language processing tasks such as arithmetic, commonsense, and logical reasoning. However, LLMs with CoT require multi-step prompting and multi-token prediction, which i...
http://arxiv.org/pdf/2212.09561v5
2023-10-19T12:24:42
2022-12-19T15:51:52
{ "title": [ 0.007041931, -0.008636475, -0.040130615, -0.1105957, -0.001824379, -0.002828598, -0.073791504, 0.00831604, -0.030334473, 0.021728516, -0.009048462, -0.013771057, -0.003917694, -0.039978027, -0.0053138733, -0.0028629303, -0.002450943, ...
http://arxiv.org/abs/2212.09553v2
Mu$^{2}$SLAM: Multitask, Multilingual Speech and Language Models
Yong Cheng, Yu Zhang, Melvin Johnson, Wolfgang Macherey, Ankur Bapna
cs.CL, cs.SD, eess.AS
We present Mu$^{2}$SLAM, a multilingual sequence-to-sequence model pre-trained jointly on unlabeled speech, unlabeled text and supervised data spanning Automatic Speech Recognition (ASR), Automatic Speech Translation (AST) and Machine Translation (MT), in over 100 languages. By leveraging a quantized representation of ...
http://arxiv.org/pdf/2212.09553v2
2023-06-27T01:18:45
2022-12-19T15:45:36
{ "title": [ -0.03729248, -0.026138306, -0.034729004, -0.027862549, -0.01802063, -0.005001068, -0.05819702, 0.0037651062, -0.019302368, 0.054718018, 0.0042419434, -0.007972717, 0.032592773, -0.03933716, -0.025604248, 0.012863159, -0.0079422, 0.02...
http://arxiv.org/abs/2212.09535v3
BLOOM+1: Adding Language Support to BLOOM for Zero-Shot Prompting
Zheng-Xin Yong, Hailey Schoelkopf, Niklas Muennighoff, Alham Fikri Aji, David Ifeoluwa Adelani, Khalid Almubarak, M Saiful Bari, Lintang Sutawika, Jungo Kasai, Ahmed Baruwa, Genta Indra Winata, Stella Biderman, Edward Raff, Dragomir Radev, Vassilina Nikoulina
cs.CL, cs.AI, cs.LG
The BLOOM model is a large publicly available multilingual language model, but its pretraining was limited to 46 languages. To extend the benefits of BLOOM to other languages without incurring prohibitively large costs, it is desirable to adapt BLOOM to new languages not seen during pretraining. In this work, we apply ...
http://arxiv.org/pdf/2212.09535v3
2023-05-27T05:48:38
2022-12-19T15:24:45
{ "title": [ 0.010498047, -0.029037476, -0.037017822, -0.06518555, -0.02204895, -0.011489868, -0.052978516, -0.0043754578, 0.018295288, 0.020828247, -0.015617371, -0.007774353, 0.027694702, -0.045928955, -0.00868988, 0.00032639503, 0.024108887, -...
http://arxiv.org/abs/2212.09512v1
Rethinking Label Smoothing on Multi-hop Question Answering
Zhangyue Yin, Yuxin Wang, Yiguang Wu, Hang Yan, Xiannian Hu, Xinyu Zhang, Zhao Cao, Xuanjing Huang, Xipeng Qiu
cs.CL
Label smoothing is a regularization technique widely used in supervised learning to improve the generalization of models on various tasks, such as image classification and machine translation. However, the effectiveness of label smoothing in multi-hop question answering (MHQA) has yet to be well studied. In this paper,...
http://arxiv.org/pdf/2212.09512v1
2022-12-19T14:48:08
2022-12-19T14:48:08
{ "title": [ -0.02192688, -0.013084412, -0.048309326, -0.02670288, 0.000045001507, -0.018112183, -0.0032444, 0.02166748, 0.007118225, 0.023498535, -0.026443481, 0.0068206787, 0.0043754578, -0.010116577, -0.032562256, -0.0020217896, -0.0060539246, ...
http://arxiv.org/abs/2212.09465v1
Improving the Generalizability of Text-Based Emotion Detection by Leveraging Transformers with Psycholinguistic Features
Sourabh Zanwar, Daniel Wiechmann, Yu Qiao, Elma Kerz
cs.CL
In recent years, there has been increased interest in building predictive models that harness natural language processing and machine learning techniques to detect emotions from various text sources, including social media posts, micro-blogs or news articles. Yet, deployment of such models in real-world sentiment and e...
http://arxiv.org/pdf/2212.09465v1
2022-12-19T13:58:48
2022-12-19T13:58:48
{ "title": [ -0.0023555756, 0.005844116, -0.012245178, -0.03378296, 0.0071525574, 0.0014476776, -0.017608643, -0.027496338, 0.025741577, 0.03543091, -0.042541504, 0.010803223, 0.014404297, -0.0034618378, -0.05645752, -0.00066900253, -0.012069702, ...
http://arxiv.org/abs/2212.09462v2
Latent Diffusion for Language Generation
Justin Lovelace, Varsha Kishore, Chao Wan, Eliot Shekhtman, Kilian Q. Weinberger
cs.CL, cs.LG
Diffusion models have achieved great success in modeling continuous data modalities such as images, audio, and video, but have seen limited use in discrete domains such as language. Recent attempts to adapt diffusion to language have presented diffusion as an alternative to existing pretrained language models. We view ...
http://arxiv.org/pdf/2212.09462v2
2023-11-07T15:35:45
2022-12-19T13:57:06
{ "title": [ -0.03074646, -0.020065308, -0.044799805, -0.047698975, 0.009719849, 0.0027427673, -0.020645142, 0.0074043274, 0.0013198853, 0.023376465, -0.0032405853, 0.020477295, -0.008506775, -0.026916504, -0.025039673, 0.02331543, -0.016601562, ...
http://arxiv.org/abs/2212.09422v1
Human in the loop: How to effectively create coherent topics by manually labeling only a few documents per class
Anton Thielmann, Christoph Weisser, Benjamin Säfken
cs.CL
Few-shot methods for accurate modeling under sparse label-settings have improved significantly. However, the applications of few-shot modeling in natural language processing remain solely in the field of document classification. With recent performance improvements, supervised few-shot methods, combined with a simple t...
http://arxiv.org/pdf/2212.09422v1
2022-12-19T12:57:11
2022-12-19T12:57:11
{ "title": [ -0.0032310486, 0.0052223206, -0.045776367, 0.00094509125, -0.022232056, -0.02468872, -0.03378296, -0.012229919, -0.037597656, 0.027893066, -0.04550171, 0.018600464, 0.004058838, -0.014221191, 0.017562866, -0.047424316, -0.007663727, ...
http://arxiv.org/abs/2212.09420v2
Large Language Models Meet NL2Code: A Survey
Daoguang Zan, Bei Chen, Fengji Zhang, Dianjie Lu, Bingchao Wu, Bei Guan, Yongji Wang, Jian-Guang Lou
cs.SE, cs.AI, cs.CL, cs.PL
The task of generating code from a natural language description, or NL2Code, is considered a pressing and significant challenge in code intelligence. Thanks to the rapid development of pre-training techniques, surging large language models are being proposed for code, sparking the advances in NL2Code. To facilitate fur...
http://arxiv.org/pdf/2212.09420v2
2023-05-08T10:25:41
2022-12-19T12:55:32
{ "title": [ -0.049102783, -0.024368286, -0.026550293, -0.076416016, -0.014549255, 0.02897644, -0.07702637, 0.027618408, -0.00881958, 0.047607422, -0.0026168823, -0.0065193176, 0.017501831, -0.011749268, -0.051940918, 0.017715454, -0.025238037, 0...
http://arxiv.org/abs/2212.09412v2
Difformer: Empowering Diffusion Models on the Embedding Space for Text Generation
Zhujin Gao, Junliang Guo, Xu Tan, Yongxin Zhu, Fang Zhang, Jiang Bian, Linli Xu
cs.CL, cs.AI, cs.LG
Diffusion models have achieved state-of-the-art synthesis quality on both visual and audio tasks, and recent works further adapt them to textual data by diffusing on the embedding space. In this paper, we conduct systematic studies and analyze the challenges between the continuous data space and the embedding space whi...
http://arxiv.org/pdf/2212.09412v2
2023-01-31T09:21:38
2022-12-19T12:44:25
{ "title": [ -0.019973755, -0.007167816, -0.040924072, -0.039886475, -0.0006814003, 0.006526947, -0.03463745, -0.017059326, 0.03463745, 0.018188477, -0.023590088, 0.01864624, 0.0020065308, -0.01600647, -0.03060913, 0.009475708, -0.0259552, 0.0490...
http://arxiv.org/abs/2212.09409v2
Multi-View Knowledge Distillation from Crowd Annotations for Out-of-Domain Generalization
Dustin Wright, Isabelle Augenstein
cs.CL
Selecting an effective training signal for tasks in natural language processing is difficult: expert annotations are expensive, and crowd-sourced annotations may not be reliable. At the same time, recent work in NLP has demonstrated that learning from a distribution over labels acquired from crowd annotations can be ef...
http://arxiv.org/pdf/2212.09409v2
2023-05-23T14:44:24
2022-12-19T12:40:18
{ "title": [ 0.02305603, 0.015319824, -0.03527832, -0.07946777, 0.012481689, -0.009422302, -0.027557373, 0.011131287, 0.02947998, 0.029785156, -0.051361084, 0.017593384, -0.02268982, 0.0064811707, -0.043548584, -0.020523071, 0.009269714, -0.00072...
http://arxiv.org/abs/2212.09410v1
Less is More: Parameter-Free Text Classification with Gzip
Zhiying Jiang, Matthew Y. R. Yang, Mikhail Tsirlin, Raphael Tang, Jimmy Lin
cs.CL
Deep neural networks (DNNs) are often used for text classification tasks as they usually achieve high levels of accuracy. However, DNNs can be computationally intensive with billions of parameters and large amounts of labeled data, which can make them expensive to use, to optimize and to transfer to out-of-distribution...
http://arxiv.org/pdf/2212.09410v1
2022-12-19T12:40:18
2022-12-19T12:40:18
{ "title": [ -0.009346008, 0.009811401, -0.043182373, -0.06640625, 0.025634766, -0.0048065186, -0.01777649, -0.001947403, -0.008056641, 0.022583008, -0.011009216, 0.018203735, -0.013175964, -0.018844604, -0.037750244, -0.009780884, -0.0018692017, ...
http://arxiv.org/abs/2212.09400v2
An Efficient Drug-Drug Interactions Prediction Technology for Molecularly Intelligent Manufacturing
Peng Gao, Feng Gao, Jian-Cheng Ni
cs.CL
Drug-Drug Interactions (DDIs) prediction is an essential issue in the molecular field. Traditional methods of observing DDIs in medical experiments require plenty of resources and labor. In this paper, we present a computational model dubbed MedKGQA based on Graph Neural Networks to automatically predict the DDIs after...
http://arxiv.org/pdf/2212.09400v2
2022-12-22T01:57:51
2022-12-19T12:24:32
{ "title": [ -0.025619507, 0.0031929016, -0.061401367, -0.08013916, 0.022750854, -0.010467529, -0.04940796, 0.0287323, -0.007156372, -0.00037550926, -0.01600647, 0.018096924, -0.048034668, -0.021606445, -0.033172607, -0.019622803, -0.009613037, 0...
http://arxiv.org/abs/2212.09387v2
An Extensible Plug-and-Play Method for Multi-Aspect Controllable Text Generation
Xuancheng Huang, Zijun Liu, Peng Li, Tao Li, Maosong Sun, Yang Liu
cs.CL, cs.AI
Recently, multi-aspect controllable text generation that controls the generated text in multiple aspects (e.g., sentiment, topic, and keywords) has attracted increasing attention. Although methods based on parameter efficient tuning like prefix-tuning could achieve multi-aspect controlling in a plug-and-play way, the m...
http://arxiv.org/pdf/2212.09387v2
2023-05-28T14:12:48
2022-12-19T11:53:59
{ "title": [ -0.03164673, -0.0053977966, -0.0345459, -0.06488037, -0.0026950836, 0.025512695, -0.07080078, -0.00056028366, -0.00016069412, 0.050048828, 0.007865906, 0.027755737, 0.005996704, -0.010856628, -0.049804688, 0.017456055, -0.024108887, ...
http://arxiv.org/abs/2212.09376v1
Enriching Relation Extraction with OpenIE
Alessandro Temperoni, Maria Biryukov, Martin Theobald
cs.CL, cs.IR, cs.LG
Relation extraction (RE) is a sub-discipline of information extraction (IE) which focuses on the prediction of a relational predicate from a natural-language input unit (such as a sentence, a clause, or even a short paragraph consisting of multiple sentences and/or clauses). Together with named-entity recognition (NER)...
http://arxiv.org/pdf/2212.09376v1
2022-12-19T11:26:23
2022-12-19T11:26:23
{ "title": [ -0.018875122, -0.006816864, -0.048431396, -0.080566406, 0.022033691, 0.021209717, -0.056121826, -0.005973816, 0.009513855, 0.027160645, -0.010658264, 0.027282715, -0.0038795471, -0.010505676, -0.031234741, 0.011566162, -0.013839722, ...
http://arxiv.org/abs/2212.09359v3
WACO: Word-Aligned Contrastive Learning for Speech Translation
Siqi Ouyang, Rong Ye, Lei Li
cs.CL, cs.SD, eess.AS
End-to-end Speech Translation (E2E ST) aims to directly translate source speech into target text. Existing ST methods perform poorly when only extremely small speech-text data are available for training. We observe that an ST model's performance closely correlates with its embedding similarity between speech and source...
http://arxiv.org/pdf/2212.09359v3
2023-07-07T04:56:14
2022-12-19T10:49:35
{ "title": [ -0.022857666, -0.00687027, -0.01739502, -0.037109375, -0.014915466, 0.0007724762, -0.03753662, -0.011398315, 0.010871887, 0.04675293, -0.028579712, 0.014427185, 0.02279663, -0.020889282, -0.04525757, 0.0006508827, -0.008369446, 0.043...
http://arxiv.org/abs/2212.09353v1
Bridging The Gap: Entailment Fused-T5 for Open-retrieval Conversational Machine Reading Comprehension
Xiao Zhang, Heyan Huang, Zewen Chi, Xian-Ling Mao
cs.CL
Open-retrieval conversational machine reading comprehension (OCMRC) simulates real-life conversational interaction scenes. Machines are required to make a decision of "Yes/No/Inquire" or generate a follow-up question when the decision is "Inquire" based on retrieved rule texts, user scenario, user question, and dialogu...
http://arxiv.org/pdf/2212.09353v1
2022-12-19T10:38:30
2022-12-19T10:38:30
{ "title": [ -0.020889282, 0.0023078918, -0.0049438477, -0.068847656, -0.034301758, 0.0309906, -0.08251953, -0.021606445, 0.01676941, 0.040527344, -0.03164673, 0.011062622, 0.04486084, -0.008979797, -0.028900146, 0.014724731, -0.017059326, 0.0014...
http://arxiv.org/abs/2212.09306v1
E-NER -- An Annotated Named Entity Recognition Corpus of Legal Text
Ting Wai Terence Au, Ingemar J. Cox, Vasileios Lampos
cs.CL
Identifying named entities such as a person, location or organization, in documents can highlight key information to readers. Training Named Entity Recognition (NER) models requires an annotated data set, which can be a time-consuming labour-intensive task. Nevertheless, there are publicly available NER data sets for g...
http://arxiv.org/pdf/2212.09306v1
2022-12-19T09:03:32
2022-12-19T09:03:32
{ "title": [ -0.025558472, -0.0068473816, -0.062316895, -0.01828003, 0.0011997223, -0.012825012, -0.023284912, -0.0004210472, 0.01158905, 0.024093628, -0.06768799, 0.026123047, 0.05142212, -0.01789856, 0.014038086, -0.001244545, -0.017562866, 0.0...
http://arxiv.org/abs/2212.09305v2
SESCORE2: Learning Text Generation Evaluation via Synthesizing Realistic Mistakes
Wenda Xu, Xian Qian, Mingxuan Wang, Lei Li, William Yang Wang
cs.CL
Is it possible to train a general metric for evaluating text generation quality without human annotated ratings? Existing learned metrics either perform unsatisfactorily across text generation tasks or require human ratings for training on specific tasks. In this paper, we propose SESCORE2, a self-supervised approach f...
http://arxiv.org/pdf/2212.09305v2
2023-07-07T17:49:31
2022-12-19T09:02:16
{ "title": [ -0.034240723, -0.014350891, -0.03111267, -0.024810791, -0.016555786, 0.0064582825, -0.0395813, -0.021835327, -0.0058288574, 0.020111084, -0.033813477, 0.02859497, -0.025146484, 0.0027770996, -0.04626465, -0.019882202, -0.037750244, 0...
http://arxiv.org/abs/2212.09292v1
ChatGPT: The End of Online Exam Integrity?
Teo Susnjak
cs.AI, cs.CL
This study evaluated the ability of ChatGPT, a recently developed artificial intelligence (AI) agent, to perform high-level cognitive tasks and produce text that is indistinguishable from human-generated text. This capacity raises concerns about the potential use of ChatGPT as a tool for academic misconduct in online e...
http://arxiv.org/pdf/2212.09292v1
2022-12-19T08:15:16
2022-12-19T08:15:16
{ "title": [ 0.016723633, -0.009864807, -0.008392334, -0.013916016, -0.016174316, 0.002046585, -0.0026550293, -0.021697998, -0.02494812, 0.035003662, 0.005683899, 0.005241394, 0.011054993, -0.04434204, -0.004840851, 0.0036849976, -0.020401001, 0....
http://arxiv.org/abs/2212.09284v1
An Investigation of Indian Native Language Phonemic Influences on L2 English Pronunciations
Shelly Jain, Priyanshi Pal, Anil Vuppala, Prasanta Ghosh, Chiranjeevi Yarra
cs.CL
Speech systems are sensitive to accent variations. This is especially challenging in the Indian context, with an abundance of languages but a dearth of linguistic studies characterising pronunciation variations. The growing number of L2 English speakers in India reinforces the need to study accents and L1-L2 interactio...
http://arxiv.org/pdf/2212.09284v1
2022-12-19T07:41:39
2022-12-19T07:41:39
{ "title": [ -0.028762817, -0.0013303757, -0.043029785, -0.03555298, -0.030426025, -0.008018494, -0.038909912, 0.018417358, -0.046447754, 0.035888672, -0.0005636215, 0.0026721954, -0.0038146973, -0.040100098, -0.011764526, -0.01676941, -0.023742676, ...
http://arxiv.org/abs/2212.09282v2
APOLLO: A Simple Approach for Adaptive Pretraining of Language Models for Logical Reasoning
Soumya Sanyal, Yichong Xu, Shuohang Wang, Ziyi Yang, Reid Pryzant, Wenhao Yu, Chenguang Zhu, Xiang Ren
cs.CL, cs.AI, cs.LG
Logical reasoning of text is an important ability that requires understanding the information present in the text, their interconnections, and then reasoning through them to infer new conclusions. Prior works on improving the logical reasoning ability of language models require complex processing of training data (e.g....
http://arxiv.org/pdf/2212.09282v2
2023-06-05T00:56:11
2022-12-19T07:40:02
{ "title": [ -0.03225708, -0.016708374, -0.050994873, -0.07165527, -0.0022888184, 0.0012922287, -0.076049805, 0.011756897, -0.0076293945, 0.020523071, -0.011299133, 0.009216309, 0.000647068, -0.024093628, -0.02796936, 0.007144928, -0.010803223, 0...
http://arxiv.org/abs/2212.09278v1
MIGA: A Unified Multi-task Generation Framework for Conversational Text-to-SQL
Yingwen Fu, Wenjie Ou, Zhou Yu, Yue Lin
cs.CL, cs.AI
Conversational text-to-SQL is designed to translate multi-turn natural language questions into their corresponding SQL queries. Most state-of-the-art conversational text- to-SQL methods are incompatible with generative pre-trained language models (PLMs), such as T5. In this paper, we present a two-stage unified MultI-t...
http://arxiv.org/pdf/2212.09278v1
2022-12-19T07:14:32
2022-12-19T07:14:32
{ "title": [ -0.00844574, -0.007820129, -0.05783081, -0.07946777, 0.014465332, 0.00223732, -0.068481445, 0.024383545, 0.026275635, 0.06414795, -0.01486969, 0.013198853, 0.022613525, -0.00299263, -0.08190918, 0.005207062, -0.011077881, 0.023895264...
http://arxiv.org/abs/2212.09272v1
Statistical Dataset Evaluation: Reliability, Difficulty, and Validity
Chengwen Wang, Qingxiu Dong, Xiaochen Wang, Haitao Wang, Zhifang Sui
cs.CL
Datasets serve as crucial training resources and model performance trackers. However, existing datasets have exposed a plethora of problems, inducing biased models and unreliable evaluation results. In this paper, we propose a model-agnostic dataset evaluation framework for automatic dataset quality evaluation. We seek...
http://arxiv.org/pdf/2212.09272v1
2022-12-19T06:55:42
2022-12-19T06:55:42
{ "title": [ -0.02420044, 0.007347107, -0.062438965, -0.04559326, 0.035858154, -0.013710022, -0.018463135, -0.023147583, -0.0007247925, 0.012626648, -0.03137207, 0.0051727295, -0.053985596, -0.027145386, -0.06286621, -0.008377075, -0.0071754456, ...
http://arxiv.org/abs/2212.09257v2
PromptBoosting: Black-Box Text Classification with Ten Forward Passes
Bairu Hou, Joe O'Connor, Jacob Andreas, Shiyu Chang, Yang Zhang
cs.CL
We describe PromptBoosting, a query-efficient procedure for building a text classifier from a neural language model (LM) without access to the LM's parameters, gradients, or hidden representations. This form of "black-box" classifier training has become increasingly important as the cost of training and inference in la...
http://arxiv.org/pdf/2212.09257v2
2023-07-03T02:28:27
2022-12-19T06:04:54
{ "title": [ 0.030578613, 0.0131073, -0.023284912, -0.061828613, 0.014778137, -0.025024414, -0.020553589, -0.033477783, 0.008865356, 0.028045654, -0.042144775, 0.04135132, -0.037902832, 0.0061454773, -0.05307007, -0.011253357, -0.016860962, 0.061...
http://arxiv.org/abs/2212.09255v1
Multi hash embeddings in spaCy
Lester James Miranda, Ákos Kádár, Adriane Boyd, Sofie Van Landeghem, Anders Søgaard, Matthew Honnibal
cs.CL, I.2.7
The distributed representation of symbols is one of the key technologies in machine learning systems today, playing a pivotal role in modern natural language processing. Traditional word embeddings associate a separate vector with each word. While this approach is simple and leads to good performance, it requires a lot...
http://arxiv.org/pdf/2212.09255v1
2022-12-19T06:03:04
2022-12-19T06:03:04
{ "title": [ 0.0012636185, -0.0039138794, -0.038330078, -0.080566406, -0.0057868958, 0.010368347, -0.045898438, 0.021377563, -0.019378662, 0.06463623, -0.011947632, -0.014419556, 0.021194458, -0.02835083, -0.015792847, -0.0065193176, 0.006839752, ...
http://arxiv.org/abs/2212.09254v1
TextGrad: Advancing Robustness Evaluation in NLP by Gradient-Driven Optimization
Bairu Hou, Jinghan Jia, Yihua Zhang, Guanhua Zhang, Yang Zhang, Sijia Liu, Shiyu Chang
cs.CL
Robustness evaluation against adversarial examples has become increasingly important to unveil the trustworthiness of the prevailing deep models in natural language processing (NLP). However, in contrast to the computer vision domain where the first-order projected gradient descent (PGD) is used as the benchmark approa...
http://arxiv.org/pdf/2212.09254v1
2022-12-19T05:55:58
2022-12-19T05:55:58
{ "title": [ -0.008232117, 0.011558533, -0.027618408, -0.064331055, 0.02116394, 0.012832642, -0.054901123, -0.029632568, 0.00041246414, 0.016082764, -0.024536133, 0.02633667, 0.00066661835, -0.00047016144, -0.03933716, -0.0074806213, -0.027252197, ...
http://arxiv.org/abs/2212.09252v2
Mind the Knowledge Gap: A Survey of Knowledge-enhanced Dialogue Systems
Sagi Shaier, Lawrence Hunter, Katharina Kann
cs.CL, cs.LG
Many dialogue systems (DSs) lack characteristics humans have, such as emotion perception, factuality, and informativeness. Enhancing DSs with knowledge alleviates this problem, but, as many ways of doing so exist, keeping track of all proposed methods is difficult. Here, we present the first survey of knowledge-enhance...
http://arxiv.org/pdf/2212.09252v2
2022-12-20T17:52:24
2022-12-19T05:17:33
{ "title": [ -0.025924683, -0.011627197, -0.051452637, -0.06738281, 0.022476196, -0.010543823, -0.025314331, 0.015457153, 0.009765625, 0.024597168, -0.019714355, 0.005493164, -0.019989014, -0.0132751465, -0.06286621, -0.0052337646, -0.0047683716, ...
http://arxiv.org/abs/2212.09251v1
Discovering Language Model Behaviors with Model-Written Evaluations
Ethan Perez, Sam Ringer, Kamilė Lukošiūtė, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Ben Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, Dario Amodei, Dawn Drain, Dustin Li, Eli Tran-Joh...
cs.CL, cs.AI, cs.LG
As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations...
http://arxiv.org/pdf/2212.09251v1
2022-12-19T05:13:52
2022-12-19T05:13:52
{ "title": [ -0.02772522, -0.00095939636, -0.028213501, -0.04727173, 0.0005955696, 0.006931305, -0.057495117, -0.0035209656, -0.008102417, 0.014511108, -0.030563354, 0.012420654, -0.02128601, -0.011077881, -0.047607422, -0.010314941, -0.029907227, ...
http://arxiv.org/abs/2212.09248v1
Natural Language to Code Generation in Interactive Data Science Notebooks
Pengcheng Yin, Wen-Ding Li, Kefan Xiao, Abhishek Rao, Yeming Wen, Kensen Shi, Joshua Howland, Paige Bailey, Michele Catasta, Henryk Michalewski, Alex Polozov, Charles Sutton
cs.CL, cs.SE
Computational notebooks, such as Jupyter notebooks, are interactive computing environments that are ubiquitous among data scientists to perform data wrangling and analytic tasks. To measure the performance of AI pair programmers that automatically synthesize programs for those tasks given natural language (NL) intents ...
http://arxiv.org/pdf/2212.09248v1
2022-12-19T05:06:00
2022-12-19T05:06:00
{ "title": [ -0.04788208, -0.038848877, -0.03427124, -0.022109985, -0.04660034, 0.006790161, -0.041046143, 0.006702423, 0.012046814, 0.019058228, -0.042297363, -0.00068855286, -0.027832031, -0.0076522827, -0.03781128, -0.029693604, 0.00046944618, ...
http://arxiv.org/abs/2212.09246v3
I2D2: Inductive Knowledge Distillation with NeuroLogic and Self-Imitation
Chandra Bhagavatula, Jena D. Hwang, Doug Downey, Ronan Le Bras, Ximing Lu, Lianhui Qin, Keisuke Sakaguchi, Swabha Swayamdipta, Peter West, Yejin Choi
cs.CL
Commonsense capabilities of pre-trained language models dramatically improve with scale, leading many to believe that scale is the only winning recipe. But is it? Here, we investigate an alternative that a priori seems impossible: can smaller language models (e.g., GPT-2) win over models that are orders of magnitude la...
http://arxiv.org/pdf/2212.09246v3
2023-05-26T17:14:27
2022-12-19T04:47:49
{ "title": [ -0.014221191, -0.0064468384, -0.04449463, -0.044036865, 0.01033783, -0.023971558, -0.019088745, -0.02279663, -0.02041626, 0.020706177, 0.0008740425, -0.00020897388, -0.022842407, -0.05029297, -0.025421143, -0.0015945435, 0.011978149, ...
http://arxiv.org/abs/2212.09235v2
PAL: Persona-Augmented Emotional Support Conversation Generation
Jiale Cheng, Sahand Sabour, Hao Sun, Zhuang Chen, Minlie Huang
cs.CL
Due to the lack of human resources for mental health support, there is an increasing demand for employing conversational agents for support. Recent work has demonstrated the effectiveness of dialogue models in providing emotional support. As previous studies have demonstrated that seekers' persona is an important facto...
http://arxiv.org/pdf/2212.09235v2
2023-05-29T06:59:14
2022-12-19T04:12:54
{ "title": [ -0.021896362, -0.0013608932, -0.037750244, -0.051940918, 0.0041999817, 0.008926392, -0.044525146, -0.0057640076, -0.006286621, 0.03881836, -0.0027046204, 0.01335907, 0.026229858, -0.035980225, -0.032226562, 0.014472961, -0.028915405, ...
http://arxiv.org/abs/2212.09233v2
OASum: Large-Scale Open Domain Aspect-based Summarization
Xianjun Yang, Kaiqiang Song, Sangwoo Cho, Xiaoyang Wang, Xiaoman Pan, Linda Petzold, Dong Yu
cs.CL
Aspect or query-based summarization has recently caught more attention, as it can generate differentiated summaries based on users' interests. However, the current dataset for aspect or query-based summarization either focuses on specific domains, contains relatively small-scale instances, or includes only a few aspect...
http://arxiv.org/pdf/2212.09233v2
2023-05-25T22:29:45
2022-12-19T04:04:17
{ "title": [ -0.00041937828, 0.0016326904, -0.042816162, -0.12817383, 0.008377075, 0.02394104, -0.0836792, 0.011642456, 0.010696411, 0.06286621, 0.014076233, 0.017333984, 0.022994995, -0.010681152, -0.04977417, 0.033996582, -0.0021686554, 0.02470...
http://arxiv.org/abs/2212.09196v3
Emergent Analogical Reasoning in Large Language Models
Taylor Webb, Keith J. Holyoak, Hongjing Lu
cs.AI, cs.CL
The recent advent of large language models has reinvigorated debate over whether human cognitive capacities might emerge in such generic models given sufficient training data. Of particular interest is the ability of these models to reason about novel problems zero-shot, without any direct training. In human cognition,...
http://arxiv.org/pdf/2212.09196v3
2023-08-03T03:44:47
2022-12-19T00:04:56
{ "title": [ -0.026565552, -0.019836426, -0.0077590942, -0.055480957, -0.018249512, 0.015640259, -0.034698486, 0.0009851456, 0.023345947, 0.03668213, -0.014839172, 0.0010662079, 0.014221191, -0.006462097, -0.045654297, 0.018585205, -0.013168335, ...
http://arxiv.org/abs/2212.09180v3
Don't Forget Your ABC's: Evaluating the State-of-the-Art in Chat-Oriented Dialogue Systems
Sarah E. Finch, James D. Finch, Jinho D. Choi
cs.CL
Despite tremendous advancements in dialogue systems, stable evaluation still requires human judgments producing notoriously high-variance metrics due to their inherent subjectivity. Moreover, methods and labels in dialogue evaluation are not fully standardized, especially for open-domain chats, with a lack of work to c...
http://arxiv.org/pdf/2212.09180v3
2023-07-28T20:55:09
2022-12-18T22:07:55
{ "title": [ -0.009719849, -0.01158905, -0.01411438, -0.046569824, 0.0035114288, -0.0058631897, -0.019302368, 0.011100769, -0.00061655045, 0.041625977, -0.018707275, 0.0070991516, -0.011299133, -0.010574341, -0.072631836, -0.0059814453, -0.012107849,...
http://arxiv.org/abs/2212.09171v2
Rainproof: An Umbrella To Shield Text Generators From Out-Of-Distribution Data
Maxime Darrin, Pablo Piantanida, Pierre Colombo
cs.CL
Implementing effective control mechanisms to ensure the proper functioning and security of deployed NLP models, from translation to chatbots, is essential. A key ingredient to ensure safe system behaviour is Out-Of-Distribution (OOD) detection, which aims to detect whether an input sample is statistically far from the ...
http://arxiv.org/pdf/2212.09171v2
2023-11-01T19:06:58
2022-12-18T21:22:28
{ "title": [ -0.01828003, -0.039001465, -0.06842041, -0.046844482, -0.042175293, -0.0029296875, -0.04147339, 0.043029785, -0.028656006, 0.0209198, -0.016708374, 0.019882202, 0.034729004, -0.04171753, 0.035125732, -0.011016846, 0.0011663437, 0.001...
http://arxiv.org/abs/2212.09170v2
On Isotropy, Contextualization and Learning Dynamics of Contrastive-based Sentence Representation Learning
Chenghao Xiao, Yang Long, Noura Al Moubayed
cs.CL, cs.AI
Incorporating contrastive learning objectives in sentence representation learning (SRL) has yielded significant improvements on many sentence-level NLP tasks. However, it is not well understood why contrastive learning works for learning sentence-level semantics. In this paper, we aim to help guide future designs of se...
http://arxiv.org/pdf/2212.09170v2
2023-05-26T20:40:57
2022-12-18T21:11:49
{ "title": [ 0.010574341, 0.024856567, -0.017074585, -0.024749756, 0.008270264, -0.009613037, -0.0041007996, -0.033843994, 0.00497818, 0.05480957, -0.041015625, 0.038146973, -0.0053215027, 0.011665344, -0.06604004, -0.0017623901, -0.014030457, 0....
http://arxiv.org/abs/2212.09146v3
Can Retriever-Augmented Language Models Reason? The Blame Game Between the Retriever and the Language Model
Parishad BehnamGhader, Santiago Miret, Siva Reddy
cs.CL
Augmenting pretrained language models with retrievers has shown promise in effectively solving common NLP problems, such as language modeling and question answering. In this paper, we evaluate the strengths and weaknesses of popular retriever-augmented language models, namely kNN-LM, REALM, DPR + FiD, Contriever + ATLA...
http://arxiv.org/pdf/2212.09146v3
2023-11-02T19:12:52
2022-12-18T19:27:41
{ "title": [ -0.008369446, -0.0036373138, -0.027816772, -0.03253174, -0.03692627, -0.012382507, -0.034942627, -0.014770508, -0.03213501, 0.0158844, -0.0131073, -0.014595032, 0.026382446, -0.04711914, 0.003709793, -0.013824463, -0.019714355, 0.059...
http://arxiv.org/abs/2212.09140v2
Unsupervised Discontinuous Constituency Parsing with Mildly Context-Sensitive Grammars
Songlin Yang, Roger P. Levy, Yoon Kim
cs.CL
We study grammar induction with mildly context-sensitive grammars for unsupervised discontinuous parsing. Using the probabilistic linear context-free rewriting system (LCFRS) formalism, our approach fixes the rule structure in advance and focuses on parameter learning with maximum likelihood. To reduce the computationa...
http://arxiv.org/pdf/2212.09140v2
2023-06-09T15:42:50
2022-12-18T18:10:45
{ "title": [ -0.016708374, 0.014701843, -0.028427124, -0.07928467, 0.023803711, 0.019500732, -0.06213379, -0.00080537796, -0.0073890686, 0.014129639, -0.040893555, 0.014465332, -0.030517578, -0.009559631, -0.03604126, -0.013931274, -0.024932861, ...
http://arxiv.org/abs/2212.09125v1
Recall, Expand and Multi-Candidate Cross-Encode: Fast and Accurate Ultra-Fine Entity Typing
Chengyue Jiang, Wenyang Hui, Yong Jiang, Xiaobin Wang, Pengjun Xie, Kewei Tu
cs.CL, cs.AI
Ultra-fine entity typing (UFET) predicts extremely free-formed types (e.g., president, politician) of a given entity mention (e.g., Joe Biden) in context. State-of-the-art (SOTA) methods use the cross-encoder (CE) based architecture. CE concatenates the mention (and its context) with each type and feeds the pairs into ...
http://arxiv.org/pdf/2212.09125v1
2022-12-18T16:42:52
2022-12-18T16:42:52
{ "title": [ -0.013916016, -0.013595581, -0.042816162, -0.09741211, 0.03186035, 0.0024662018, -0.03704834, 0.008163452, 0.01928711, 0.045410156, -0.00919342, 0.033813477, 0.044006348, -0.006931305, -0.028915405, 0.041900635, -0.018661499, 0.04296...
http://arxiv.org/abs/2212.11126v2
Chatbots in a Botnet World
Forrest McKee, David Noever
cs.CR, cs.CL, cs.LG
Question-and-answer formats provide a novel experimental platform for investigating cybersecurity questions. Unlike previous chatbots, the latest ChatGPT model from OpenAI supports an advanced understanding of complex coding questions. The research demonstrates thirteen coding tasks that generally qualify as stages in ...
http://arxiv.org/pdf/2212.11126v2
2022-12-22T17:20:36
2022-12-18T16:08:40
{ "title": [ 0.03942871, -0.02571106, -0.0019779205, -0.006504059, -0.05908203, -0.016662598, 0.038879395, 0.01826477, -0.044921875, 0.06768799, -0.04827881, 0.0015010834, 0.022262573, -0.0073776245, -0.000041663647, -0.032226562, 0.0029754639, 0...
http://arxiv.org/abs/2212.09114v2
CAPSTONE: Curriculum Sampling for Dense Retrieval with Document Expansion
Xingwei He, Yeyun Gong, A-Long Jin, Hang Zhang, Anlei Dong, Jian Jiao, Siu Ming Yiu, Nan Duan
cs.CL
The dual-encoder has become the de facto architecture for dense retrieval. Typically, it computes the latent representations of the query and document independently, thus failing to fully capture the interactions between the query and document. To alleviate this, recent research has focused on obtaining query-informed ...
http://arxiv.org/pdf/2212.09114v2
2023-10-29T09:32:07
2022-12-18T15:57:46
{ "title": [ -0.015731812, -0.013931274, -0.030654907, -0.03137207, 0.00015354156, -0.0127334595, -0.0056648254, -0.025405884, 0.032958984, 0.016921997, -0.06390381, 0.03677368, -0.005153656, 0.0033187866, -0.032043457, 0.0048980713, -0.0015897751, ...
http://arxiv.org/abs/2212.09104v1
LaSQuE: Improved Zero-Shot Classification from Explanations Through Quantifier Modeling and Curriculum Learning
Sayan Ghosh, Rakesh R Menon, Shashank Srivastava
cs.CL
A hallmark of human intelligence is the ability to learn new concepts purely from language. Several recent approaches have explored training machine learning models via natural language supervision. However, these approaches fall short in leveraging linguistic quantifiers (such as 'always' or 'rarely') and mimicking hu...
http://arxiv.org/pdf/2212.09104v1
2022-12-18T15:10:05
2022-12-18T15:10:05
{ "title": [ 0.0068092346, -0.018707275, -0.044647217, -0.04446411, -0.0011825562, -0.038391113, 0.0031108856, -0.011947632, -0.001996994, 0.0026226044, -0.003900528, 0.0057792664, -0.041259766, -0.03250122, -0.010398865, -0.019454956, -0.017974854, ...
http://arxiv.org/abs/2212.09097v2
Continual Knowledge Distillation for Neural Machine Translation
Yuanchi Zhang, Peng Li, Maosong Sun, Yang Liu
cs.CL
While many parallel corpora are not publicly accessible for data copyright, data privacy and competitive differentiation reasons, trained translation models are increasingly available on open platforms. In this work, we propose a method called continual knowledge distillation to take advantage of existing translation m...
http://arxiv.org/pdf/2212.09097v2
2023-06-12T12:00:20
2022-12-18T14:41:13
{ "title": [ -0.018157959, 0.00078201294, -0.039978027, -0.085754395, 0.0064926147, -0.0008149147, -0.06750488, 0.018600464, 0.024139404, 0.052947998, -0.022415161, -0.012550354, 0.018676758, -0.0017108917, -0.06317139, 0.010848999, 0.011375427, ...
http://arxiv.org/abs/2212.09095v2
Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale
Hritik Bansal, Karthik Gopalakrishnan, Saket Dingliwal, Sravan Bodapati, Katrin Kirchhoff, Dan Roth
cs.CL, cs.AI
Language models have been shown to perform better with an increase in scale on a wide variety of tasks via the in-context learning paradigm. In this paper, we investigate the hypothesis that the ability of a large language model to in-context learn-perform a task is not uniformly spread across all of its underlying com...
http://arxiv.org/pdf/2212.09095v2
2023-08-16T09:09:53
2022-12-18T14:36:07
{ "title": [ -0.018920898, 0.00038170815, -0.046813965, -0.027877808, 0.0060539246, -0.018203735, -0.013771057, -0.023391724, -0.014549255, -0.003982544, -0.054016113, 0.000852108, -0.036224365, -0.019485474, -0.02319336, -0.0473938, -0.0011529922, ...
http://arxiv.org/abs/2212.09086v4
PVGRU: Generating Diverse and Relevant Dialogue Responses via Pseudo-Variational Mechanism
Yongkang Liu, Shi Feng, Daling Wang, Yifei Zhang, Hinrich Schütze
cs.CL
We investigate response generation for multi-turn dialogue in generative-based chatbots. Existing generative models based on RNNs (Recurrent Neural Networks) usually employ the last hidden state to summarize the sequences, which makes models unable to capture the subtle variability observed in different dialogues and c...
http://arxiv.org/pdf/2212.09086v4
2023-05-16T11:29:24
2022-12-18T13:36:07
{ "title": [ -0.006904602, -0.014251709, -0.04888916, -0.07183838, 0.0074691772, -0.008888245, -0.032409668, 0.015304565, -0.006626129, 0.0262146, 0.027877808, -0.0011692047, -0.01977539, -0.03869629, -0.059417725, 0.0003874302, -0.0018167496, 0....
http://arxiv.org/abs/2212.09080v1
Synthesis and Evaluation of a Domain-specific Large Data Set for Dungeons & Dragons
Akila Peiris, Nisansa de Silva
cs.CL, cs.LG
This paper introduces the Forgotten Realms Wiki (FRW) data set and domain specific natural language generation using FRW along with related analyses. Forgotten Realms is the de-facto default setting of the popular open ended tabletop fantasy role playing game, Dungeons & Dragons. The data set was extracted from the For...
http://arxiv.org/pdf/2212.09080v1
2022-12-18T12:54:45
2022-12-18T12:54:45
{ "title": [ 0.006713867, 0.009483337, -0.066467285, -0.038085938, 0.020965576, -0.041137695, -0.0078048706, -0.006526947, -0.02508545, 0.007621765, -0.027145386, -0.007949829, 0.0069999695, -0.040985107, 0.010253906, -0.032409668, 0.005508423, 0...
http://arxiv.org/abs/2212.09072v1
Let's Negotiate! A Survey of Negotiation Dialogue Systems
Haolan Zhan, Yufei Wang, Tao Feng, Yuncheng Hua, Suraj Sharma, Zhuang Li, Lizhen Qu, Gholamreza Haffari
cs.CL
Negotiation is one of the crucial abilities in human communication, and there has been a resurgent research interest in negotiation dialogue systems recently, which goal is to empower intelligent agents with such ability that can efficiently help humans resolve conflicts or reach beneficial agreements. Although there h...
http://arxiv.org/pdf/2212.09072v1
2022-12-18T12:03:53
2022-12-18T12:03:53
{ "title": [ -0.021469116, -0.00970459, -0.019989014, -0.043304443, 0.015281677, 0.0016546249, -0.008850098, -0.0032539368, -0.008361816, 0.021728516, -0.015289307, 0.008781433, -0.038238525, -0.023742676, -0.050598145, -0.02017212, -0.019042969, ...
http://arxiv.org/abs/2212.09058v1
BEATs: Audio Pre-Training with Acoustic Tokenizers
Sanyuan Chen, Yu Wu, Chengyi Wang, Shujie Liu, Daniel Tompkins, Zhuo Chen, Furu Wei
eess.AS, cs.AI, cs.CL, cs.LG, cs.SD
The massive growth of self-supervised learning (SSL) has been witnessed in language, vision, speech, and audio domains over the past few years. While discrete label prediction is widely adopted for other modalities, the state-of-the-art audio SSL models still employ reconstruction loss for pre-training. Compared with r...
http://arxiv.org/pdf/2212.09058v1
2022-12-18T10:41:55
2022-12-18T10:41:55
{ "title": [ -0.011299133, -0.012924194, -0.04699707, -0.059783936, -0.0049476624, -0.0063934326, -0.046173096, 0.0054244995, -0.05001831, 0.052246094, -0.02355957, 0.04067993, 0.026397705, -0.025558472, -0.00004953146, -0.007965088, 0.0082473755, ...
http://arxiv.org/abs/2212.09056v1
Beyond Digital "Echo Chambers": The Role of Viewpoint Diversity in Political Discussion
Rishav Hada, Amir Ebrahimi Fard, Sarah Shugars, Federico Bianchi, Patricia Rossini, Dirk Hovy, Rebekah Tromble, Nava Tintarev
cs.CL
Increasingly taking place in online spaces, modern political conversations are typically perceived to be unproductively affirming -- siloed in so called ``echo chambers'' of exclusively like-minded discussants. Yet, to date we lack sufficient means to measure viewpoint diversity in conversations. To this end, in this p...
http://arxiv.org/pdf/2212.09056v1
2022-12-18T10:18:15
2022-12-18T10:18:15
{ "title": [ 0.016067505, 0.015853882, -0.048065186, -0.051757812, 0.047332764, -0.04147339, -0.0008134842, -0.02017212, -0.04522705, 0.022521973, -0.004852295, 0.017242432, -0.033355713, -0.029022217, -0.033966064, -0.01637268, -0.003929138, 0.0...
http://arxiv.org/abs/2212.09052v1
A Better Choice: Entire-space Datasets for Aspect Sentiment Triplet Extraction
Yuncong Li, Fang Wang, Sheng-Hua Zhong
cs.CL
Aspect sentiment triplet extraction (ASTE) aims to extract aspect term, sentiment and opinion term triplets from sentences. Since the initial datasets used to evaluate models on ASTE had flaws, several studies later corrected the initial datasets and released new versions of the datasets independently. As a result, dif...
http://arxiv.org/pdf/2212.09052v1
2022-12-18T09:57:40
2022-12-18T09:57:40
{ "title": [ 0.01473999, -0.0074539185, -0.023254395, -0.07556152, 0.008735657, -0.0076408386, -0.032989502, -0.0030345917, 0.040496826, 0.055664062, -0.024673462, 0.0065574646, -0.005756378, 0.011390686, -0.09265137, 0.015541077, 0.00021111965, ...
http://arxiv.org/abs/2212.09045v1
Task Preferences across Languages on Community Question Answering Platforms
Sebastin Santy, Prasanta Bhattacharya, Rishabh Mehrotra
cs.CL, cs.CY, cs.HC, cs.SI
With the steady emergence of community question answering (CQA) platforms like Quora, StackExchange, and WikiHow, users now have an unprecedented access to information on various kind of queries and tasks. Moreover, the rapid proliferation and localization of these platforms spanning geographic and linguistic boundarie...
http://arxiv.org/pdf/2212.09045v1
2022-12-18T09:37:20
2022-12-18T09:37:20
{ "title": [ -0.011947632, -0.01826477, -0.030670166, -0.045043945, -0.011192322, -0.014045715, -0.020095825, 0.019332886, 0.0037975311, 0.037963867, -0.014816284, -0.0019168854, 0.022155762, -0.009841919, -0.04006958, 0.02003479, -0.0006914139, ...
http://arxiv.org/abs/2212.09028v1
Neural Coreference Resolution based on Reinforcement Learning
Yu Wang, Hongxia Jin
cs.CL
The target of a coreference resolution system is to cluster all mentions that refer to the same entity in a given context. All coreference resolution systems need to solve two subtasks; one task is to detect all of the potential mentions, and the other is to learn the linking of an antecedent for each possible mention....
http://arxiv.org/pdf/2212.09028v1
2022-12-18T07:36:35
2022-12-18T07:36:35
{ "title": [ 0.017913818, 0.010314941, -0.009017944, -0.055999756, 0.007385254, -0.016815186, -0.014343262, -0.0016155243, -0.022994995, 0.057800293, -0.0030155182, 0.00945282, 0.0059280396, -0.018798828, -0.034210205, 0.0141067505, -0.005554199, ...
http://arxiv.org/abs/2212.08999v1
Sentence-level Feedback Generation for English Language Learners: Does Data Augmentation Help?
Shabnam Behzad, Amir Zeldes, Nathan Schneider
cs.CL
In this paper, we present strong baselines for the task of Feedback Comment Generation for Writing Learning. Given a sentence and an error span, the task is to generate a feedback comment explaining the error. Sentences and feedback comments are both in English. We experiment with LLMs and also create multiple pseudo d...
http://arxiv.org/pdf/2212.08999v1
2022-12-18T03:53:44
2022-12-18T03:53:44
{ "title": [ 0.0026073456, -0.00881958, -0.0061836243, -0.013381958, -0.032989502, 0.0033950806, -0.017120361, -0.01977539, -0.026260376, 0.028167725, -0.030563354, 0.033569336, -0.016281128, -0.0073432922, -0.038513184, -0.017578125, -0.027008057, ...
http://arxiv.org/abs/2212.08995v1
Impact of Sentiment Analysis in Fake Review Detection
Amira Yousif, James Buckley
cs.CL, cs.LG
Fake review identification is an important topic and has gained the interest of experts all around the world. Identifying fake reviews is challenging for researchers, and there are several primary challenges to fake review detection. We propose developing an initial research paper for investigating fake reviews by usin...
http://arxiv.org/pdf/2212.08995v1
2022-12-18T03:17:47
2022-12-18T03:17:47
{ "title": [ 0.0016469955, 0.002128601, -0.008705139, -0.001789093, 0.0049209595, -0.01461792, 0.010108948, -0.0418396, -0.0413208, 0.03149414, -0.03024292, 0.007663727, -0.010108948, -0.011642456, -0.029067993, -0.016906738, -0.026565552, 0.0626...
http://arxiv.org/abs/2212.08992v1
PoE: a Panel of Experts for Generalized Automatic Dialogue Assessment
Chen Zhang, Luis Fernando D'Haro, Qiquan Zhang, Thomas Friedrichs, Haizhou Li
cs.CL
Chatbots are expected to be knowledgeable across multiple domains, e.g. for daily chit-chat, exchange of information, and grounding in emotional situations. To effectively measure the quality of such conversational agents, a model-based automatic dialogue evaluation metric (ADEM) is expected to perform well across mult...
http://arxiv.org/pdf/2212.08992v1
2022-12-18T02:26:50
2022-12-18T02:26:50
{ "title": [ -0.010269165, -0.00023329258, -0.04196167, -0.059020996, 0.008796692, -0.0030593872, -0.033966064, -0.0052604675, 0.02130127, 0.016113281, -0.0036449432, -0.008178711, -0.008995056, -0.023361206, -0.060302734, 0.004306793, -0.02267456, ...
http://arxiv.org/abs/2212.08987v1
A Robust Semantic Frame Parsing Pipeline on a New Complex Twitter Dataset
Yu Wang, Hongxia Jin
cs.CL, cs.AI
Most recent semantic frame parsing systems for spoken language understanding (SLU) are designed based on recurrent neural networks. These systems display decent performance on benchmark SLU datasets such as ATIS or SNIPS, which contain short utterances with relatively simple patterns. However, the current semantic fram...
http://arxiv.org/pdf/2212.08987v1
2022-12-18T01:59:49
2022-12-18T01:59:49
{ "title": [ 0.0061035156, 0.002565384, -0.034484863, -0.099853516, 0.016983032, 0.0137786865, -0.06286621, 0.00548172, -0.0110321045, 0.04159546, -0.04815674, 0.03643799, 0.0010795593, 0.012252808, -0.04168701, 0.0036449432, -0.01424408, 0.04830...
http://arxiv.org/abs/2212.08986v2
Low-Resource Authorship Style Transfer: Can Non-Famous Authors Be Imitated?
Ajay Patel, Nicholas Andrews, Chris Callison-Burch
cs.CL
Authorship style transfer involves altering text to match the style of a target author whilst preserving the original meaning. Existing unsupervised approaches like STRAP have largely focused on style transfer to target authors with many examples of their writing style in books, speeches, or other published works. This...
http://arxiv.org/pdf/2212.08986v2
2023-08-23T16:31:34
2022-12-18T01:57:30
{ "title": [ -0.035095215, 0.0012464523, -0.043548584, 0.0012559891, -0.011650085, -0.019134521, -0.013877869, -0.033172607, -0.050933838, 0.04434204, 0.013465881, 0.03262329, 0.014892578, -0.023239136, -0.0041160583, 0.011604309, -0.018829346, 0...
http://arxiv.org/abs/2212.08979v1
Language model acceptability judgements are not always robust to context
Koustuv Sinha, Jon Gauthier, Aaron Mueller, Kanishka Misra, Keren Fuentes, Roger Levy, Adina Williams
cs.CL, cs.LG
Targeted syntactic evaluations of language models ask whether models show stable preferences for syntactically acceptable content over minimal-pair unacceptable inputs. Most targeted syntactic evaluation datasets ask models to make these judgements with just a single context-free sentence as input. This does not match ...
http://arxiv.org/pdf/2212.08979v1
2022-12-18T00:11:06
2022-12-18T00:11:06
{ "title": [ -0.0018100739, 0.023590088, -0.009277344, -0.004863739, -0.027664185, -0.017471313, -0.0008263588, -0.029571533, -0.010009766, 0.024047852, -0.0680542, -0.032836914, -0.0029697418, -0.028442383, -0.027053833, -0.037261963, -0.0007953644,...
http://arxiv.org/abs/2212.08950v1
Beyond the C: Retargetable Decompilation using Neural Machine Translation
Iman Hosseini, Brendan Dolan-Gavitt
cs.CR, cs.CL, cs.PL
The problem of reversing the compilation process, decompilation, is an important tool in reverse engineering of computer software. Recently, researchers have proposed using techniques from neural machine translation to automate the process in decompilation. Although such techniques hold the promise of targeting a wider...
http://arxiv.org/pdf/2212.08950v1
2022-12-17T20:45:59
2022-12-17T20:45:59
{ "title": [ -0.044525146, 0.0016908646, 0.0058555603, -0.07208252, -0.010101318, 0.05215454, -0.09350586, -0.00969696, -0.012512207, 0.027893066, 0.009338379, 0.0138168335, 0.0129776, -0.022872925, 0.002840042, 0.026184082, -0.0027179718, 0.0199...
http://arxiv.org/abs/2212.08946v1
Towards leveraging latent knowledge and Dialogue context for real-world conversational question answering
Shaomu Tan, Denis Paperno
cs.CL
In many real-world scenarios, the absence of external knowledge source like Wikipedia restricts question answering systems to rely on latent internal knowledge in limited dialogue data. In addition, humans often seek answers by asking several questions for more comprehensive information. As the dialog becomes more exte...
http://arxiv.org/pdf/2212.08946v1
2022-12-17T20:36:17
2022-12-17T20:36:17
{ "title": [ -0.003929138, -0.017166138, -0.043823242, -0.067871094, 0.0012397766, -0.014389038, -0.026153564, 0.031433105, 0.023529053, 0.029586792, -0.023834229, 0.00018334389, 0.00042581558, -0.0049209595, -0.056854248, -0.003534317, 0.009262085, ...
http://arxiv.org/abs/2212.08929v2
Modeling Instance Interactions for Joint Information Extraction with Neural High-Order Conditional Random Field
Zixia Jia, Zhaohui Yan, Wenjuan Han, Zilong Zheng, Kewei Tu
cs.CL
Prior works on joint Information Extraction (IE) typically model instance (e.g., event triggers, entities, roles, relations) interactions by representation enhancement, type dependencies scoring, or global decoding. We find that the previous models generally consider binary type dependency scoring of a pair of instance...
http://arxiv.org/pdf/2212.08929v2
2023-05-28T09:48:24
2022-12-17T18:45:23
{ "title": [ 0.0034923553, 0.017333984, -0.06750488, -0.08654785, 0.017990112, -0.009887695, -0.05895996, -0.010765076, -0.0062065125, 0.0051574707, -0.024856567, 0.0067100525, -0.015655518, -0.01574707, -0.02142334, -0.020004272, -0.016174316, 0...
http://arxiv.org/abs/2212.08926v1
A Simple Baseline for Beam Search Reranking
Lior Vassertail, Omer Levy
cs.CL, cs.AI, cs.LG
Reranking methods in machine translation aim to close the gap between common evaluation metrics (e.g. BLEU) and maximum likelihood learning and decoding algorithms. Prior works address this challenge by training models to rerank beam search candidates according to their predicted BLEU scores, building upon large models...
http://arxiv.org/pdf/2212.08926v1
2022-12-17T18:22:20
2022-12-17T18:22:20
{ "title": [ -0.01576233, 0.0075149536, 0.0074768066, -0.028518677, -0.021072388, 0.02154541, -0.042510986, -0.06088257, 0.01902771, 0.043762207, 0.0059547424, -0.0072517395, -0.00042128563, -0.034057617, -0.049591064, 0.025238037, 0.001033783, -...
http://arxiv.org/abs/2212.08913v2
Claim Optimization in Computational Argumentation
Gabriella Skitalinskaya, Maximilian Spliethöver, Henning Wachsmuth
cs.CL
An optimal delivery of arguments is key to persuasion in any debate, both for humans and for AI systems. This requires the use of clear and fluent claims relevant to the given debate. Prior work has studied the automatic assessment of argument quality extensively. Yet, no approach actually improves the quality so far. ...
http://arxiv.org/pdf/2212.08913v2
2023-09-07T09:01:20
2022-12-17T16:30:27
{ "title": [ 0.016311646, -0.0024700165, -0.0025596619, -0.0703125, -0.009941101, -0.0037727356, -0.018218994, 0.002243042, -0.04135132, 0.025787354, 0.01108551, 0.011299133, -0.02519226, -0.03665161, -0.016998291, -0.005012512, -0.015716553, 0.0...
http://arxiv.org/abs/2212.08911v1
AdaTranS: Adapting with Boundary-based Shrinking for End-to-End Speech Translation
Xingshan Zeng, Liangyou Li, Qun Liu
cs.CL, cs.SD, eess.AS
To alleviate the data scarcity problem in End-to-end speech translation (ST), pre-training on data for speech recognition and machine translation is considered as an important technique. However, the modality gap between speech and text prevents the ST model from efficiently inheriting knowledge from the pre-trained mo...
http://arxiv.org/pdf/2212.08911v1
2022-12-17T16:14:30
2022-12-17T16:14:30
{ "title": [ -0.049102783, -0.011581421, -0.033966064, -0.05279541, -0.001083374, 0.021835327, -0.050567627, 0.014389038, -0.016189575, 0.06060791, -0.021469116, 0.027389526, 0.008155823, -0.00983429, -0.047302246, 0.0124053955, -0.004047394, 0.0...
http://arxiv.org/abs/2212.08909v2
Controlling Styles in Neural Machine Translation with Activation Prompt
Yifan Wang, Zewei Sun, Shanbo Cheng, Weiguo Zheng, Mingxuan Wang
cs.CL
Controlling styles in neural machine translation (NMT) has attracted wide attention, as it is crucial for enhancing user experience. Earlier studies on this topic typically concentrate on regulating the level of formality and achieve some progress in this area. However, they still encounter two major challenges. The fi...
http://arxiv.org/pdf/2212.08909v2
2023-05-29T02:59:07
2022-12-17T16:05:50
{ "title": [ -0.033996582, -0.00017893314, -0.032562256, -0.014579773, -0.008666992, -0.0017881393, -0.022598267, -0.00025892258, -0.008277893, 0.06304932, 0.014152527, 0.004421234, 0.03387451, -0.022415161, -0.038970947, 0.03817749, -0.0069465637, ...
http://arxiv.org/abs/2212.08902v2
Know What I don't Know: Handling Ambiguous and Unanswerable Questions for Text-to-SQL
Bing Wang, Yan Gao, Zhoujun Li, Jian-Guang Lou
cs.CL
The task of text-to-SQL aims to convert a natural language question into its corresponding SQL query within the context of relational tables. Existing text-to-SQL parsers generate a "plausible" SQL query for an arbitrary user question, thereby failing to correctly handle problematic user questions. To formalize this pr...
http://arxiv.org/pdf/2212.08902v2
2023-05-19T14:27:31
2022-12-17T15:32:00
{ "title": [ -0.032470703, 0.00037670135, -0.053741455, -0.054595947, 0.0012235641, 0.001745224, -0.021408081, -0.0042915344, 0.02015686, 0.025665283, -0.04800415, 0.0027885437, -0.0037975311, -0.009437561, -0.058563232, -0.0357666, 0.012817383, ...
http://arxiv.org/abs/2212.08897v1
Improving Question Answering Performance through Manual Annotation: Costs, Benefits and Strategies
Piotr Rybak, Piotr Przybyła, Maciej Ogrodniczuk
cs.CL
Recently proposed systems for open-domain question answering (OpenQA) require large amounts of training data to achieve state-of-the-art performance. However, data annotation is known to be time-consuming and therefore expensive to acquire. As a result, the appropriate datasets are available only for a handful of langu...
http://arxiv.org/pdf/2212.08897v1
2022-12-17T15:20:18
2022-12-17T15:20:18
{ "title": [ 0.003829956, -0.0041923523, -0.012130737, -0.049560547, -0.011001587, -0.008102417, -0.018737793, -0.006839752, -0.002149582, 0.05529785, 0.0035190582, 0.024002075, 0.014076233, -0.0047721863, -0.0715332, 0.018432617, -0.0020046234, ...
http://arxiv.org/abs/2212.08888v1
Exploiting Rich Textual User-Product Context for Improving Sentiment Analysis
Chenyang Lyu, Linyi Yang, Yue Zhang, Yvette Graham, Jennifer Foster
cs.CL
User and product information associated with a review is useful for sentiment polarity prediction. Typical approaches incorporating such information focus on modeling users and products as implicitly learned representation vectors. Most do not exploit the potential of historical reviews, or those that currently do requ...
http://arxiv.org/pdf/2212.08888v1
2022-12-17T14:57:52
2022-12-17T14:57:52
{ "title": [ 0.0051002502, -0.021484375, -0.019104004, -0.052001953, 0.008216858, -0.0045051575, -0.0077705383, -0.008735657, 0.023162842, 0.04537964, -0.038879395, 0.008201599, 0.000021576881, 0.008018494, -0.07849121, -0.003168106, -0.014892578, ...