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/2305.16263v1
Unified Modeling of Multi-Talker Overlapped Speech Recognition and Diarization with a Sidecar Separator
Lingwei Meng, Jiawen Kang, Mingyu Cui, Haibin Wu, Xixin Wu, Helen Meng
cs.SD, cs.AI, cs.CL, cs.LG, eess.AS
Multi-talker overlapped speech poses a significant challenge for speech recognition and diarization. Recent research indicated that these two tasks are inter-dependent and complementary, motivating us to explore a unified modeling method to address them in the context of overlapped speech. A recent study proposed a cos...
http://arxiv.org/pdf/2305.16263v1
2023-05-25T17:18:37
2023-05-25T17:18:37
{ "title": [ 0.0023822784, 0.01537323, -0.061584473, -0.08135986, 0.009651184, -0.001742363, -0.05886841, 0.024368286, -0.05392456, 0.05682373, -0.035491943, 0.010108948, 0.025268555, -0.015487671, 0.011123657, -0.019943237, -0.021865845, 0.04483...
http://arxiv.org/abs/2305.16259v5
Neural Natural Language Processing for Long Texts: A Survey of the State-of-the-Art
Dimitrios Tsirmpas, Ioannis Gkionis, Ioannis Mademlis, Georgios Papadopoulos
cs.CL, cs.AI, I.2.7
The adoption of Deep Neural Networks (DNNs) has greatly benefited Natural Language Processing (NLP) during the past decade. However, the demands of long document analysis are quite different from those of shorter texts, while the ever increasing size of documents uploaded on-line renders automated understanding of leng...
http://arxiv.org/pdf/2305.16259v5
2023-07-23T20:00:46
2023-05-25T17:13:44
{ "title": [ -0.019805908, 0.0046653748, -0.019439697, -0.042877197, 0.00548172, 0.007167816, -0.025863647, -0.0034713745, 0.0014438629, 0.031219482, -0.007843018, -0.001162529, -0.0016841888, -0.010375977, -0.046295166, -0.0016307831, -0.022491455, ...
http://arxiv.org/abs/2305.16253v2
Uncovering and Categorizing Social Biases in Text-to-SQL
Yan Liu, Yan Gao, Zhe Su, Xiaokang Chen, Elliott Ash, Jian-Guang Lou
cs.CL
Content Warning: This work contains examples that potentially implicate stereotypes, associations, and other harms that could be offensive to individuals in certain social groups.} Large pre-trained language models are acknowledged to carry social biases towards different demographics, which can further amplify existin...
http://arxiv.org/pdf/2305.16253v2
2023-06-07T13:30:39
2023-05-25T17:08:56
{ "title": [ 0.0047187805, 0.0014076233, -0.03665161, -0.054626465, 0.025054932, -0.015312195, 0.0012197495, -0.00050115585, 0.011421204, 0.03781128, -0.04257202, 0.009918213, 0.0034637451, 0.0021476746, -0.05368042, -0.027709961, -0.025466919, 0...
http://arxiv.org/abs/2305.16252v1
Overcoming Catastrophic Forgetting in Massively Multilingual Continual Learning
Genta Indra Winata, Lingjue Xie, Karthik Radhakrishnan, Shijie Wu, Xisen Jin, Pengxiang Cheng, Mayank Kulkarni, Daniel Preotiuc-Pietro
cs.CL
Real-life multilingual systems should be able to efficiently incorporate new languages as data distributions fed to the system evolve and shift over time. To do this, systems need to handle the issue of catastrophic forgetting, where the model performance drops for languages or tasks seen further in its past. In this p...
http://arxiv.org/pdf/2305.16252v1
2023-05-25T17:06:34
2023-05-25T17:06:34
{ "title": [ -0.009468079, 0.009727478, -0.039764404, -0.07696533, 0.011688232, 0.0014591217, -0.03262329, 0.0069007874, -0.00027608871, 0.044067383, -0.0014810562, 0.012062073, -0.009086609, -0.013710022, -0.042144775, 0.006111145, -0.0034694672, ...
http://arxiv.org/abs/2305.16243v3
Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models
Ehsan Doostmohammadi, Tobias Norlund, Marco Kuhlmann, Richard Johansson
cs.CL
Augmenting language models with a retrieval mechanism has been shown to significantly improve their performance while keeping the number of parameters low. Retrieval-augmented models commonly rely on a semantic retrieval mechanism based on the similarity between dense representations of the query chunk and potential ne...
http://arxiv.org/pdf/2305.16243v3
2023-07-04T07:59:15
2023-05-25T16:56:26
{ "title": [ -0.005882263, 0.01235199, -0.00080394745, -0.017745972, -0.028640747, -0.016494751, -0.04714966, -0.036499023, 0.001045227, 0.04458618, -0.011268616, 0.010108948, 0.03677368, -0.030685425, -0.020446777, 0.019088745, -0.004711151, 0.0...
http://arxiv.org/abs/2305.16199v2
Diversity-Aware Coherence Loss for Improving Neural Topic Models
Raymond Li, Felipe González-Pizarro, Linzi Xing, Gabriel Murray, Giuseppe Carenini
cs.CL, cs.LG
The standard approach for neural topic modeling uses a variational autoencoder (VAE) framework that jointly minimizes the KL divergence between the estimated posterior and prior, in addition to the reconstruction loss. Since neural topic models are trained by recreating individual input documents, they do not explicitl...
http://arxiv.org/pdf/2305.16199v2
2023-05-26T09:59:49
2023-05-25T16:01:56
{ "title": [ -0.018127441, 0.023208618, -0.046020508, -0.028640747, 0.020889282, -0.017715454, -0.03201294, -0.03326416, 0.015586853, 0.054870605, 0.011489868, 0.007843018, 0.020874023, -0.015434265, -0.041900635, 0.035858154, -0.013122559, 0.018...
http://arxiv.org/abs/2305.16380v4
Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer
Yuandong Tian, Yiping Wang, Beidi Chen, Simon Du
cs.CL, cs.AI, cs.LG
Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \emph{training dynamic...
http://arxiv.org/pdf/2305.16380v4
2023-10-30T17:32:08
2023-05-25T15:59:13
{ "title": [ -0.001373291, 0.00756073, -0.027572632, -0.046722412, 0.0045394897, -0.013946533, -0.011756897, -0.010253906, -0.021820068, 0.01651001, -0.015777588, 0.0138549805, -0.0065956116, -0.030136108, -0.00554657, -0.013084412, -0.015380859, ...
http://arxiv.org/abs/2305.16195v1
Abstractive Summary Generation for the Urdu Language
Ali Raza, Hadia Sultan Raja, Usman Maratib
cs.CL, cs.AI, 68T50 (Primary) 03B65, 91F20 (Secondary), I.2; I.7
Abstractive summary generation is a challenging task that requires the model to comprehend the source text and generate a concise and coherent summary that captures the essential information. In this paper, we explore the use of an encoder/decoder approach for abstractive summary generation in the Urdu language. We emp...
http://arxiv.org/pdf/2305.16195v1
2023-05-25T15:55:42
2023-05-25T15:55:42
{ "title": [ -0.02961731, -0.0074310303, -0.042175293, -0.079833984, -0.02961731, 0.016464233, -0.085998535, 0.023269653, -0.0026779175, 0.05596924, -0.023605347, 0.013450623, -0.000114142895, -0.015083313, -0.048309326, 0.002670288, 0.0079422, 0...
http://arxiv.org/abs/2305.16183v2
Passive learning of active causal strategies in agents and language models
Andrew Kyle Lampinen, Stephanie C Y Chan, Ishita Dasgupta, Andrew J Nam, Jane X Wang
cs.LG, cs.AI, cs.CL
What can be learned about causality and experimentation from passive data? This question is salient given recent successes of passively-trained language models in interactive domains such as tool use. Passive learning is inherently limited. However, we show that purely passive learning can in fact allow an agent to lea...
http://arxiv.org/pdf/2305.16183v2
2023-10-02T13:04:59
2023-05-25T15:39:46
{ "title": [ -0.006790161, -0.009597778, -0.0027332306, -0.05307007, -0.0078125, -0.0074920654, -0.016067505, -0.0024738312, -0.009033203, 0.026168823, -0.004070282, 0.017669678, -0.0043029785, -0.024871826, -0.010818481, 0.012229919, -0.004234314, ...
http://arxiv.org/abs/2305.16171v1
Multi-lingual and Multi-cultural Figurative Language Understanding
Anubha Kabra, Emmy Liu, Simran Khanuja, Alham Fikri Aji, Genta Indra Winata, Samuel Cahyawijaya, Anuoluwapo Aremu, Perez Ogayo, Graham Neubig
cs.CL
Figurative language permeates human communication, but at the same time is relatively understudied in NLP. Datasets have been created in English to accelerate progress towards measuring and improving figurative language processing in language models (LMs). However, the use of figurative language is an expression of our...
http://arxiv.org/pdf/2305.16171v1
2023-05-25T15:30:31
2023-05-25T15:30:31
{ "title": [ -0.003929138, 0.0023994446, -0.023483276, -0.045410156, 0.0019836426, -0.023162842, -0.031082153, -0.014381409, -0.009483337, 0.035308838, 0.00047421455, -0.018508911, 0.040924072, -0.043121338, -0.0077705383, 0.012908936, 0.0020256042, ...
http://arxiv.org/abs/2305.16166v1
Multimodal Relation Extraction with Cross-Modal Retrieval and Synthesis
Xuming Hu, Zhijiang Guo, Zhiyang Teng, Irwin King, Philip S. Yu
cs.CL
Multimodal relation extraction (MRE) is the task of identifying the semantic relationships between two entities based on the context of the sentence image pair. Existing retrieval-augmented approaches mainly focused on modeling the retrieved textual knowledge, but this may not be able to accurately identify complex rel...
http://arxiv.org/pdf/2305.16166v1
2023-05-25T15:26:13
2023-05-25T15:26:13
{ "title": [ -0.014137268, 0.0068244934, -0.05609131, -0.06915283, 0.011657715, -0.0045166016, -0.054840088, -0.011070251, 0.04550171, 0.021896362, -0.014511108, 0.0047721863, 0.013710022, -0.013114929, -0.048065186, 0.013000488, -0.012496948, 0....
http://arxiv.org/abs/2305.16162v1
Feature Collapse
Thomas Laurent, James H. von Brecht, Xavier Bresson
cs.LG, cs.AI, cs.CL
We formalize and study a phenomenon called feature collapse that makes precise the intuitive idea that entities playing a similar role in a learning task receive similar representations. As feature collapse requires a notion of task, we leverage a simple but prototypical NLP task to study it. We start by showing experi...
http://arxiv.org/pdf/2305.16162v1
2023-05-25T15:25:34
2023-05-25T15:25:34
{ "title": [ 0.020904541, 0.020584106, -0.015548706, -0.017974854, -0.011169434, -0.02204895, -0.036956787, -0.046020508, 0.032104492, 0.044525146, -0.048217773, -0.035949707, 0.0032196045, -0.025970459, -0.020462036, -0.019470215, 0.03933716, 0....
http://arxiv.org/abs/2305.16157v1
Training Data Extraction From Pre-trained Language Models: A Survey
Shotaro Ishihara
cs.CL
As the deployment of pre-trained language models (PLMs) expands, pressing security concerns have arisen regarding the potential for malicious extraction of training data, posing a threat to data privacy. This study is the first to provide a comprehensive survey of training data extraction from PLMs. Our review covers m...
http://arxiv.org/pdf/2305.16157v1
2023-05-25T15:23:29
2023-05-25T15:23:29
{ "title": [ -0.030700684, -0.003124237, -0.043914795, -0.07043457, 0.0075263977, 0.0024223328, -0.063964844, 0.0059165955, 0.0033512115, 0.028137207, -0.007358551, -0.0013093948, -0.01600647, -0.018463135, -0.05810547, 0.00712204, -0.011177063, ...
http://arxiv.org/abs/2305.16155v2
Revisiting Non-Autoregressive Translation at Scale
Zhihao Wang, Longyue Wang, Jinsong Su, Junfeng Yao, Zhaopeng Tu
cs.CL
In real-world systems, scaling has been critical for improving the translation quality in autoregressive translation (AT), which however has not been well studied for non-autoregressive translation (NAT). In this work, we bridge the gap by systematically studying the impact of scaling on NAT behaviors. Extensive experi...
http://arxiv.org/pdf/2305.16155v2
2023-06-02T13:58:00
2023-05-25T15:22:47
{ "title": [ -0.062683105, -0.007896423, -0.0423584, -0.042907715, -0.021697998, -0.0002310276, -0.062805176, -0.0054359436, -0.014480591, 0.04248047, -0.022598267, 0.025161743, 0.0003941059, -0.002828598, -0.041503906, 0.007587433, 0.0028476715, ...
http://arxiv.org/abs/2305.18340v2
Mapping ChatGPT in Mainstream Media to Unravel Jobs and Diversity Challenges: Early Quantitative Insights through Sentiment Analysis and Word Frequency Analysis
Maya Karanouh
cs.CY, cs.CL
The exponential growth in user acquisition and popularity of OpenAIs ChatGPT, an artificial intelligence(AI) powered chatbot, was accompanied by widespread mainstream media coverage. This article presents a quantitative data analysis of the early trends and sentiments revealed by conducting text mining and NLP methods ...
http://arxiv.org/pdf/2305.18340v2
2023-08-03T19:21:02
2023-05-25T15:10:51
{ "title": [ 0.009017944, -0.009262085, 0.007171631, 0.03765869, -0.019699097, -0.026748657, 0.029846191, -0.03491211, -0.029953003, 0.034576416, -0.028411865, 0.03010559, 0.005580902, -0.009048462, -0.012626648, -0.02458191, -0.026123047, 0.0206...
http://arxiv.org/abs/2305.18339v2
A Survey on ChatGPT: AI-Generated Contents, Challenges, and Solutions
Yuntao Wang, Yanghe Pan, Miao Yan, Zhou Su, Tom H. Luan
cs.CY, cs.AI, cs.CL
With the widespread use of large artificial intelligence (AI) models such as ChatGPT, AI-generated content (AIGC) has garnered increasing attention and is leading a paradigm shift in content creation and knowledge representation. AIGC uses generative large AI algorithms to assist or replace humans in creating massive, ...
http://arxiv.org/pdf/2305.18339v2
2023-07-30T02:31:58
2023-05-25T15:09:11
{ "title": [ -0.013755798, -0.013587952, -0.013320923, -0.030807495, -0.0017547607, 0.006351471, -0.0075950623, -0.017608643, -0.0063934326, 0.031234741, -0.006668091, 0.03604126, -0.010879517, -0.015701294, -0.017669678, 0.02722168, -0.01939392, ...
http://arxiv.org/abs/2305.16130v2
A Mechanism for Solving Relational Tasks in Transformer Language Models
Jack Merullo, Carsten Eickhoff, Ellie Pavlick
cs.CL, cs.LG
A primary criticism towards language models (LMs) is their inscrutability. This paper presents evidence that, despite their size and complexity, LMs sometimes exploit a simple computational mechanism to solve one-to-one relational tasks (e.g., capital_of(Poland)=Warsaw). We investigate a range of language model sizes (...
http://arxiv.org/pdf/2305.16130v2
2023-10-12T21:43:18
2023-05-25T15:04:01
{ "title": [ -0.022445679, -0.013496399, -0.031829834, -0.061920166, 0.00086927414, 0.0043945312, -0.07324219, 0.005050659, 0.005092621, 0.032440186, -0.0037326813, -0.012619019, 0.026031494, -0.02520752, -0.032470703, 0.013046265, -0.013175964, ...
http://arxiv.org/abs/2305.16128v1
Give Me More Details: Improving Fact-Checking with Latent Retrieval
Xuming Hu, Zhijiang Guo, Guanyu Wu, Lijie Wen, Philip S. Yu
cs.CL
Evidence plays a crucial role in automated fact-checking. When verifying real-world claims, existing fact-checking systems either assume the evidence sentences are given or use the search snippets returned by the search engine. Such methods ignore the challenges of collecting evidence and may not provide sufficient inf...
http://arxiv.org/pdf/2305.16128v1
2023-05-25T15:01:19
2023-05-25T15:01:19
{ "title": [ 0.03125, 0.0026016235, -0.03933716, -0.0703125, 0.0064888, -0.030319214, -0.012710571, -0.014640808, -0.05166626, 0.037200928, -0.029586792, 0.044158936, -0.0025100708, -0.01474762, -0.0085372925, -0.01285553, -0.007675171, 0.0377502...
http://arxiv.org/abs/2305.16107v1
VioLA: Unified Codec Language Models for Speech Recognition, Synthesis, and Translation
Tianrui Wang, Long Zhou, Ziqiang Zhang, Yu Wu, Shujie Liu, Yashesh Gaur, Zhuo Chen, Jinyu Li, Furu Wei
cs.CL, cs.SD, eess.AS
Recent research shows a big convergence in model architecture, training objectives, and inference methods across various tasks for different modalities. In this paper, we propose VioLA, a single auto-regressive Transformer decoder-only network that unifies various cross-modal tasks involving speech and text, such as sp...
http://arxiv.org/pdf/2305.16107v1
2023-05-25T14:39:47
2023-05-25T14:39:47
{ "title": [ -0.04345703, -0.019805908, -0.043823242, -0.060668945, -0.0060539246, 0.008575439, -0.07244873, 0.01902771, 0.0004377365, 0.05178833, -0.01234436, 0.008636475, 0.02671814, -0.019958496, -0.044189453, 0.02645874, -0.012825012, 0.04241...
http://arxiv.org/abs/2305.16106v1
Multijugate Dual Learning for Low-Resource Task-Oriented Dialogue System
Shimin Li, Xiaotian Zhang, Yanjun Zheng, Linyang Li, Xipeng Qiu
cs.CL
Dialogue data in real scenarios tend to be sparsely available, rendering data-starved end-to-end dialogue systems trained inadequately. We discover that data utilization efficiency in low-resource scenarios can be enhanced by mining alignment information uncertain utterance and deterministic dialogue state. Therefore, ...
http://arxiv.org/pdf/2305.16106v1
2023-05-25T14:38:05
2023-05-25T14:38:05
{ "title": [ -0.016662598, -0.008804321, -0.043701172, -0.07171631, 0.0069885254, -0.0070266724, -0.030944824, 0.031036377, -0.036376953, 0.062347412, -0.012557983, 0.013534546, -0.023757935, -0.009254456, -0.047546387, -0.023223877, -0.014175415, ...
http://arxiv.org/abs/2305.16103v1
ChatBridge: Bridging Modalities with Large Language Model as a Language Catalyst
Zijia Zhao, Longteng Guo, Tongtian Yue, Sihan Chen, Shuai Shao, Xinxin Zhu, Zehuan Yuan, Jing Liu
cs.CV, cs.AI, cs.CL, cs.MM
Building general-purpose models that can perceive diverse real-world modalities and solve various tasks is an appealing target in artificial intelligence. In this paper, we present ChatBridge, a novel multimodal language model that leverages the expressive capabilities of language as the catalyst to bridge the gap betw...
http://arxiv.org/pdf/2305.16103v1
2023-05-25T14:34:08
2023-05-25T14:34:08
{ "title": [ -0.028549194, -0.016082764, -0.0579834, -0.08355713, -0.0057296753, 0.014419556, -0.07647705, 0.008666992, -0.009437561, 0.03363037, 0.0053977966, -0.0051231384, 0.00014734268, -0.032043457, -0.047424316, -0.0021438599, -0.021820068, ...
http://arxiv.org/abs/2305.16093v2
End-to-End Simultaneous Speech Translation with Differentiable Segmentation
Shaolei Zhang, Yang Feng
cs.CL, cs.AI, cs.SD, eess.AS
End-to-end simultaneous speech translation (SimulST) outputs translation while receiving the streaming speech inputs (a.k.a. streaming speech translation), and hence needs to segment the speech inputs and then translate based on the current received speech. However, segmenting the speech inputs at unfavorable moments c...
http://arxiv.org/pdf/2305.16093v2
2023-06-18T03:18:12
2023-05-25T14:25:12
{ "title": [ -0.030090332, -0.004047394, -0.021408081, -0.07751465, 0.010139465, 0.009979248, -0.04107666, 0.017822266, -0.030136108, 0.04171753, 0.0052337646, 0.017486572, 0.009223938, -0.025970459, -0.01626587, 0.013000488, -0.0017051697, 0.022...
http://arxiv.org/abs/2305.16065v2
ASR and Emotional Speech: A Word-Level Investigation of the Mutual Impact of Speech and Emotion Recognition
Yuanchao Li, Zeyu Zhao, Ondrej Klejch, Peter Bell, Catherine Lai
eess.AS, cs.CL, cs.SD
In Speech Emotion Recognition (SER), textual data is often used alongside audio signals to address their inherent variability. However, the reliance on human annotated text in most research hinders the development of practical SER systems. To overcome this challenge, we investigate how Automatic Speech Recognition (ASR...
http://arxiv.org/pdf/2305.16065v2
2023-05-28T17:26:58
2023-05-25T13:56:09
{ "title": [ -0.0070266724, 0.010658264, -0.048828125, -0.04977417, 0.0033493042, -0.026275635, -0.030593872, 0.008262634, -0.017364502, 0.035705566, -0.017349243, -0.0039024353, 0.019927979, -0.029266357, -0.029830933, 0.0012636185, 0.009109497, ...
http://arxiv.org/abs/2305.16057v1
Fake News Detection and Behavioral Analysis: Case of COVID-19
Chih-Yuan Li, Navya Martin Kollapally, Soon Ae Chun, James Geller
cs.LG, cs.CL, 68
While the world has been combating COVID-19 for over three years, an ongoing "Infodemic" due to the spread of fake news regarding the pandemic has also been a global issue. The existence of the fake news impact different aspect of our daily lives, including politics, public health, economic activities, etc. Readers cou...
http://arxiv.org/pdf/2305.16057v1
2023-05-25T13:42:08
2023-05-25T13:42:08
{ "title": [ 0.009864807, 0.012481689, -0.04977417, -0.045318604, 0.01574707, -0.02229309, -0.0051612854, -0.012428284, -0.05307007, 0.025360107, -0.014564514, 0.016677856, 0.003501892, -0.041809082, 0.008079529, -0.02003479, -0.012825012, 0.0515...
http://arxiv.org/abs/2305.16051v1
What about em? How Commercial Machine Translation Fails to Handle (Neo-)Pronouns
Anne Lauscher, Debora Nozza, Archie Crowley, Ehm Miltersen, Dirk Hovy
cs.CL, cs.AI, cs.CY
As 3rd-person pronoun usage shifts to include novel forms, e.g., neopronouns, we need more research on identity-inclusive NLP. Exclusion is particularly harmful in one of the most popular NLP applications, machine translation (MT). Wrong pronoun translations can discriminate against marginalized groups, e.g., non-binar...
http://arxiv.org/pdf/2305.16051v1
2023-05-25T13:34:09
2023-05-25T13:34:09
{ "title": [ -0.028564453, -0.011436462, -0.00712204, -0.017974854, -0.022918701, 0.0059318542, -0.011512756, -0.022781372, -0.0028839111, 0.045928955, -0.013343811, -0.005382538, 0.067871094, -0.03286743, -0.029922485, 0.013801575, -0.015312195, ...
http://arxiv.org/abs/2305.16048v1
UFO: Unified Fact Obtaining for Commonsense Question Answering
Zhifeng Li, Yifan Fan, Bowei Zou, Yu Hong
cs.CL, cs.AI
Leveraging external knowledge to enhance the reasoning ability is crucial for commonsense question answering. However, the existing knowledge bases heavily rely on manual annotation which unavoidably causes deficiency in coverage of world-wide commonsense knowledge. Accordingly, the knowledge bases fail to be flexible ...
http://arxiv.org/pdf/2305.16048v1
2023-05-25T13:25:49
2023-05-25T13:25:49
{ "title": [ 0.009002686, -0.03479004, -0.026245117, -0.08074951, -0.0051994324, -0.0048179626, -0.019500732, 0.011116028, 0.013725281, 0.030075073, -0.015419006, 0.0126953125, -0.007091522, -0.008987427, -0.047027588, 0.009048462, -0.013031006, ...
http://arxiv.org/abs/2305.16031v1
Efficient Document Embeddings via Self-Contrastive Bregman Divergence Learning
Daniel Saggau, Mina Rezaei, Bernd Bisch, Ilias Chalkidis
cs.CL
Learning quality document embeddings is a fundamental problem in natural language processing (NLP), information retrieval (IR), recommendation systems, and search engines. Despite recent advances in the development of transformer-based models that produce sentence embeddings with self-contrastive learning, the encoding...
http://arxiv.org/pdf/2305.16031v1
2023-05-25T13:08:10
2023-05-25T13:08:10
{ "title": [ 0.008201599, 0.015113831, -0.02178955, -0.044647217, 0.017074585, -0.0064086914, -0.0129776, -0.022323608, 0.01890564, 0.021469116, -0.03555298, 0.020370483, -0.018554688, -0.005001068, -0.04626465, -0.0044784546, -0.0035037994, 0.05...
http://arxiv.org/abs/2305.16371v1
INTapt: Information-Theoretic Adversarial Prompt Tuning for Enhanced Non-Native Speech Recognition
Eunseop Yoon, Hee Suk Yoon, John Harvill, Mark Hasegawa-Johnson, Chang D. Yoo
cs.CL, cs.SD, eess.AS
Automatic Speech Recognition (ASR) systems have attained unprecedented performance with large speech models pre-trained based on self-supervised speech representation learning. However, these pre-trained speech models suffer from representational bias as they tend to better represent those prominent accents (i.e., nati...
http://arxiv.org/pdf/2305.16371v1
2023-05-25T13:06:01
2023-05-25T13:06:01
{ "title": [ -0.014175415, -0.008224487, -0.023284912, -0.053253174, -0.011680603, 0.0110321045, -0.056640625, 0.009140015, -0.03375244, 0.022659302, -0.018875122, 0.024368286, -0.007751465, -0.022247314, -0.017181396, -0.010269165, 0.0033454895, ...
http://arxiv.org/abs/2305.16023v1
NaSGEC: a Multi-Domain Chinese Grammatical Error Correction Dataset from Native Speaker Texts
Yue Zhang, Bo Zhang, Haochen Jiang, Zhenghua Li, Chen Li, Fei Huang, Min Zhang
cs.CL
We introduce NaSGEC, a new dataset to facilitate research on Chinese grammatical error correction (CGEC) for native speaker texts from multiple domains. Previous CGEC research primarily focuses on correcting texts from a single domain, especially learner essays. To broaden the target domain, we annotate multiple refere...
http://arxiv.org/pdf/2305.16023v1
2023-05-25T13:05:52
2023-05-25T13:05:52
{ "title": [ -0.007865906, 0.013908386, -0.017807007, -0.040985107, 0.0043029785, -0.00043582916, -0.03237915, -0.0107040405, 0.011009216, 0.023757935, -0.027191162, 0.009643555, 0.021270752, -0.017440796, -0.041015625, 0.0010766983, -0.041046143, ...
http://arxiv.org/abs/2305.16000v1
Do You Hear The People Sing? Key Point Analysis via Iterative Clustering and Abstractive Summarisation
Hao Li, Viktor Schlegel, Riza Batista-Navarro, Goran Nenadic
cs.CL, cs.AI
Argument summarisation is a promising but currently under-explored field. Recent work has aimed to provide textual summaries in the form of concise and salient short texts, i.e., key points (KPs), in a task known as Key Point Analysis (KPA). One of the main challenges in KPA is finding high-quality key point candidates...
http://arxiv.org/pdf/2305.16000v1
2023-05-25T12:43:29
2023-05-25T12:43:29
{ "title": [ -0.0076026917, 0.0104904175, -0.058502197, -0.044921875, 0.018371582, -0.019454956, -0.048461914, -0.041534424, 0.03451538, 0.026123047, -0.013969421, 0.029037476, -0.018600464, -0.013389587, -0.051818848, 0.030639648, -0.021606445, ...
http://arxiv.org/abs/2305.16367v1
Role-Play with Large Language Models
Murray Shanahan, Kyle McDonell, Laria Reynolds
cs.CL, cs.AI, cs.LG
As dialogue agents become increasingly human-like in their performance, it is imperative that we develop effective ways to describe their behaviour in high-level terms without falling into the trap of anthropomorphism. In this paper, we foreground the concept of role-play. Casting dialogue agent behaviour in terms of r...
http://arxiv.org/pdf/2305.16367v1
2023-05-25T11:36:52
2023-05-25T11:36:52
{ "title": [ -0.012702942, -0.018722534, -0.033233643, -0.070373535, -0.006248474, 0.00409317, -0.058258057, 0.011001587, -0.025558472, 0.023605347, -0.0053482056, -0.017715454, -0.007598877, -0.03286743, -0.012893677, -0.004371643, -0.011528015, ...
http://arxiv.org/abs/2305.16366v1
Decomposing the Enigma: Subgoal-based Demonstration Learning for Formal Theorem Proving
Xueliang Zhao, Wenda Li, Lingpeng Kong
cs.CL, cs.AI, cs.LG, cs.LO
Large language models~(LLMs) present an intriguing avenue of exploration in the domain of formal theorem proving. Nonetheless, the full utilization of these models, particularly in terms of demonstration formatting and organization, remains an underexplored area. In an endeavor to enhance the efficacy of LLMs, we intro...
http://arxiv.org/pdf/2305.16366v1
2023-05-25T11:35:52
2023-05-25T11:35:52
{ "title": [ -0.0029716492, 0.01638794, -0.03656006, -0.06604004, 0.025650024, -0.013305664, -0.027740479, 0.0006117821, -0.0027656555, 0.009109497, -0.05630493, 0.016189575, -0.0018129349, -0.003408432, -0.021820068, -0.0262146, -0.018341064, 0....
http://arxiv.org/abs/2305.15937v1
Visually grounded few-shot word acquisition with fewer shots
Leanne Nortje, Benjamin van Niekerk, Herman Kamper
cs.CL, cs.AI, eess.AS
We propose a visually grounded speech model that acquires new words and their visual depictions from just a few word-image example pairs. Given a set of test images and a spoken query, we ask the model which image depicts the query word. Previous work has simplified this problem by either using an artificial setting wi...
http://arxiv.org/pdf/2305.15937v1
2023-05-25T11:05:54
2023-05-25T11:05:54
{ "title": [ 0.008575439, -0.01424408, -0.029953003, -0.006137848, -0.040771484, -0.04220581, 0.014656067, -0.0061035156, -0.002790451, 0.030792236, -0.060455322, 0.018051147, 0.011619568, -0.012756348, -0.026412964, -0.014816284, 0.0024490356, 0...
http://arxiv.org/abs/2305.15932v2
BUCA: A Binary Classification Approach to Unsupervised Commonsense Question Answering
Jie He, Simon Chi Lok U, Víctor Gutiérrez-Basulto, Jeff Z. Pan
cs.CL
Unsupervised commonsense reasoning (UCR) is becoming increasingly popular as the construction of commonsense reasoning datasets is expensive, and they are inevitably limited in their scope. A popular approach to UCR is to fine-tune language models with external knowledge (e.g., knowledge graphs), but this usually requi...
http://arxiv.org/pdf/2305.15932v2
2023-06-07T20:33:09
2023-05-25T10:59:47
{ "title": [ -0.013061523, -0.019714355, -0.02357483, -0.0496521, -0.002216339, 0.00031399727, -0.022888184, -0.00028657913, 0.041900635, 0.016983032, -0.022827148, -0.008071899, -0.018798828, -0.013877869, -0.06237793, -0.0019207001, -0.01802063, ...
http://arxiv.org/abs/2305.15929v2
Emergence of a phonological bias in ChatGPT
Juan Manuel Toro
cs.CL
Current large language models, such as OpenAI's ChatGPT, have captured the public's attention because how remarkable they are in the use of language. Here, I demonstrate that ChatGPT displays phonological biases that are a hallmark of human language processing. More concretely, just like humans, ChatGPT has a consonant...
http://arxiv.org/pdf/2305.15929v2
2023-05-27T09:19:54
2023-05-25T10:57:43
{ "title": [ -0.015159607, -0.008041382, -0.014907837, 0.0001283884, -0.03225708, -0.0065574646, -0.0013542175, -0.00422287, -0.01828003, 0.034301758, -0.030380249, 0.00687027, 0.013008118, -0.023422241, -0.0018949509, 0.0037994385, -0.03353882, ...
http://arxiv.org/abs/2305.15914v2
Reliable Detection and Quantification of Selective Forces in Language Change
Juan Guerrero Montero, Andres Karjus, Kenny Smith, Richard A. Blythe
cs.CL
Language change is a cultural evolutionary process in which variants of linguistic variables change in frequency through processes analogous to mutation, selection and genetic drift. In this work, we apply a recently-introduced method to corpus data to quantify the strength of selection in specific instances of histori...
http://arxiv.org/pdf/2305.15914v2
2023-08-21T12:51:10
2023-05-25T10:20:15
{ "title": [ -0.019943237, 0.0020885468, -0.04598999, -0.045806885, 0.00907135, -0.006099701, -0.02897644, -0.0051498413, -0.0098724365, 0.016067505, -0.004558563, 0.014961243, -0.0026283264, -0.030639648, -0.023956299, 0.009811401, -0.015731812, ...
http://arxiv.org/abs/2305.15913v2
MEMEX: Detecting Explanatory Evidence for Memes via Knowledge-Enriched Contextualization
Shivam Sharma, Ramaneswaran S, Udit Arora, Md. Shad Akhtar, Tanmoy Chakraborty
cs.CL, cs.CY, cs.MM
Memes are a powerful tool for communication over social media. Their affinity for evolving across politics, history, and sociocultural phenomena makes them an ideal communication vehicle. To comprehend the subtle message conveyed within a meme, one must understand the background that facilitates its holistic assimilati...
http://arxiv.org/pdf/2305.15913v2
2023-05-27T13:09:46
2023-05-25T10:19:35
{ "title": [ 0.031280518, 0.011238098, -0.02670288, -0.031219482, 0.0047340393, -0.026763916, 0.000688076, -0.04864502, 0.004524231, 0.036499023, -0.015853882, 0.03857422, 0.0048980713, -0.021209717, -0.015258789, 0.016052246, -0.020492554, 0.066...
http://arxiv.org/abs/2305.15908v1
Response Generation in Longitudinal Dialogues: Which Knowledge Representation Helps?
Seyed Mahed Mousavi, Simone Caldarella, Giuseppe Riccardi
cs.CL
Longitudinal Dialogues (LD) are the most challenging type of conversation for human-machine dialogue systems. LDs include the recollections of events, personal thoughts, and emotions specific to each individual in a sparse sequence of dialogue sessions. Dialogue systems designed for LDs should uniquely interact with th...
http://arxiv.org/pdf/2305.15908v1
2023-05-25T10:13:53
2023-05-25T10:13:53
{ "title": [ -0.007911682, -0.015808105, -0.054870605, -0.07727051, -0.0029716492, -0.011978149, -0.043884277, 0.03112793, 0.01663208, 0.014389038, -0.008575439, 0.012756348, -0.05834961, -0.02178955, -0.06542969, -0.023605347, 0.0050315857, 0.02...
http://arxiv.org/abs/2305.15904v1
MTCue: Learning Zero-Shot Control of Extra-Textual Attributes by Leveraging Unstructured Context in Neural Machine Translation
Sebastian Vincent, Robert Flynn, Carolina Scarton
cs.CL, cs.AI, cs.LG
Efficient utilisation of both intra- and extra-textual context remains one of the critical gaps between machine and human translation. Existing research has primarily focused on providing individual, well-defined types of context in translation, such as the surrounding text or discrete external variables like the speak...
http://arxiv.org/pdf/2305.15904v1
2023-05-25T10:06:08
2023-05-25T10:06:08
{ "title": [ -0.015281677, -0.012924194, -0.0069770813, -0.035003662, -0.018936157, 0.010765076, -0.018127441, -0.0062713623, -0.0043029785, 0.059265137, -0.0029315948, 0.020217896, 0.03604126, -0.013969421, -0.03074646, 0.023986816, -0.01121521, ...
http://arxiv.org/abs/2305.15895v1
Collective Knowledge Graph Completion with Mutual Knowledge Distillation
Weihang Zhang, Ovidiu Serban, Jiahao Sun, Yi-ke Guo
cs.CL
Knowledge graph completion (KGC), the task of predicting missing information based on the existing relational data inside a knowledge graph (KG), has drawn significant attention in recent years. However, the predictive power of KGC methods is often limited by the completeness of the existing knowledge graphs from diffe...
http://arxiv.org/pdf/2305.15895v1
2023-05-25T09:49:40
2023-05-25T09:49:40
{ "title": [ -0.030029297, -0.0020389557, -0.07727051, -0.10296631, 0.012710571, 0.015075684, -0.08178711, 0.009353638, 0.024627686, 0.028930664, -0.032592773, 0.0028762817, 0.0005121231, 0.0011129379, -0.046173096, 0.00011104345, 0.0024414062, 0...
http://arxiv.org/abs/2305.15894v1
Private Meeting Summarization Without Performance Loss
Seolhwa Lee, Anders Søgaard
cs.CL, cs.CR
Meeting summarization has an enormous business potential, but in addition to being a hard problem, roll-out is challenged by privacy concerns. We explore the problem of meeting summarization under differential privacy constraints and find, to our surprise, that while differential privacy leads to slightly lower perform...
http://arxiv.org/pdf/2305.15894v1
2023-05-25T09:48:50
2023-05-25T09:48:50
{ "title": [ -0.004863739, -0.0030593872, -0.044952393, -0.08972168, -0.011795044, -0.00042247772, -0.085510254, 0.0017604828, -0.0070114136, 0.054595947, -0.0027618408, 0.0071144104, 0.012672424, -0.0069007874, -0.032592773, -0.0008635521, -0.003076...
http://arxiv.org/abs/2305.15891v1
CSS: A Large-scale Cross-schema Chinese Text-to-SQL Medical Dataset
Hanchong Zhang, Jieyu Li, Lu Chen, Ruisheng Cao, Yunyan Zhang, Yu Huang, Yefeng Zheng, Kai Yu
cs.CL
The cross-domain text-to-SQL task aims to build a system that can parse user questions into SQL on complete unseen databases, and the single-domain text-to-SQL task evaluates the performance on identical databases. Both of these setups confront unavoidable difficulties in real-world applications. To this end, we introd...
http://arxiv.org/pdf/2305.15891v1
2023-05-25T09:44:44
2023-05-25T09:44:44
{ "title": [ -0.016479492, 0.003452301, -0.059295654, -0.07647705, 0.024993896, -0.010139465, -0.035583496, 0.011222839, 0.0065841675, 0.04324341, -0.030532837, 0.013252258, 0.018554688, -0.023712158, -0.024291992, -0.007434845, 0.013977051, 0.01...
http://arxiv.org/abs/2305.15878v2
LFTK: Handcrafted Features in Computational Linguistics
Bruce W. Lee, Jason Hyung-Jong Lee
cs.CL, cs.LG
Past research has identified a rich set of handcrafted linguistic features that can potentially assist various tasks. However, their extensive number makes it difficult to effectively select and utilize existing handcrafted features. Coupled with the problem of inconsistent implementation across research works, there h...
http://arxiv.org/pdf/2305.15878v2
2023-06-01T17:42:21
2023-05-25T09:20:27
{ "title": [ -0.024765015, -0.009147644, -0.010505676, -0.05419922, 0.0075302124, 0.024642944, -0.04598999, -0.012077332, 0.00034332275, 0.031204224, 0.0040359497, 0.0023384094, 0.011116028, -0.016601562, -0.040924072, 0.018707275, -0.018920898, ...
http://arxiv.org/abs/2305.15875v2
Linguistic Properties of Truthful Response
Bruce W. Lee, Benedict Florance Arockiaraj, Helen Jin
cs.CL, cs.AI
We investigate the phenomenon of an LLM's untruthful response using a large set of 220 handcrafted linguistic features. We focus on GPT-3 models and find that the linguistic profiles of responses are similar across model sizes. That is, how varying-sized LLMs respond to given prompts stays similar on the linguistic pro...
http://arxiv.org/pdf/2305.15875v2
2023-06-02T12:39:12
2023-05-25T09:17:39
{ "title": [ -0.006439209, -0.010795593, -0.029418945, -0.032287598, -0.047821045, -0.0234375, -0.033996582, -0.009681702, -0.03555298, 0.06530762, 0.026184082, -0.015701294, 0.0035171509, -0.042999268, -0.032440186, 0.010414124, -0.00087976456, ...
http://arxiv.org/abs/2305.15872v1
Jointprop: Joint Semi-supervised Learning for Entity and Relation Extraction with Heterogeneous Graph-based Propagation
Yandan Zheng, Anran Hao, Anh Tuan Luu
cs.CL, cs.AI
Semi-supervised learning has been an important approach to address challenges in extracting entities and relations from limited data. However, current semi-supervised works handle the two tasks (i.e., Named Entity Recognition and Relation Extraction) separately and ignore the cross-correlation of entity and relation in...
http://arxiv.org/pdf/2305.15872v1
2023-05-25T09:07:04
2023-05-25T09:07:04
{ "title": [ -0.01675415, 0.006629944, -0.068603516, -0.076416016, 0.027526855, -0.0069122314, -0.04827881, 0.0008292198, 0.022872925, 0.007762909, -0.007675171, 0.016708374, 0.014320374, -0.028900146, -0.0066833496, 0.016555786, -0.027389526, 0....
http://arxiv.org/abs/2305.15867v1
Extracting Text Representations for Terms and Phrases in Technical Domains
Francesco Fusco, Diego Antognini
cs.CL, cs.AI, cs.LG
Extracting dense representations for terms and phrases is a task of great importance for knowledge discovery platforms targeting highly-technical fields. Dense representations are used as features for downstream components and have multiple applications ranging from ranking results in search to summarization. Common ap...
http://arxiv.org/pdf/2305.15867v1
2023-05-25T08:59:36
2023-05-25T08:59:36
{ "title": [ -0.0037193298, 0.010345459, -0.036895752, -0.039642334, 0.023086548, -0.012451172, -0.018127441, -0.017852783, 0.007801056, 0.044830322, -0.010375977, 0.017974854, 0.003578186, 0.0049934387, -0.057617188, 0.01272583, -0.01979065, 0.0...
http://arxiv.org/abs/2305.15853v1
Sequential Integrated Gradients: a simple but effective method for explaining language models
Joseph Enguehard
cs.CL, cs.AI, cs.LG
Several explanation methods such as Integrated Gradients (IG) can be characterised as path-based methods, as they rely on a straight line between the data and an uninformative baseline. However, when applied to language models, these methods produce a path for each word of a sentence simultaneously, which could lead to...
http://arxiv.org/pdf/2305.15853v1
2023-05-25T08:44:11
2023-05-25T08:44:11
{ "title": [ -0.01260376, -0.008125305, -0.019989014, -0.08062744, 0.0031223297, 0.016998291, -0.06713867, -0.020568848, 0.0005440712, 0.020187378, 0.00044870377, -0.0012292862, -0.0039138794, -0.034942627, -0.039276123, 0.018218994, -0.013175964, ...
http://arxiv.org/abs/2305.15852v2
Self-contradictory Hallucinations of Large Language Models: Evaluation, Detection and Mitigation
Niels Mündler, Jingxuan He, Slobodan Jenko, Martin Vechev
cs.CL, cs.AI, cs.LG
Large language models (large LMs) are susceptible to producing text that contains hallucinated content. An important instance of this problem is self-contradiction, where the LM generates two contradictory sentences within the same context. In this work, we present a comprehensive investigation into self-contradiction ...
http://arxiv.org/pdf/2305.15852v2
2023-10-01T07:22:39
2023-05-25T08:43:46
{ "title": [ -0.03125, 0.017852783, -0.04244995, -0.027664185, 0.011070251, -0.008872986, -0.029556274, -0.026550293, -0.0064086914, 0.03262329, -0.00522995, -0.0024375916, 0.004486084, -0.014526367, -0.056732178, 0.003545761, -0.020050049, 0.057...
http://arxiv.org/abs/2305.15814v3
Bhasha-Abhijnaanam: Native-script and romanized Language Identification for 22 Indic languages
Yash Madhani, Mitesh M. Khapra, Anoop Kunchukuttan
cs.CL
We create publicly available language identification (LID) datasets and models in all 22 Indian languages listed in the Indian constitution in both native-script and romanized text. First, we create Bhasha-Abhijnaanam, a language identification test set for native-script as well as romanized text which spans all 22 Ind...
http://arxiv.org/pdf/2305.15814v3
2023-10-26T05:57:27
2023-05-25T07:53:23
{ "title": [ -0.010673523, 0.018600464, -0.059936523, -0.024856567, -0.004383087, -0.03427124, -0.040649414, -0.011016846, -0.050720215, 0.021102905, -0.025604248, 0.020568848, -0.002210617, -0.04727173, 0.025115967, -0.031066895, 0.015258789, 0....
http://arxiv.org/abs/2305.15805v2
Dynamic Context Pruning for Efficient and Interpretable Autoregressive Transformers
Sotiris Anagnostidis, Dario Pavllo, Luca Biggio, Lorenzo Noci, Aurelien Lucchi, Thomas Hofmann
cs.CL, cs.LG
Autoregressive Transformers adopted in Large Language Models (LLMs) are hard to scale to long sequences. Despite several works trying to reduce their computational cost, most of LLMs still adopt attention layers between all pairs of tokens in the sequence, thus incurring a quadratic cost. In this study, we present a no...
http://arxiv.org/pdf/2305.15805v2
2023-05-28T12:11:11
2023-05-25T07:39:41
{ "title": [ 0.00969696, 0.010803223, -0.033599854, -0.10284424, 0.041137695, 0.019042969, -0.027694702, -0.003522873, 0.013015747, 0.031280518, -0.027664185, 0.011993408, -0.00024223328, 0.0010538101, -0.052520752, -0.002407074, 0.013397217, 0.0...
http://arxiv.org/abs/2305.15769v2
MERGE: Fast Private Text Generation
Zi Liang, Pinghui Wang, Ruofei Zhang, Lifeng Xing, Nuo Xu, Shuo Zhang
cs.CL, cs.AI
Recent years have seen increasing concerns about the private inference of NLP services and Transformer models. However, existing two-party privacy-preserving methods solely consider NLU scenarios, while the private inference of text generation such as translation, dialogue, and code completion remains unsolved. Besides...
http://arxiv.org/pdf/2305.15769v2
2023-07-02T11:56:55
2023-05-25T06:27:19
{ "title": [ -0.038879395, -0.025741577, -0.05444336, -0.07727051, -0.015266418, 0.018325806, -0.07293701, 0.014839172, -0.028549194, 0.02355957, -0.017471313, 0.015579224, 0.01071167, -0.020141602, -0.01687622, -0.010032654, -0.016815186, 0.0577...
http://arxiv.org/abs/2305.16357v1
EDM3: Event Detection as Multi-task Text Generation
Ujjwala Anantheswaran, Himanshu Gupta, Mihir Parmar, Kuntal Kumar Pal, Chitta Baral
cs.CL
Event detection refers to identifying event occurrences in a text and comprises of two subtasks; event identification and classification. We present EDM3, a novel approach for Event Detection that formulates three generative tasks: identification, classification, and combined detection. We show that EDM3 helps to learn...
http://arxiv.org/pdf/2305.16357v1
2023-05-25T06:25:16
2023-05-25T06:25:16
{ "title": [ -0.010475159, -0.0029144287, -0.039794922, -0.06896973, 0.008888245, 0.0031051636, -0.06567383, -0.025115967, 0.013031006, 0.039215088, 0.01008606, 0.019973755, 0.0028915405, -0.015472412, -0.046661377, 0.027023315, -0.01928711, 0.03...
http://arxiv.org/abs/2305.15760v1
Svarah: Evaluating English ASR Systems on Indian Accents
Tahir Javed, Sakshi Joshi, Vignesh Nagarajan, Sai Sundaresan, Janki Nawale, Abhigyan Raman, Kaushal Bhogale, Pratyush Kumar, Mitesh M. Khapra
cs.CL, cs.SD, eess.AS
India is the second largest English-speaking country in the world with a speaker base of roughly 130 million. Thus, it is imperative that automatic speech recognition (ASR) systems for English should be evaluated on Indian accents. Unfortunately, Indian speakers find a very poor representation in existing English ASR b...
http://arxiv.org/pdf/2305.15760v1
2023-05-25T06:20:29
2023-05-25T06:20:29
{ "title": [ -0.004508972, 0.00390625, -0.04537964, -0.05105591, 0.01727295, -0.014183044, -0.019927979, 0.0062065125, -0.03152466, 0.026977539, -0.020233154, 0.00349617, 0.007545471, -0.035003662, -0.03086853, -0.0065307617, -0.00894165, 0.04589...
http://arxiv.org/abs/2305.15757v1
Healing Unsafe Dialogue Responses with Weak Supervision Signals
Zi Liang, Pinghui Wang, Ruofei Zhang, Shuo Zhang, Xiaofan Ye Yi Huang, Junlan Feng
cs.CL, cs.AI
Recent years have seen increasing concerns about the unsafe response generation of large-scale dialogue systems, where agents will learn offensive or biased behaviors from the real-world corpus. Some methods are proposed to address the above issue by detecting and replacing unsafe training examples in a pipeline style....
http://arxiv.org/pdf/2305.15757v1
2023-05-25T06:15:53
2023-05-25T06:15:53
{ "title": [ 0.019317627, 0.0034313202, -0.017364502, -0.06628418, -0.0038280487, -0.0052871704, -0.019851685, 0.003206253, -0.017578125, 0.029678345, -0.011932373, 0.004383087, 0.012184143, -0.029205322, -0.013626099, 0.0017910004, -0.00027656555, ...
http://arxiv.org/abs/2305.15756v1
UniTRec: A Unified Text-to-Text Transformer and Joint Contrastive Learning Framework for Text-based Recommendation
Zhiming Mao, Huimin Wang, Yiming Du, Kam-fai Wong
cs.CL
Prior study has shown that pretrained language models (PLM) can boost the performance of text-based recommendation. In contrast to previous works that either use PLM to encode user history as a whole input text, or impose an additional aggregation network to fuse multi-turn history representations, we propose a unified...
http://arxiv.org/pdf/2305.15756v1
2023-05-25T06:11:31
2023-05-25T06:11:31
{ "title": [ 0.013725281, -0.000080645084, -0.008781433, -0.044128418, 0.007499695, -0.0005207062, -0.0054016113, -0.021469116, 0.0034313202, 0.03564453, -0.04309082, 0.058380127, 0.014137268, -0.004398346, -0.019699097, 0.0012569427, -0.028869629, ...
http://arxiv.org/abs/2305.15749v1
Multilingual Text-to-Speech Synthesis for Turkic Languages Using Transliteration
Rustem Yeshpanov, Saida Mussakhojayeva, Yerbolat Khassanov
eess.AS, cs.CL
This work aims to build a multilingual text-to-speech (TTS) synthesis system for ten lower-resourced Turkic languages: Azerbaijani, Bashkir, Kazakh, Kyrgyz, Sakha, Tatar, Turkish, Turkmen, Uyghur, and Uzbek. We specifically target the zero-shot learning scenario, where a TTS model trained using the data of one language...
http://arxiv.org/pdf/2305.15749v1
2023-05-25T05:57:54
2023-05-25T05:57:54
{ "title": [ 0.0004646778, 0.014419556, -0.03479004, -0.038482666, 0.0048980713, -0.00712204, -0.045440674, 0.0020370483, -0.022277832, 0.07141113, -0.015945435, -0.008300781, 0.06750488, -0.03062439, -0.028686523, 0.0069732666, -0.0107803345, 0....
http://arxiv.org/abs/2305.15725v1
Learn to Not Link: Exploring NIL Prediction in Entity Linking
Fangwei Zhu, Jifan Yu, Hailong Jin, Juanzi Li, Lei Hou, Zhifang Sui
cs.CL
Entity linking models have achieved significant success via utilizing pretrained language models to capture semantic features. However, the NIL prediction problem, which aims to identify mentions without a corresponding entity in the knowledge base, has received insufficient attention. We categorize mentions linking to...
http://arxiv.org/pdf/2305.15725v1
2023-05-25T05:12:33
2023-05-25T05:12:33
{ "title": [ 0.0040397644, 0.00680542, -0.034729004, -0.08782959, 0.030532837, 0.00086307526, -0.02784729, -0.0037555695, -0.010101318, 0.025726318, -0.0025424957, 0.0024642944, 0.025985718, -0.026504517, -0.00793457, 0.0027389526, -0.026367188, ...
http://arxiv.org/abs/2305.15722v2
Comparative Study of Pre-Trained BERT Models for Code-Mixed Hindi-English Data
Aryan Patil, Varad Patwardhan, Abhishek Phaltankar, Gauri Takawane, Raviraj Joshi
cs.CL, cs.LG
The term "Code Mixed" refers to the use of more than one language in the same text. This phenomenon is predominantly observed on social media platforms, with an increasing amount of adaptation as time goes on. It is critical to detect foreign elements in a language and process them correctly, as a considerable number o...
http://arxiv.org/pdf/2305.15722v2
2023-05-26T06:51:22
2023-05-25T05:10:28
{ "title": [ -0.013130188, -0.022399902, -0.0146484375, -0.05899048, -0.024551392, 0.008575439, -0.06732178, 0.019042969, -0.017913818, 0.038604736, -0.0107421875, 0.003709793, 0.01550293, -0.0395813, -0.025177002, 0.0039711, -0.019500732, 0.0381...
http://arxiv.org/abs/2305.15718v1
Towards Higher Pareto Frontier in Multilingual Machine Translation
Yichong Huang, Xiaocheng Feng, Xinwei Geng, Baohang Li, Bing Qin
cs.CL
Multilingual neural machine translation has witnessed remarkable progress in recent years. However, the long-tailed distribution of multilingual corpora poses a challenge of Pareto optimization, i.e., optimizing for some languages may come at the cost of degrading the performance of others. Existing balancing training ...
http://arxiv.org/pdf/2305.15718v1
2023-05-25T05:01:33
2023-05-25T05:01:33
{ "title": [ -0.028823853, -0.014625549, 0.001707077, -0.061065674, -0.00705719, 0.017425537, -0.053588867, 0.0003900528, -0.0020828247, 0.07122803, 0.0077209473, 0.010322571, 0.04437256, -0.013809204, -0.058380127, 0.029708862, -0.02619934, 0.04...
http://arxiv.org/abs/2305.15717v1
The False Promise of Imitating Proprietary LLMs
Arnav Gudibande, Eric Wallace, Charlie Snell, Xinyang Geng, Hao Liu, Pieter Abbeel, Sergey Levine, Dawn Song
cs.CL
An emerging method to cheaply improve a weaker language model is to finetune it on outputs from a stronger model, such as a proprietary system like ChatGPT (e.g., Alpaca, Self-Instruct, and others). This approach looks to cheaply imitate the proprietary model's capabilities using a weaker open-source model. In this wor...
http://arxiv.org/pdf/2305.15717v1
2023-05-25T05:00:12
2023-05-25T05:00:12
{ "title": [ -0.028396606, -0.026412964, -0.00042510033, 0.018585205, -0.07800293, 0.0035057068, 0.005077362, -0.00077295303, -0.07940674, 0.042816162, -0.017623901, 0.006916046, -0.019317627, -0.057159424, 0.02470398, -0.039978027, -0.009269714, ...
http://arxiv.org/abs/2305.16355v1
PandaGPT: One Model To Instruction-Follow Them All
Yixuan Su, Tian Lan, Huayang Li, Jialu Xu, Yan Wang, Deng Cai
cs.CL, cs.CV
We present PandaGPT, an approach to emPower large lANguage moDels with visual and Auditory instruction-following capabilities. Our pilot experiments show that PandaGPT can perform complex tasks such as detailed image description generation, writing stories inspired by videos, and answering questions about audios. More ...
http://arxiv.org/pdf/2305.16355v1
2023-05-25T04:16:07
2023-05-25T04:16:07
{ "title": [ 0.007820129, -0.0077819824, -0.03805542, -0.047454834, -0.016448975, -0.02178955, -0.053100586, -0.027450562, -0.026916504, 0.01889038, 0.012710571, -0.003921509, 0.007865906, -0.060699463, 0.0076179504, 0.010429382, -0.015975952, 0....
http://arxiv.org/abs/2305.15689v2
Zero-shot Approach to Overcome Perturbation Sensitivity of Prompts
Mohna Chakraborty, Adithya Kulkarni, Qi Li
cs.CL, cs.AI
Recent studies have demonstrated that natural-language prompts can help to leverage the knowledge learned by pre-trained language models for the binary sentence-level sentiment classification task. Specifically, these methods utilize few-shot learning settings to fine-tune the sentiment classification model using manua...
http://arxiv.org/pdf/2305.15689v2
2023-07-01T23:02:18
2023-05-25T03:36:43
{ "title": [ 0.060028076, 0.010475159, -0.0031108856, -0.021026611, -0.01499939, -0.04284668, 0.013809204, -0.03515625, 0.027420044, -0.0020828247, -0.018478394, -0.028137207, -0.004558563, -0.06506348, -0.005432129, -0.010421753, 0.025024414, 0....
http://arxiv.org/abs/2305.15685v1
RewriteLM: An Instruction-Tuned Large Language Model for Text Rewriting
Lei Shu, Liangchen Luo, Jayakumar Hoskere, Yun Zhu, Canoee Liu, Simon Tong, Jindong Chen, Lei Meng
cs.CL, cs.AI
Large Language Models (LLMs) have demonstrated impressive zero-shot capabilities in long-form text generation tasks expressed through natural language instructions. However, user expectations for long-form text rewriting is high, and unintended rewrites (''hallucinations'') produced by the model can negatively impact i...
http://arxiv.org/pdf/2305.15685v1
2023-05-25T03:26:26
2023-05-25T03:26:26
{ "title": [ -0.055786133, -0.04107666, -0.038513184, -0.09033203, -0.018966675, 0.03488159, -0.093444824, 0.029815674, -0.011703491, 0.028030396, -0.019363403, 0.012763977, 0.00856781, -0.008895874, -0.0335083, 0.00017344952, -0.022201538, 0.037...
http://arxiv.org/abs/2305.15684v1
Perturbation-based Self-supervised Attention for Attention Bias in Text Classification
Huawen Feng, Zhenxi Lin, Qianli Ma
cs.CL
In text classification, the traditional attention mechanisms usually focus too much on frequent words, and need extensive labeled data in order to learn. This paper proposes a perturbation-based self-supervised attention approach to guide attention learning without any annotation overhead. Specifically, we add as much ...
http://arxiv.org/pdf/2305.15684v1
2023-05-25T03:18:18
2023-05-25T03:18:18
{ "title": [ 0.007797241, 0.02456665, -0.024169922, -0.0287323, 0.014091492, -0.026107788, 0.0014600754, -0.025985718, 0.023651123, 0.0184021, -0.047454834, 0.0067253113, -0.002828598, -0.0068206787, -0.03237915, -0.013633728, -0.011528015, 0.051...
http://arxiv.org/abs/2305.15678v1
Revisiting non-English Text Simplification: A Unified Multilingual Benchmark
Michael J. Ryan, Tarek Naous, Wei Xu
cs.CL, cs.AI
Recent advancements in high-quality, large-scale English resources have pushed the frontier of English Automatic Text Simplification (ATS) research. However, less work has been done on multilingual text simplification due to the lack of a diverse evaluation benchmark that covers complex-simple sentence pairs in many la...
http://arxiv.org/pdf/2305.15678v1
2023-05-25T03:03:29
2023-05-25T03:03:29
{ "title": [ -0.051239014, -0.014549255, -0.03201294, -0.06933594, -0.0030498505, 0.013168335, -0.10076904, -0.000060260296, -0.008407593, 0.06842041, 0.013847351, 0.02709961, 0.02532959, -0.009407043, -0.053771973, 0.019699097, -0.0045547485, 0....
http://arxiv.org/abs/2305.15676v2
Enhancing Grammatical Error Correction Systems with Explanations
Yuejiao Fei, Leyang Cui, Sen Yang, Wai Lam, Zhenzhong Lan, Shuming Shi
cs.CL
Grammatical error correction systems improve written communication by detecting and correcting language mistakes. To help language learners better understand why the GEC system makes a certain correction, the causes of errors (evidence words) and the corresponding error types are two key factors. To enhance GEC systems...
http://arxiv.org/pdf/2305.15676v2
2023-06-10T05:41:34
2023-05-25T03:00:49
{ "title": [ 0.00843811, -0.002632141, 0.0008096695, -0.07232666, -0.0063667297, 0.015319824, -0.033233643, -0.0038394928, 0.0026874542, 0.020095825, -0.03250122, -0.0038070679, -0.0009484291, -0.010139465, -0.040130615, -0.008094788, -0.03652954, ...
http://arxiv.org/abs/2306.09194v1
Undetectable Watermarks for Language Models
Miranda Christ, Sam Gunn, Or Zamir
cs.CR, cs.CL, cs.LG
Recent advances in the capabilities of large language models such as GPT-4 have spurred increasing concern about our ability to detect AI-generated text. Prior works have suggested methods of embedding watermarks in model outputs, by noticeably altering the output distribution. We ask: Is it possible to introduce a wat...
http://arxiv.org/pdf/2306.09194v1
2023-05-25T02:57:16
2023-05-25T02:57:16
{ "title": [ -0.0077209473, 0.011177063, -0.031143188, -0.06488037, 0.02482605, -0.0023498535, -0.05126953, -0.0181427, -0.006374359, 0.04095459, -0.00038909912, -0.003665924, 0.0102005005, -0.019821167, -0.045196533, 0.017242432, -0.015457153, 0...
http://arxiv.org/abs/2305.16353v1
Betray Oneself: A Novel Audio DeepFake Detection Model via Mono-to-Stereo Conversion
Rui Liu, Jinhua Zhang, Guanglai Gao, Haizhou Li
cs.SD, cs.AI, cs.CL
Audio Deepfake Detection (ADD) aims to detect the fake audio generated by text-to-speech (TTS), voice conversion (VC) and replay, etc., which is an emerging topic. Traditionally we take the mono signal as input and focus on robust feature extraction and effective classifier design. However, the dual-channel stereo info...
http://arxiv.org/pdf/2305.16353v1
2023-05-25T02:54:29
2023-05-25T02:54:29
{ "title": [ 0.014350891, 0.039978027, -0.04837036, -0.039215088, 0.040527344, -0.032226562, 0.003282547, -0.022750854, -0.049865723, 0.047088623, 0.009635925, 0.018218994, 0.027893066, -0.031921387, -0.0090408325, 0.014328003, -0.03100586, 0.036...
http://arxiv.org/abs/2305.15673v1
BookGPT: A General Framework for Book Recommendation Empowered by Large Language Model
Aakas Zhiyuli, Yanfang Chen, Xuan Zhang, Xun Liang
cs.IR, cs.CL
With the continuous development and change exhibited by large language model (LLM) technology, represented by generative pretrained transformers (GPTs), many classic scenarios in various fields have re-emerged with new opportunities. This paper takes ChatGPT as the modeling object, incorporates LLM technology into the ...
http://arxiv.org/pdf/2305.15673v1
2023-05-25T02:45:22
2023-05-25T02:45:22
{ "title": [ -0.006881714, -0.0068244934, -0.049743652, -0.03527832, -0.0010290146, -0.019042969, -0.02482605, -0.016082764, -0.00058841705, 0.020339966, -0.0231781, 0.028701782, 0.0061035156, -0.012557983, -0.004901886, -0.0002667904, -0.011520386, ...
http://arxiv.org/abs/2305.15663v1
Mixture-of-Expert Conformer for Streaming Multilingual ASR
Ke Hu, Bo Li, Tara N. Sainath, Yu Zhang, Francoise Beaufays
cs.CL, cs.SD, eess.AS
End-to-end models with large capacity have significantly improved multilingual automatic speech recognition, but their computation cost poses challenges for on-device applications. We propose a streaming truly multilingual Conformer incorporating mixture-of-expert (MoE) layers that learn to only activate a subset of pa...
http://arxiv.org/pdf/2305.15663v1
2023-05-25T02:16:32
2023-05-25T02:16:32
{ "title": [ -0.008964539, 0.0052490234, -0.059906006, -0.11682129, 0.009979248, 0.008392334, -0.079833984, 0.04638672, -0.024642944, 0.06365967, 0.0138549805, -0.011802673, 0.037719727, -0.030883789, -0.031280518, 0.014381409, 0.009109497, 0.050...
http://arxiv.org/abs/2305.15645v2
ConvGQR: Generative Query Reformulation for Conversational Search
Fengran Mo, Kelong Mao, Yutao Zhu, Yihong Wu, Kaiyu Huang, Jian-Yun Nie
cs.IR, cs.CL
In conversational search, the user's real search intent for the current turn is dependent on the previous conversation history. It is challenging to determine a good search query from the whole conversation context. To avoid the expensive re-training of the query encoder, most existing methods try to learn a rewriting ...
http://arxiv.org/pdf/2305.15645v2
2023-05-26T02:32:18
2023-05-25T01:45:06
{ "title": [ 0.0012102127, -0.015777588, -0.02633667, -0.059631348, 0.0000795722, 0.009429932, -0.049743652, -0.007881165, 0.037231445, 0.039031982, -0.021774292, 0.0072021484, 0.0045928955, -0.0067329407, -0.06762695, 0.0065841675, -0.016494751, ...
http://arxiv.org/abs/2305.15637v1
Morphological Inflection: A Reality Check
Jordan Kodner, Sarah Payne, Salam Khalifa, Zoey Liu
cs.CL
Morphological inflection is a popular task in sub-word NLP with both practical and cognitive applications. For years now, state-of-the-art systems have reported high, but also highly variable, performance across data sets and languages. We investigate the causes of this high performance and high variability; we find se...
http://arxiv.org/pdf/2305.15637v1
2023-05-25T01:27:29
2023-05-25T01:27:29
{ "title": [ -0.029663086, 0.00013744831, -0.035125732, -0.039276123, -0.008003235, -0.008132935, -0.032196045, -0.0017690659, -0.02822876, 0.039398193, -0.01272583, 0.00013399124, 0.005756378, -0.03567505, -0.024368286, 0.003917694, -0.020843506, ...
http://arxiv.org/abs/2305.15605v1
Revisiting Sentence Union Generation as a Testbed for Text Consolidation
Eran Hirsch, Valentina Pyatkin, Ruben Wolhandler, Avi Caciularu, Asi Shefer, Ido Dagan
cs.CL
Tasks involving text generation based on multiple input texts, such as multi-document summarization, long-form question answering and contemporary dialogue applications, challenge models for their ability to properly consolidate partly-overlapping multi-text information. However, these tasks entangle the consolidation ...
http://arxiv.org/pdf/2305.15605v1
2023-05-24T22:34:01
2023-05-24T22:34:01
{ "title": [ -0.029266357, -0.008781433, -0.028900146, -0.057617188, -0.000313282, 0.022125244, -0.04840088, 0.0096588135, -0.010444641, 0.056030273, -0.0049057007, 0.022445679, 0.024230957, -0.00016748905, -0.05517578, 0.0039482117, -0.03878784, ...
http://arxiv.org/abs/2305.15597v1
Text-Augmented Open Knowledge Graph Completion via Pre-Trained Language Models
Pengcheng Jiang, Shivam Agarwal, Bowen Jin, Xuan Wang, Jimeng Sun, Jiawei Han
cs.CL, cs.AI, cs.IR
The mission of open knowledge graph (KG) completion is to draw new findings from known facts. Existing works that augment KG completion require either (1) factual triples to enlarge the graph reasoning space or (2) manually designed prompts to extract knowledge from a pre-trained language model (PLM), exhibiting limite...
http://arxiv.org/pdf/2305.15597v1
2023-05-24T22:09:35
2023-05-24T22:09:35
{ "title": [ -0.022964478, -0.012016296, -0.042297363, -0.10571289, -0.013282776, 0.028808594, -0.09918213, 0.0151901245, -0.014091492, 0.032073975, -0.018859863, 0.03189087, -0.004211426, -0.0043182373, -0.030151367, -0.00030064583, -0.0039749146, ...
http://arxiv.org/abs/2305.15594v1
Flocks of Stochastic Parrots: Differentially Private Prompt Learning for Large Language Models
Haonan Duan, Adam Dziedzic, Nicolas Papernot, Franziska Boenisch
cs.LG, cs.CL, cs.CR
Large language models (LLMs) are excellent in-context learners. However, the sensitivity of data contained in prompts raises privacy concerns. Our work first shows that these concerns are valid: we instantiate a simple but highly effective membership inference attack against the data used to prompt LLMs. To address thi...
http://arxiv.org/pdf/2305.15594v1
2023-05-24T22:06:08
2023-05-24T22:06:08
{ "title": [ -0.027053833, -0.01197052, -0.040924072, -0.048583984, -0.0017747879, -0.0002784729, -0.05718994, 0.0071525574, -0.006095886, -0.008392334, -0.06695557, 0.02619934, -0.018661499, -0.009033203, 0.005332947, -0.023895264, -0.0019073486, ...
http://arxiv.org/abs/2305.15587v1
How do humans perceive adversarial text? A reality check on the validity and naturalness of word-based adversarial attacks
Salijona Dyrmishi, Salah Ghamizi, Maxime Cordy
cs.CL, cs.AI
Natural Language Processing (NLP) models based on Machine Learning (ML) are susceptible to adversarial attacks -- malicious algorithms that imperceptibly modify input text to force models into making incorrect predictions. However, evaluations of these attacks ignore the property of imperceptibility or study it under l...
http://arxiv.org/pdf/2305.15587v1
2023-05-24T21:52:13
2023-05-24T21:52:13
{ "title": [ -0.0140686035, 0.014030457, -0.022125244, -0.0037994385, 0.011703491, -0.021865845, 0.005580902, -0.037200928, -0.058929443, 0.03652954, -0.0014286041, 0.029296875, -0.0055770874, -0.01121521, -0.03253174, -0.016677856, -0.0063591003, ...
http://arxiv.org/abs/2305.15582v1
Balancing Effect of Training Dataset Distribution of Multiple Styles for Multi-Style Text Transfer
Debarati Das, David Ma, Dongyeop Kang
cs.CL
Text style transfer is an exciting task within the field of natural language generation that is often plagued by the need for high-quality paired datasets. Furthermore, training a model for multi-attribute text style transfer requires datasets with sufficient support across all combinations of the considered stylistic ...
http://arxiv.org/pdf/2305.15582v1
2023-05-24T21:36:15
2023-05-24T21:36:15
{ "title": [ -0.027313232, 0.0178833, -0.07647705, -0.026229858, 0.00919342, -0.03567505, -0.029769897, -0.002357483, -0.009712219, 0.07080078, -0.011634827, 0.031555176, 0.050811768, -0.015914917, -0.022201538, 0.016799927, -0.002161026, 0.03341...
http://arxiv.org/abs/2305.15542v2
TOAST: Transfer Learning via Attention Steering
Baifeng Shi, Siyu Gai, Trevor Darrell, Xin Wang
cs.CV, cs.CL, cs.LG
Transfer learning involves adapting a pre-trained model to novel downstream tasks. However, we observe that current transfer learning methods often fail to focus on task-relevant features. In this work, we explore refocusing model attention for transfer learning. We introduce Top-Down Attention Steering (TOAST), a nove...
http://arxiv.org/pdf/2305.15542v2
2023-07-11T17:57:06
2023-05-24T20:03:04
{ "title": [ -0.0050086975, 0.005092621, -0.04046631, -0.041137695, 0.020126343, -0.013023376, -0.009841919, -0.026321411, -0.0026512146, 0.021209717, -0.014770508, 0.018295288, 0.0063476562, -0.008163452, -0.017532349, 0.0067825317, -0.006702423, ...
http://arxiv.org/abs/2305.15541v1
Harnessing the Power of Large Language Models for Natural Language to First-Order Logic Translation
Yuan Yang, Siheng Xiong, Ali Payani, Ehsan Shareghi, Faramarz Fekri
cs.CL, cs.AI
Translating natural language sentences to first-order logic (NL-FOL translation) is a longstanding challenge in the NLP and formal logic literature. This paper introduces LogicLLaMA, a LLaMA-7B model fine-tuned for NL-FOL translation using LoRA on a single GPU. LogicLLaMA is capable of directly translating natural lang...
http://arxiv.org/pdf/2305.15541v1
2023-05-24T19:59:51
2023-05-24T19:59:51
{ "title": [ -0.045928955, -0.019638062, -0.042663574, -0.072021484, -0.0058517456, 0.011116028, -0.08154297, 0.013267517, 0.0031719208, 0.03643799, -0.020263672, 0.008476257, 0.026275635, -0.008926392, -0.036712646, 0.020370483, -0.019760132, 0....
http://arxiv.org/abs/2305.15533v1
Automated Refugee Case Analysis: An NLP Pipeline for Supporting Legal Practitioners
Claire Barale, Michael Rovatsos, Nehal Bhuta
cs.CL
In this paper, we introduce an end-to-end pipeline for retrieving, processing, and extracting targeted information from legal cases. We investigate an under-studied legal domain with a case study on refugee law in Canada. Searching case law for past similar cases is a key part of legal work for both lawyers and judges,...
http://arxiv.org/pdf/2305.15533v1
2023-05-24T19:37:23
2023-05-24T19:37:23
{ "title": [ -0.027923584, -0.014526367, -0.0395813, -0.053741455, -0.019119263, 0.014656067, -0.051696777, -0.0064582825, 0.014190674, 0.027359009, -0.049102783, 0.006832123, 0.018859863, -0.008850098, -0.036376953, -0.0011510849, -0.01449585, 0...
http://arxiv.org/abs/2305.15525v1
Large Language Models are Few-Shot Health Learners
Xin Liu, Daniel McDuff, Geza Kovacs, Isaac Galatzer-Levy, Jacob Sunshine, Jiening Zhan, Ming-Zher Poh, Shun Liao, Paolo Di Achille, Shwetak Patel
cs.CL, cs.LG
Large language models (LLMs) can capture rich representations of concepts that are useful for real-world tasks. However, language alone is limited. While existing LLMs excel at text-based inferences, health applications require that models be grounded in numerical data (e.g., vital signs, laboratory values in clinical ...
http://arxiv.org/pdf/2305.15525v1
2023-05-24T19:25:16
2023-05-24T19:25:16
{ "title": [ 0.024459839, -0.013999939, -0.038269043, -0.06878662, -0.040405273, -0.02935791, -0.05206299, 0.02178955, -0.040771484, 0.015975952, -0.039154053, -0.0075683594, -0.024505615, -0.04434204, 0.0146102905, -0.04473877, 0.02758789, -0.00...
http://arxiv.org/abs/2305.15520v1
Exploring Automatically Perturbed Natural Language Explanations in Relation Extraction
Wanyun Cui, Xingran Chen
cs.CL
Previous research has demonstrated that natural language explanations provide valuable inductive biases that guide models, thereby improving the generalization ability and data efficiency. In this paper, we undertake a systematic examination of the effectiveness of these explanations. Remarkably, we find that corrupted...
http://arxiv.org/pdf/2305.15520v1
2023-05-24T19:17:13
2023-05-24T19:17:13
{ "title": [ -0.0011892319, -0.015472412, -0.020935059, -0.07385254, -0.01071167, 0.0066337585, -0.032806396, -0.003206253, 0.013389587, -0.0072135925, -0.034606934, 0.009529114, -0.018356323, -0.00945282, -0.025466919, -0.011566162, -0.019180298, ...
http://arxiv.org/abs/2305.15516v1
Free Lunch for Efficient Textual Commonsense Integration in Language Models
Wanyun Cui, Xingran Chen
cs.CL
Recent years have witnessed the emergence of textual commonsense knowledge bases, aimed at providing more nuanced and context-rich knowledge. The integration of external commonsense into language models has been shown to be a key enabler in advancing the state-of-the-art for a wide range of NLP tasks. However, incorpor...
http://arxiv.org/pdf/2305.15516v1
2023-05-24T19:14:57
2023-05-24T19:14:57
{ "title": [ 0.0023841858, -0.02330017, -0.014564514, -0.080322266, -0.0033435822, 0.01007843, -0.049743652, 0.0059661865, 0.035461426, 0.02545166, -0.040740967, 0.013427734, 0.015960693, -0.0020618439, -0.044891357, 0.015113831, -0.02696228, 0.0...
http://arxiv.org/abs/2305.16349v1
Lexinvariant Language Models
Qian Huang, Eric Zelikman, Sarah Li Chen, Yuhuai Wu, Gregory Valiant, Percy Liang
cs.CL, cs.AI, cs.LG
Token embeddings, a mapping from discrete lexical symbols to continuous vectors, are at the heart of any language model (LM). However, lexical symbol meanings can also be determined and even redefined by their structural role in a long context. In this paper, we ask: is it possible for a language model to be performant...
http://arxiv.org/pdf/2305.16349v1
2023-05-24T19:10:46
2023-05-24T19:10:46
{ "title": [ -0.027664185, -0.01826477, -0.03829956, -0.07165527, -0.052642822, 0.0071372986, -0.09851074, 0.025512695, 0.008148193, 0.055389404, -0.013832092, -0.019622803, 0.059387207, -0.046966553, -0.011421204, 0.035339355, -0.00017559528, 0....
http://arxiv.org/abs/2305.15507v1
The Larger They Are, the Harder They Fail: Language Models do not Recognize Identifier Swaps in Python
Antonio Valerio Miceli-Barone, Fazl Barez, Ioannis Konstas, Shay B. Cohen
cs.CL, cs.AI
Large Language Models (LLMs) have successfully been applied to code generation tasks, raising the question of how well these models understand programming. Typical programming languages have invariances and equivariances in their semantics that human programmers intuitively understand and exploit, such as the (near) in...
http://arxiv.org/pdf/2305.15507v1
2023-05-24T18:54:39
2023-05-24T18:54:39
{ "title": [ -0.03994751, 0.006137848, -0.042633057, 0.010299683, -0.015235901, -0.0287323, -0.018676758, -0.020904541, 0.001996994, 0.030593872, -0.038146973, -0.007411957, -0.00630188, -0.023208618, -0.05114746, -0.025665283, -0.017852783, 0.02...
http://arxiv.org/abs/2305.15501v1
Deriving Language Models from Masked Language Models
Lucas Torroba Hennigen, Yoon Kim
cs.CL
Masked language models (MLM) do not explicitly define a distribution over language, i.e., they are not language models per se. However, recent work has implicitly treated them as such for the purposes of generation and scoring. This paper studies methods for deriving explicit joint distributions from MLMs, focusing on ...
http://arxiv.org/pdf/2305.15501v1
2023-05-24T18:42:45
2023-05-24T18:42:45
{ "title": [ -0.05697632, -0.0098724365, -0.047576904, -0.066589355, -0.003660202, 0.021591187, -0.08862305, 0.011940002, -0.0064926147, 0.0211792, -0.0056152344, -0.018066406, 0.011817932, -0.032714844, -0.03387451, 0.010093689, -0.020751953, 0....
http://arxiv.org/abs/2305.15498v1
Large Language Models for User Interest Journeys
Konstantina Christakopoulou, Alberto Lalama, Cj Adams, Iris Qu, Yifat Amir, Samer Chucri, Pierce Vollucci, Fabio Soldo, Dina Bseiso, Sarah Scodel, Lucas Dixon, Ed H. Chi, Minmin Chen
cs.CL, cs.AI, cs.IR
Large language models (LLMs) have shown impressive capabilities in natural language understanding and generation. Their potential for deeper user understanding and improved personalized user experience on recommendation platforms is, however, largely untapped. This paper aims to address this gap. Recommender systems to...
http://arxiv.org/pdf/2305.15498v1
2023-05-24T18:40:43
2023-05-24T18:40:43
{ "title": [ -0.0035934448, -0.0072898865, -0.026229858, -0.051116943, 0.0028572083, -0.004714966, -0.031280518, -0.003894806, 0.015670776, 0.03894043, -0.007572174, -0.010375977, -0.008384705, -0.020446777, -0.04345703, 0.01876831, -0.014411926, ...
http://arxiv.org/abs/2305.15408v4
Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspective
Guhao Feng, Bohang Zhang, Yuntian Gu, Haotian Ye, Di He, Liwei Wang
cs.LG, cs.CC, cs.CL, stat.ML
Recent studies have discovered that Chain-of-Thought prompting (CoT) can dramatically improve the performance of Large Language Models (LLMs), particularly when dealing with complex tasks involving mathematics or reasoning. Despite the enormous empirical success, the underlying mechanisms behind CoT and how it unlocks ...
http://arxiv.org/pdf/2305.15408v4
2023-11-05T11:40:50
2023-05-24T17:59:21
{ "title": [ -0.010253906, -0.007160187, -0.04171753, -0.019226074, -0.00052690506, -0.02545166, 0.0072135925, -0.002199173, -0.020050049, 0.0073509216, -0.008216858, -0.013519287, -0.012298584, -0.043823242, -0.015853882, -0.009712219, 0.0005311966,...
http://arxiv.org/abs/2305.15405v1
Unit-based Speech-to-Speech Translation Without Parallel Data
Anuj Diwan, Anirudh Srinivasan, David Harwath, Eunsol Choi
cs.CL, eess.AS
We propose an unsupervised speech-to-speech translation (S2ST) system that does not rely on parallel data between the source and target languages. Our approach maps source and target language speech signals into automatically discovered, discrete units and reformulates the problem as unsupervised unit-to-unit machine t...
http://arxiv.org/pdf/2305.15405v1
2023-05-24T17:59:05
2023-05-24T17:59:05
{ "title": [ -0.03363037, -0.0074691772, -0.04522705, -0.070007324, 0.0023880005, 0.0009584427, -0.049804688, 0.025161743, -0.026062012, 0.05645752, -0.017974854, 0.013534546, 0.02645874, -0.016708374, -0.04043579, 0.009712219, -0.002937317, 0.02...
http://arxiv.org/abs/2305.15403v1
AV-TranSpeech: Audio-Visual Robust Speech-to-Speech Translation
Rongjie Huang, Huadai Liu, Xize Cheng, Yi Ren, Linjun Li, Zhenhui Ye, Jinzheng He, Lichao Zhang, Jinglin Liu, Xiang Yin, Zhou Zhao
cs.CL, cs.SD, eess.AS
Direct speech-to-speech translation (S2ST) aims to convert speech from one language into another, and has demonstrated significant progress to date. Despite the recent success, current S2ST models still suffer from distinct degradation in noisy environments and fail to translate visual speech (i.e., the movement of lip...
http://arxiv.org/pdf/2305.15403v1
2023-05-24T17:59:03
2023-05-24T17:59:03
{ "title": [ -0.009979248, -0.007358551, -0.021713257, -0.07409668, 0.007549286, 0.0061950684, -0.043151855, -0.0018901825, -0.0077552795, 0.055023193, -0.027511597, 0.023223877, 0.027053833, -0.006454468, -0.054016113, -0.00091838837, -0.0011615753,...
http://arxiv.org/abs/2305.15389v1
Comparing Humans and Models on a Similar Scale: Towards Cognitive Gender Bias Evaluation in Coreference Resolution
Gili Lior, Gabriel Stanovsky
cs.CL
Spurious correlations were found to be an important factor explaining model performance in various NLP tasks (e.g., gender or racial artifacts), often considered to be ''shortcuts'' to the actual task. However, humans tend to similarly make quick (and sometimes wrong) predictions based on societal and cognitive presupp...
http://arxiv.org/pdf/2305.15389v1
2023-05-24T17:51:44
2023-05-24T17:51:44
{ "title": [ -0.0014381409, 0.0079193115, -0.041259766, -0.03186035, -0.003709793, -0.042022705, -0.0053901672, -0.0071983337, -0.008163452, 0.065979004, -0.033081055, 0.007987976, 0.0501709, 0.010093689, -0.061645508, -0.004299164, -0.03488159, ...
http://arxiv.org/abs/2305.15387v1
Peek Across: Improving Multi-Document Modeling via Cross-Document Question-Answering
Avi Caciularu, Matthew E. Peters, Jacob Goldberger, Ido Dagan, Arman Cohan
cs.CL, cs.AI
The integration of multi-document pre-training objectives into language models has resulted in remarkable improvements in multi-document downstream tasks. In this work, we propose extending this idea by pre-training a generic multi-document model from a novel cross-document question answering pre-training objective. To...
http://arxiv.org/pdf/2305.15387v1
2023-05-24T17:48:40
2023-05-24T17:48:40
{ "title": [ -0.023376465, -0.0040397644, -0.03353882, -0.061157227, -0.0012292862, 0.0079574585, -0.04559326, -0.007663727, 0.018432617, 0.019195557, -0.04095459, 0.016906738, 0.0058288574, 0.006706238, -0.043182373, -0.0011806488, -0.0065956116, ...
http://arxiv.org/abs/2305.15386v2
Vistaar: Diverse Benchmarks and Training Sets for Indian Language ASR
Kaushal Santosh Bhogale, Sai Sundaresan, Abhigyan Raman, Tahir Javed, Mitesh M. Khapra, Pratyush Kumar
cs.CL, cs.SD, eess.AS
Improving ASR systems is necessary to make new LLM-based use-cases accessible to people across the globe. In this paper, we focus on Indian languages, and make the case that diverse benchmarks are required to evaluate and improve ASR systems for Indian languages. To address this, we collate Vistaar as a set of 59 bench...
http://arxiv.org/pdf/2305.15386v2
2023-08-02T13:29:31
2023-05-24T17:46:03
{ "title": [ -0.033996582, -0.006198883, -0.054382324, -0.06744385, -0.0042915344, -0.0027828217, -0.061065674, 0.019500732, -0.03643799, 0.03652954, -0.0059318542, 0.021499634, 0.025314331, -0.036346436, -0.009552002, 0.009315491, 0.004207611, 0...
http://arxiv.org/abs/2305.15380v1
Sentiment Analysis Using Aligned Word Embeddings for Uralic Languages
Khalid Alnajjar, Mika Hämäläinen, Jack Rueter
cs.CL
In this paper, we present an approach for translating word embeddings from a majority language into 4 minority languages: Erzya, Moksha, Udmurt and Komi-Zyrian. Furthermore, we align these word embeddings and present a novel neural network model that is trained on English data to conduct sentiment analysis and then app...
http://arxiv.org/pdf/2305.15380v1
2023-05-24T17:40:20
2023-05-24T17:40:20
{ "title": [ 0.00087594986, 0.003232956, -0.037200928, -0.040740967, 0.010765076, -0.005683899, -0.0357666, -0.009277344, 0.015625, 0.06286621, -0.041748047, 0.002439499, 0.050750732, 0.011558533, -0.053009033, 0.017364502, -0.013191223, 0.030364...
http://arxiv.org/abs/2305.15377v1
Uncovering and Quantifying Social Biases in Code Generation
Yan Liu, Xiaokang Chen, Yan Gao, Zhe Su, Fengji Zhang, Daoguang Zan, Jian-Guang Lou, Pin-Yu Chen, Tsung-Yi Ho
cs.CL
With the popularity of automatic code generation tools, such as Copilot, the study of the potential hazards of these tools is gaining importance. In this work, we explore the social bias problem in pre-trained code generation models. We propose a new paradigm to construct code prompts and successfully uncover social bi...
http://arxiv.org/pdf/2305.15377v1
2023-05-24T17:37:33
2023-05-24T17:37:33
{ "title": [ -0.032958984, -0.023254395, -0.040893555, -0.03866577, -0.012420654, -0.01777649, -0.024230957, 0.0019435883, -0.0044021606, 0.012413025, -0.023803711, 0.013114929, -0.022857666, -0.016784668, -0.01638794, -0.0149383545, -0.02758789, ...
http://arxiv.org/abs/2305.15374v1
ASPER: Answer Set Programming Enhanced Neural Network Models for Joint Entity-Relation Extraction
Trung Hoang Le, Huiping Cao, Tran Cao Son
cs.CL, cs.AI, cs.LG
A plethora of approaches have been proposed for joint entity-relation (ER) extraction. Most of these methods largely depend on a large amount of manually annotated training data. However, manual data annotation is time consuming, labor intensive, and error prone. Human beings learn using both data (through induction) a...
http://arxiv.org/pdf/2305.15374v1
2023-05-24T17:32:58
2023-05-24T17:32:58
{ "title": [ -0.019836426, 0.011413574, -0.046142578, -0.07269287, 0.028015137, 0.004722595, -0.036895752, -0.0044822693, 0.02027893, 0.014862061, 0.0031490326, 0.0007853508, -0.009506226, -0.014884949, -0.04876709, 0.011497498, -0.01902771, 0.03...
http://arxiv.org/abs/2305.15358v1
Context-Aware Transformer Pre-Training for Answer Sentence Selection
Luca Di Liello, Siddhant Garg, Alessandro Moschitti
cs.CL, cs.LG
Answer Sentence Selection (AS2) is a core component for building an accurate Question Answering pipeline. AS2 models rank a set of candidate sentences based on how likely they answer a given question. The state of the art in AS2 exploits pre-trained transformers by transferring them on large annotated datasets, while u...
http://arxiv.org/pdf/2305.15358v1
2023-05-24T17:10:45
2023-05-24T17:10:45
{ "title": [ 0.014083862, 0.018417358, -0.027130127, -0.04171753, 0.009147644, -0.022277832, 0.001996994, -0.0008058548, 0.002500534, 0.04562378, -0.025970459, 0.014328003, 0.034332275, -0.014122009, -0.03164673, 0.013916016, 0.0030288696, 0.0347...
http://arxiv.org/abs/2305.15344v1
Learning Answer Generation using Supervision from Automatic Question Answering Evaluators
Matteo Gabburo, Siddhant Garg, Rik Koncel-Kedziorski, Alessandro Moschitti
cs.CL, cs.LG
Recent studies show that sentence-level extractive QA, i.e., based on Answer Sentence Selection (AS2), is outperformed by Generation-based QA (GenQA) models, which generate answers using the top-k answer sentences ranked by AS2 models (a la retrieval-augmented generation style). In this paper, we propose a novel traini...
http://arxiv.org/pdf/2305.15344v1
2023-05-24T16:57:04
2023-05-24T16:57:04
{ "title": [ -0.007926941, -0.0019721985, -0.022872925, -0.0345459, -0.010238647, 0.0009803772, -0.01424408, 0.0014133453, 0.018585205, 0.015472412, -0.026733398, 0.0065727234, -0.028213501, -0.0066184998, -0.060668945, -0.011932373, -0.0068626404, ...
http://arxiv.org/abs/2305.15338v1
Measuring and Mitigating Constraint Violations of In-Context Learning for Utterance-to-API Semantic Parsing
Shufan Wang, Sebastien Jean, Sailik Sengupta, James Gung, Nikolaos Pappas, Yi Zhang
cs.AI, cs.CL
In executable task-oriented semantic parsing, the system aims to translate users' utterances in natural language to machine-interpretable programs (API calls) that can be executed according to pre-defined API specifications. With the popularity of Large Language Models (LLMs), in-context learning offers a strong baseli...
http://arxiv.org/pdf/2305.15338v1
2023-05-24T16:50:36
2023-05-24T16:50:36
{ "title": [ -0.006706238, -0.007686615, -0.049072266, -0.05795288, 0.018417358, -0.013679504, -0.022781372, 0.017913818, -0.00818634, 0.022857666, -0.068603516, 0.03933716, 0.0027751923, 0.008926392, -0.022644043, -0.024765015, -0.019607544, 0.0...
http://arxiv.org/abs/2305.15334v1
Gorilla: Large Language Model Connected with Massive APIs
Shishir G. Patil, Tianjun Zhang, Xin Wang, Joseph E. Gonzalez
cs.CL, cs.AI
Large Language Models (LLMs) have seen an impressive wave of advances recently, with models now excelling in a variety of tasks, such as mathematical reasoning and program synthesis. However, their potential to effectively use tools via API calls remains unfulfilled. This is a challenging task even for today's state-of...
http://arxiv.org/pdf/2305.15334v1
2023-05-24T16:48:11
2023-05-24T16:48:11
{ "title": [ -0.033172607, -0.029891968, -0.055877686, -0.07550049, -0.007259369, 0.011352539, -0.065979004, 0.023345947, 0.0033016205, 0.024368286, -0.013916016, -0.00198555, 0.026229858, -0.024734497, -0.025619507, 0.018920898, -0.007881165, 0....