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/2306.08909v1
Bridging the Gap between Decision and Logits in Decision-based Knowledge Distillation for Pre-trained Language Models
Qinhong Zhou, Zonghan Yang, Peng Li, Yang Liu
cs.CL
Conventional knowledge distillation (KD) methods require access to the internal information of teachers, e.g., logits. However, such information may not always be accessible for large pre-trained language models (PLMs). In this work, we focus on decision-based KD for PLMs, where only teacher decisions (i.e., top-1 labe...
http://arxiv.org/pdf/2306.08909v1
2023-06-15T07:23:44
2023-06-15T07:23:44
{ "title": [ -0.021743774, -0.020477295, -0.04284668, -0.08709717, -0.023468018, 0.0020103455, -0.09313965, 0.0076408386, 0.018600464, 0.023910522, -0.038909912, 0.00010961294, 0.008888245, -0.016662598, -0.033935547, -0.010437012, 0.001613617, 0...
http://arxiv.org/abs/2306.08906v1
Participatory Research as a Path to Community-Informed, Gender-Fair Machine Translation
Dagmar Gromann, Manuel Lardelli, Katta Spiel, Sabrina Burtscher, Lukas Daniel Klausner, Arthur Mettinger, Igor Miladinovic, Sigrid Schefer-Wenzl, Daniela Duh, Katharina Bühn
cs.CL, cs.CY
Recent years have seen a strongly increased visibility of non-binary people in public discourse. Accordingly, considerations of gender-fair language go beyond a binary conception of male/female. However, language technology, especially machine translation (MT), still suffers from binary gender bias. Proposing a solutio...
http://arxiv.org/pdf/2306.08906v1
2023-06-15T07:20:14
2023-06-15T07:20:14
{ "title": [ -0.052856445, -0.025161743, -0.036315918, -0.033203125, -0.015571594, -0.0063323975, -0.009628296, 0.030395508, -0.02911377, 0.066345215, -0.008888245, 0.018157959, 0.029586792, -0.005958557, -0.0574646, 0.012901306, -0.014274597, 0....
http://arxiv.org/abs/2306.08896v1
Multilingual End to End Entity Linking
Mikhail Plekhanov, Nora Kassner, Kashyap Popat, Louis Martin, Simone Merello, Borislav Kozlovskii, Frédéric A. Dreyer, Nicola Cancedda
cs.CL
Entity Linking is one of the most common Natural Language Processing tasks in practical applications, but so far efficient end-to-end solutions with multilingual coverage have been lacking, leading to complex model stacks. To fill this gap, we release and open source BELA, the first fully end-to-end multilingual entity...
http://arxiv.org/pdf/2306.08896v1
2023-06-15T07:02:00
2023-06-15T07:02:00
{ "title": [ 0.0146102905, -0.00027370453, -0.034820557, -0.09289551, 0.009254456, -0.008804321, -0.045898438, 0.008354187, -0.0024433136, 0.062438965, 0.036102295, -0.016098022, 0.06903076, -0.068847656, 0.01777649, 0.049102783, 0.0025367737, 0....
http://arxiv.org/abs/2306.08892v1
MetricPrompt: Prompting Model as a Relevance Metric for Few-shot Text Classification
Hongyuan Dong, Weinan Zhang, Wanxiang Che
cs.CL
Prompting methods have shown impressive performance in a variety of text mining tasks and applications, especially few-shot ones. Despite the promising prospects, the performance of prompting model largely depends on the design of prompt template and verbalizer. In this work, we propose MetricPrompt, which eases verbal...
http://arxiv.org/pdf/2306.08892v1
2023-06-15T06:51:35
2023-06-15T06:51:35
{ "title": [ -0.0025596619, -0.024963379, -0.0064086914, -0.031021118, -0.02545166, -0.006511688, -0.019805908, -0.0032138824, 0.007457733, 0.013771057, -0.045013428, 0.0234375, -0.004875183, -0.007774353, -0.042114258, -0.03086853, 0.00283432, 0...
http://arxiv.org/abs/2306.08891v1
Interleaving Pre-Trained Language Models and Large Language Models for Zero-Shot NL2SQL Generation
Zihui Gu, Ju Fan, Nan Tang, Songyue Zhang, Yuxin Zhang, Zui Chen, Lei Cao, Guoliang Li, Sam Madden, Xiaoyong Du
cs.CL
Zero-shot NL2SQL is crucial in achieving natural language to SQL that is adaptive to new environments (e.g., new databases, new linguistic phenomena or SQL structures) with zero annotated NL2SQL samples from such environments. Existing approaches either fine-tune pre-trained language models (PLMs) based on annotated da...
http://arxiv.org/pdf/2306.08891v1
2023-06-15T06:50:51
2023-06-15T06:50:51
{ "title": [ -0.020446777, -0.020401001, -0.046173096, -0.088256836, -0.0038318634, 0.01977539, -0.06774902, 0.044708252, 0.01965332, 0.04827881, -0.02279663, -0.016113281, 0.01626587, -0.007972717, -0.051513672, -0.0017433167, -0.0025157928, 0.0...
http://arxiv.org/abs/2307.11519v1
Multi-modal Hate Speech Detection using Machine Learning
Fariha Tahosin Boishakhi, Ponkoj Chandra Shill, Md. Golam Rabiul Alam
cs.AI, cs.CL, cs.CV, cs.LG, cs.SD, eess.AS
With the continuous growth of internet users and media content, it is very hard to track down hateful speech in audio and video. Converting video or audio into text does not detect hate speech accurately as human sometimes uses hateful words as humorous or pleasant in sense and also uses different voice tones or show d...
http://arxiv.org/pdf/2307.11519v1
2023-06-15T06:46:52
2023-06-15T06:46:52
{ "title": [ 0.009239197, 0.008987427, -0.042175293, -0.040496826, 0.016082764, -0.034698486, -0.0113220215, -0.01737976, -0.010772705, 0.03805542, -0.012901306, -0.002456665, 0.015266418, -0.022384644, -0.036315918, -0.0058021545, -0.013793945, ...
http://arxiv.org/abs/2306.08877v2
Linguistic Binding in Diffusion Models: Enhancing Attribute Correspondence through Attention Map Alignment
Royi Rassin, Eran Hirsch, Daniel Glickman, Shauli Ravfogel, Yoav Goldberg, Gal Chechik
cs.CL, cs.CV
Text-conditioned image generation models often generate incorrect associations between entities and their visual attributes. This reflects an impaired mapping between linguistic binding of entities and modifiers in the prompt and visual binding of the corresponding elements in the generated image. As one notable exampl...
http://arxiv.org/pdf/2306.08877v2
2023-10-20T12:01:44
2023-06-15T06:21:44
{ "title": [ -0.016738892, 0.0004005432, -0.021972656, -0.009590149, -0.03024292, -0.0011911392, -0.018478394, -0.030197144, 0.00944519, 0.052825928, 0.0143585205, -0.010627747, 0.01638794, -0.02456665, -0.03894043, 0.011550903, -0.01852417, 0.02...
http://arxiv.org/abs/2306.08872v1
Neural models for Factual Inconsistency Classification with Explanations
Tathagata Raha, Mukund Choudhary, Abhinav Menon, Harshit Gupta, KV Aditya Srivatsa, Manish Gupta, Vasudeva Varma
cs.CL, cs.AI
Factual consistency is one of the most important requirements when editing high quality documents. It is extremely important for automatic text generation systems like summarization, question answering, dialog modeling, and language modeling. Still, automated factual inconsistency detection is rather under-studied. Exi...
http://arxiv.org/pdf/2306.08872v1
2023-06-15T06:06:50
2023-06-15T06:06:50
{ "title": [ -0.011497498, 0.019607544, -0.032958984, -0.051086426, 0.008850098, -0.003408432, -0.021026611, -0.013763428, -0.016998291, 0.023925781, -0.02130127, -0.014305115, 0.0036373138, -0.028152466, -0.03668213, -0.018707275, -0.029876709, ...
http://arxiv.org/abs/2306.08852v1
BED: Bi-Encoder-Based Detectors for Out-of-Distribution Detection
Louis Owen, Biddwan Ahmed, Abhay Kumar
cs.CL
This paper introduces a novel method leveraging bi-encoder-based detectors along with a comprehensive study comparing different out-of-distribution (OOD) detection methods in NLP using different feature extractors. The feature extraction stage employs popular methods such as Universal Sentence Encoder (USE), BERT, MPNE...
http://arxiv.org/pdf/2306.08852v1
2023-06-15T04:41:28
2023-06-15T04:41:28
{ "title": [ -0.0009803772, 0.013298035, -0.016052246, -0.02986145, 0.023269653, 0.0007901192, 0.021606445, -0.035705566, 0.011390686, 0.02760315, -0.035827637, 0.0050582886, -0.00021636486, -0.032440186, -0.027450562, -0.009651184, -0.010017395, ...
http://arxiv.org/abs/2306.08847v1
Improving Reading Comprehension Question Generation with Data Augmentation and Overgenerate-and-rank
Nischal Ashok Kumar, Nigel Fernandez, Zichao Wang, Andrew Lan
cs.CL, cs.CY, cs.LG
Reading comprehension is a crucial skill in many aspects of education, including language learning, cognitive development, and fostering early literacy skills in children. Automated answer-aware reading comprehension question generation has significant potential to scale up learner support in educational activities. On...
http://arxiv.org/pdf/2306.08847v1
2023-06-15T04:23:25
2023-06-15T04:23:25
{ "title": [ -0.000395298, -0.002937317, -0.025726318, -0.017364502, -0.01991272, -0.01146698, -0.014213562, -0.027145386, 0.024353027, 0.005012512, -0.04171753, 0.013328552, -0.02003479, -0.0051460266, -0.03894043, -0.019134521, -0.015625, 0.009...
http://arxiv.org/abs/2306.14905v1
PRISMA-DFLLM: An Extension of PRISMA for Systematic Literature Reviews using Domain-specific Finetuned Large Language Models
Teo Susnjak
cs.CL, cs.AI
With the proliferation of open-sourced Large Language Models (LLMs) and efficient finetuning techniques, we are on the cusp of the emergence of numerous domain-specific LLMs that have been finetuned for expertise across specialized fields and applications for which the current general-purpose LLMs are unsuitable. In ac...
http://arxiv.org/pdf/2306.14905v1
2023-06-15T02:52:50
2023-06-15T02:52:50
{ "title": [ -0.035064697, -0.012741089, -0.045440674, -0.06915283, -0.0082473755, 0.008232117, -0.08312988, -0.031585693, 0.008216858, 0.033355713, 0.0031986237, 0.007980347, 0.02331543, -0.026565552, -0.04208374, 0.029708862, -0.027786255, 0.03...
http://arxiv.org/abs/2306.08818v1
Pragmatic Inference with a CLIP Listener for Contrastive Captioning
Jiefu Ou, Benno Krojer, Daniel Fried
cs.CL
We propose a simple yet effective and robust method for contrastive captioning: generating discriminative captions that distinguish target images from very similar alternative distractor images. Our approach is built on a pragmatic inference procedure that formulates captioning as a reference game between a speaker, wh...
http://arxiv.org/pdf/2306.08818v1
2023-06-15T02:22:28
2023-06-15T02:22:28
{ "title": [ 0.03286743, -0.000010848045, -0.0048675537, -0.066345215, -0.012001038, -0.013160706, -0.028213501, -0.013160706, 0.0072364807, 0.02268982, -0.028381348, -0.0018501282, -0.016418457, -0.023452759, -0.041412354, -0.023468018, -0.001936912...
http://arxiv.org/abs/2306.08817v1
Description-Enhanced Label Embedding Contrastive Learning for Text Classification
Kun Zhang, Le Wu, Guangyi Lv, Enhong Chen, Shulan Ruan, Jing Liu, Zhiqiang Zhang, Jun Zhou, Meng Wang
cs.CL, cs.IR
Text Classification is one of the fundamental tasks in natural language processing, which requires an agent to determine the most appropriate category for input sentences. Recently, deep neural networks have achieved impressive performance in this area, especially Pre-trained Language Models (PLMs). Usually, these meth...
http://arxiv.org/pdf/2306.08817v1
2023-06-15T02:19:34
2023-06-15T02:19:34
{ "title": [ -0.005176544, 0.0071144104, -0.040283203, -0.03994751, 0.017440796, -0.02861023, 0.007472992, -0.015083313, 0.006248474, 0.023849487, -0.037963867, 0.030288696, -0.002981186, -0.007965088, -0.032928467, -0.010498047, -0.02835083, 0.0...
http://arxiv.org/abs/2306.08804v2
PEACE: Cross-Platform Hate Speech Detection- A Causality-guided Framework
Paras Sheth, Tharindu Kumarage, Raha Moraffah, Aman Chadha, Huan Liu
cs.CL, cs.LG
Hate speech detection refers to the task of detecting hateful content that aims at denigrating an individual or a group based on their religion, gender, sexual orientation, or other characteristics. Due to the different policies of the platforms, different groups of people express hate in different ways. Furthermore, d...
http://arxiv.org/pdf/2306.08804v2
2023-10-08T21:44:47
2023-06-15T01:18:02
{ "title": [ 0.022964478, 0.014228821, -0.0368042, -0.046020508, 0.02166748, -0.025665283, -0.014442444, -0.02519226, -0.018798828, 0.024093628, -0.010383606, 0.027648926, 0.019577026, -0.019607544, -0.015129089, 0.00034856796, -0.024368286, 0.06...
http://arxiv.org/abs/2306.08798v1
MPSA-DenseNet: A novel deep learning model for English accent classification
Tianyu Song, Linh Thi Hoai Nguyen, Ton Viet Ta
cs.CL, stat.ML
This paper presents three innovative deep learning models for English accent classification: Multi-DenseNet, PSA-DenseNet, and MPSE-DenseNet, that combine multi-task learning and the PSA module attention mechanism with DenseNet. We applied these models to data collected from six dialects of English across native Englis...
http://arxiv.org/pdf/2306.08798v1
2023-06-15T01:03:54
2023-06-15T01:03:54
{ "title": [ 0.01776123, 0.0061035156, -0.02128601, -0.03692627, 0.0040893555, -0.010772705, -0.015014648, -0.0154418945, -0.00039982796, 0.03173828, -0.0055732727, -0.010292053, 0.023284912, -0.039764404, -0.04095459, 0.0076828003, -0.010284424, ...
http://arxiv.org/abs/2306.08756v1
Recipes for Sequential Pre-training of Multilingual Encoder and Seq2Seq Models
Saleh Soltan, Andy Rosenbaum, Tobias Falke, Qin Lu, Anna Rumshisky, Wael Hamza
cs.CL, cs.AI, cs.LG
Pre-trained encoder-only and sequence-to-sequence (seq2seq) models each have advantages, however training both model types from scratch is computationally expensive. We explore recipes to improve pre-training efficiency by initializing one model from the other. (1) Extracting the encoder from a seq2seq model, we show i...
http://arxiv.org/pdf/2306.08756v1
2023-06-14T21:41:52
2023-06-14T21:41:52
{ "title": [ -0.019607544, -0.0017061234, -0.027832031, -0.062316895, -0.0076522827, 0.004173279, -0.07171631, -0.020019531, -0.01272583, 0.046661377, 0.026153564, 0.004299164, 0.013145447, -0.048339844, -0.0109939575, 0.03086853, 0.0061187744, 0...
http://arxiv.org/abs/2306.08753v3
Unified model for code-switching speech recognition and language identification based on a concatenated tokenizer
Kunal Dhawan, Dima Rekesh, Boris Ginsburg
eess.AS, cs.CL, cs.SD
Code-Switching (CS) multilingual Automatic Speech Recognition (ASR) models can transcribe speech containing two or more alternating languages during a conversation. This paper proposes (1) a new method for creating code-switching ASR datasets from purely monolingual data sources, and (2) a novel Concatenated Tokenizer ...
http://arxiv.org/pdf/2306.08753v3
2023-09-16T05:32:12
2023-06-14T21:24:11
{ "title": [ -0.0055351257, 0.014381409, -0.046966553, -0.040924072, 0.0009613037, -0.013664246, -0.050842285, -0.009567261, -0.038482666, 0.051879883, -0.01436615, -0.0024204254, 0.01663208, -0.034332275, -0.011787415, -0.011405945, -0.018463135, ...
http://arxiv.org/abs/2306.08749v2
Utilizing Longitudinal Chest X-Rays and Reports to Pre-Fill Radiology Reports
Qingqing Zhu, Tejas Sudharshan Mathai, Pritam Mukherjee, Yifan Peng, Ronald M. Summers, Zhiyong Lu
cs.CL, cs.LG
Despite the reduction in turn-around times in radiology reports with the use of speech recognition software, persistent communication errors can significantly impact the interpretation of the radiology report. Pre-filling a radiology report holds promise in mitigating reporting errors, and despite efforts in the litera...
http://arxiv.org/pdf/2306.08749v2
2023-10-10T17:29:52
2023-06-14T21:17:31
{ "title": [ 0.04812622, -0.00032687187, -0.040527344, -0.079711914, 0.0053596497, -0.026016235, -0.016159058, -0.003648758, -0.028671265, 0.007709503, -0.060028076, 0.016204834, -0.017791748, -0.013572693, 0.031829834, -0.050964355, 0.004627228, ...
http://arxiv.org/abs/2306.08702v3
Does mBERT understand Romansh? Evaluating word embeddings using word alignment
Eyal Liron Dolev
cs.CL
We test similarity-based word alignment models (SimAlign and awesome-align) in combination with word embeddings from mBERT and XLM-R on parallel sentences in German and Romansh. Since Romansh is an unseen language, we are dealing with a zero-shot setting. Using embeddings from mBERT, both models reach an alignment erro...
http://arxiv.org/pdf/2306.08702v3
2023-08-17T08:54:42
2023-06-14T19:00:12
{ "title": [ -0.010673523, 0.017959595, -0.029830933, -0.033447266, 0.008415222, -0.009552002, -0.045898438, -0.030639648, -0.007972717, 0.046051025, -0.022964478, 0.019378662, 0.033355713, -0.014694214, -0.026473999, 0.019256592, -0.01096344, 0....
http://arxiv.org/abs/2306.08685v1
World-to-Words: Grounded Open Vocabulary Acquisition through Fast Mapping in Vision-Language Models
Ziqiao Ma, Jiayi Pan, Joyce Chai
cs.CL, cs.AI, cs.CV
The ability to connect language units to their referents in the physical world, referred to as grounding, is crucial to learning and understanding grounded meanings of words. While humans demonstrate fast mapping in new word learning, it remains unclear whether modern vision-language models can truly represent language...
http://arxiv.org/pdf/2306.08685v1
2023-06-14T18:10:05
2023-06-14T18:10:05
{ "title": [ -0.035461426, -0.0066719055, -0.017852783, -0.02583313, -0.022827148, -0.0055656433, -0.023010254, -0.021652222, 0.014877319, 0.016677856, -0.053466797, 0.022201538, -0.00712204, -0.0110321045, -0.0317688, -0.017166138, -0.0015926361, ...
http://arxiv.org/abs/2306.08667v1
When to Use Efficient Self Attention? Profiling Text, Speech and Image Transformer Variants
Anuj Diwan, Eunsol Choi, David Harwath
cs.CL, cs.SD, eess.AS
We present the first unified study of the efficiency of self-attention-based Transformer variants spanning text, speech and vision. We identify input length thresholds (tipping points) at which efficient Transformer variants become more efficient than vanilla models, using a variety of efficiency metrics (latency, thro...
http://arxiv.org/pdf/2306.08667v1
2023-06-14T17:59:02
2023-06-14T17:59:02
{ "title": [ 0.01574707, 0.0284729, -0.052856445, -0.022445679, 0.022994995, -0.049926758, 0.005634308, -0.013946533, -0.03640747, 0.05053711, -0.013801575, -0.00018131733, 0.02746582, -0.027404785, -0.01272583, -0.01637268, 0.009437561, 0.054473...
http://arxiv.org/abs/2306.08666v1
Radiology-GPT: A Large Language Model for Radiology
Zhengliang Liu, Aoxiao Zhong, Yiwei Li, Longtao Yang, Chao Ju, Zihao Wu, Chong Ma, Peng Shu, Cheng Chen, Sekeun Kim, Haixing Dai, Lin Zhao, Dajiang Zhu, Jun Liu, Wei Liu, Dinggang Shen, Xiang Li, Quanzheng Li, Tianming Liu
cs.CL, cs.AI
We introduce Radiology-GPT, a large language model for radiology. Using an instruction tuning approach on an extensive dataset of radiology domain knowledge, Radiology-GPT demonstrates superior performance compared to general language models such as StableLM, Dolly and LLaMA. It exhibits significant versatility in radi...
http://arxiv.org/pdf/2306.08666v1
2023-06-14T17:57:24
2023-06-14T17:57:24
{ "title": [ -0.025817871, -0.005077362, -0.040985107, -0.061676025, 0.010360718, 0.009391785, -0.05404663, -0.0074768066, 0.009262085, 0.0317688, -0.005207062, -0.009010315, 0.00084257126, -0.030380249, -0.0446167, 0.013572693, 0.009033203, 0.00...
http://arxiv.org/abs/2306.08658v1
Babel-ImageNet: Massively Multilingual Evaluation of Vision-and-Language Representations
Gregor Geigle, Radu Timofte, Goran Glavaš
cs.CL, cs.CV
Vision-and-language (VL) models with separate encoders for each modality (e.g., CLIP) have become the go-to models for zero-shot image classification and image-text retrieval. The bulk of the evaluation of these models is, however, performed with English text only: the costly creation of language-specific image-caption...
http://arxiv.org/pdf/2306.08658v1
2023-06-14T17:53:06
2023-06-14T17:53:06
{ "title": [ 0.0003812313, -0.00415802, -0.023864746, -0.050231934, 0.013618469, -0.0035629272, -0.037506104, -0.02796936, 0.024154663, 0.031219482, -0.0132369995, -0.0041503906, -0.0011453629, -0.019317627, -0.046173096, 0.013259888, 0.0054740906, ...
http://arxiv.org/abs/2306.08588v1
Improving Code-Switching and Named Entity Recognition in ASR with Speech Editing based Data Augmentation
Zheng Liang, Zheshu Song, Ziyang Ma, Chenpeng Du, Kai Yu, Xie Chen
cs.CL, cs.SD, eess.AS
Recently, end-to-end (E2E) automatic speech recognition (ASR) models have made great strides and exhibit excellent performance in general speech recognition. However, there remain several challenging scenarios that E2E models are not competent in, such as code-switching and named entity recognition (NER). Data augmenta...
http://arxiv.org/pdf/2306.08588v1
2023-06-14T15:50:13
2023-06-14T15:50:13
{ "title": [ 0.00674057, -0.0016517639, -0.04827881, -0.059387207, 0.011711121, -0.007171631, -0.028121948, 0.027008057, -0.02519226, 0.020202637, -0.019546509, 0.004825592, 0.015167236, -0.02960205, -0.0075035095, -0.001080513, -0.0098724365, 0....
http://arxiv.org/abs/2306.10062v1
Revealing the structure of language model capabilities
Ryan Burnell, Han Hao, Andrew R. A. Conway, Jose Hernandez Orallo
cs.CL, cs.AI, cs.LG
Building a theoretical understanding of the capabilities of large language models (LLMs) is vital for our ability to predict and explain the behavior of these systems. Here, we investigate the structure of LLM capabilities by extracting latent capabilities from patterns of individual differences across a varied populat...
http://arxiv.org/pdf/2306.10062v1
2023-06-14T15:43:25
2023-06-14T15:43:25
{ "title": [ -0.018875122, -0.013679504, -0.03604126, -0.06945801, -0.042633057, -0.0045318604, -0.08239746, 0.013511658, -0.02645874, 0.032928467, -0.0024375916, -0.010650635, -0.0036506653, -0.041046143, -0.030654907, 0.005760193, 0.01007843, 0...
http://arxiv.org/abs/2306.08582v1
Tagged End-to-End Simultaneous Speech Translation Training using Simultaneous Interpretation Data
Yuka Ko, Ryo Fukuda, Yuta Nishikawa, Yasumasa Kano, Katsuhito Sudoh, Satoshi Nakamura
cs.CL, cs.SD, eess.AS
Simultaneous speech translation (SimulST) translates partial speech inputs incrementally. Although the monotonic correspondence between input and output is preferable for smaller latency, it is not the case for distant language pairs such as English and Japanese. A prospective approach to this problem is to mimic simul...
http://arxiv.org/pdf/2306.08582v1
2023-06-14T15:42:06
2023-06-14T15:42:06
{ "title": [ 0.0024852753, 0.012390137, -0.047180176, -0.07757568, -0.0029029846, -0.017700195, -0.04776001, 0.02571106, -0.06323242, 0.048461914, -0.0077056885, 0.0043296814, 0.0211792, -0.045410156, 0.026687622, -0.018218994, 0.0001437664, 0.03...
http://arxiv.org/abs/2306.08577v1
Learning Cross-lingual Mappings for Data Augmentation to Improve Low-Resource Speech Recognition
Muhammad Umar Farooq, Thomas Hain
cs.CL, cs.SD, eess.AS
Exploiting cross-lingual resources is an effective way to compensate for data scarcity of low resource languages. Recently, a novel multilingual model fusion technique has been proposed where a model is trained to learn cross-lingual acoustic-phonetic similarities as a mapping function. However, handcrafted lexicons ha...
http://arxiv.org/pdf/2306.08577v1
2023-06-14T15:24:31
2023-06-14T15:24:31
{ "title": [ 0.000702858, -0.0012865067, -0.04736328, -0.07098389, 0.0072784424, -0.010879517, -0.053710938, 0.009918213, -0.009788513, 0.035217285, -0.03149414, 0.012031555, 0.008560181, -0.012283325, -0.030166626, -0.0068244934, -0.0038852692, ...
http://arxiv.org/abs/2306.08568v1
WizardCoder: Empowering Code Large Language Models with Evol-Instruct
Ziyang Luo, Can Xu, Pu Zhao, Qingfeng Sun, Xiubo Geng, Wenxiang Hu, Chongyang Tao, Jing Ma, Qingwei Lin, Daxin Jiang
cs.CL, cs.AI
Code Large Language Models (Code LLMs), such as StarCoder, have demonstrated exceptional performance in code-related tasks. However, most existing models are solely pre-trained on extensive raw code data without instruction fine-tuning. In this paper, we introduce WizardCoder, which empowers Code LLMs with complex inst...
http://arxiv.org/pdf/2306.08568v1
2023-06-14T15:18:48
2023-06-14T15:18:48
{ "title": [ -0.02796936, -0.029220581, -0.04815674, -0.08001709, -0.0059776306, 0.00724411, -0.060791016, 0.014175415, 0.0018939972, 0.02357483, -0.001244545, -0.0014219284, 0.014091492, -0.044281006, -0.021621704, 0.016647339, -0.0025367737, 0....
http://arxiv.org/abs/2306.08543v1
Knowledge Distillation of Large Language Models
Yuxian Gu, Li Dong, Furu Wei, Minlie Huang
cs.CL, cs.AI
Knowledge Distillation (KD) is a promising technique for reducing the high computational demand of large language models (LLMs). However, previous KD methods are primarily applied to white-box classification models or training small models to imitate black-box model APIs like ChatGPT. How to effectively distill the kno...
http://arxiv.org/pdf/2306.08543v1
2023-06-14T14:44:03
2023-06-14T14:44:03
{ "title": [ -0.03656006, -0.016433716, -0.06549072, -0.097839355, 0.004348755, 0.0068359375, -0.097839355, 0.024597168, 0.03918457, 0.023086548, -0.033355713, -0.02709961, 0.007762909, -0.0126571655, -0.04815674, -0.00064468384, 0.00198555, 0.02...
http://arxiv.org/abs/2306.08526v1
AlbMoRe: A Corpus of Movie Reviews for Sentiment Analysis in Albanian
Erion Çano
cs.CL, cs.AI, cs.LG
Lack of available resources such as text corpora for low-resource languages seriously hinders research on natural language processing and computational linguistics. This paper presents AlbMoRe, a corpus of 800 sentiment annotated movie reviews in Albanian. Each text is labeled as positive or negative and can be used fo...
http://arxiv.org/pdf/2306.08526v1
2023-06-14T14:21:55
2023-06-14T14:21:55
{ "title": [ 0.0012245178, -0.022216797, -0.013931274, -0.0289917, -0.009391785, -0.000056624413, -0.0045776367, -0.013389587, 0.004043579, 0.0519104, -0.031021118, -0.0099105835, 0.03842163, -0.0057640076, -0.05734253, 0.004245758, -0.0076789856, ...
http://arxiv.org/abs/2306.08505v1
DiffuDetox: A Mixed Diffusion Model for Text Detoxification
Griffin Floto, Mohammad Mahdi Abdollah Pour, Parsa Farinneya, Zhenwei Tang, Ali Pesaranghader, Manasa Bharadwaj, Scott Sanner
cs.CL, cs.LG
Text detoxification is a conditional text generation task aiming to remove offensive content from toxic text. It is highly useful for online forums and social media, where offensive content is frequently encountered. Intuitively, there are diverse ways to detoxify sentences while preserving their meanings, and we can s...
http://arxiv.org/pdf/2306.08505v1
2023-06-14T13:41:23
2023-06-14T13:41:23
{ "title": [ -0.0011453629, 0.007171631, -0.041534424, -0.055847168, 0.025878906, -0.0053710938, -0.019042969, -0.016494751, 0.023376465, 0.026916504, -0.013618469, -0.0031337738, -0.004547119, -0.030151367, -0.047180176, 0.0029087067, -0.020614624, ...
http://arxiv.org/abs/2306.08502v1
ITALIC: An Italian Intent Classification Dataset
Alkis Koudounas, Moreno La Quatra, Lorenzo Vaiani, Luca Colomba, Giuseppe Attanasio, Eliana Pastor, Luca Cagliero, Elena Baralis
cs.CL, cs.SD, eess.AS
Recent large-scale Spoken Language Understanding datasets focus predominantly on English and do not account for language-specific phenomena such as particular phonemes or words in different lects. We introduce ITALIC, the first large-scale speech dataset designed for intent classification in Italian. The dataset compri...
http://arxiv.org/pdf/2306.08502v1
2023-06-14T13:36:24
2023-06-14T13:36:24
{ "title": [ -0.025482178, 0.0031280518, -0.06210327, -0.022460938, 0.008735657, -0.027008057, -0.032562256, -0.01902771, 0.015792847, 0.037109375, -0.001531601, -0.0104904175, 0.024475098, -0.03237915, -0.045013428, 0.011940002, -0.0109939575, 0...
http://arxiv.org/abs/2306.08492v1
A Relaxed Optimization Approach for Adversarial Attacks against Neural Machine Translation Models
Sahar Sadrizadeh, Clément Barbier, Ljiljana Dolamic, Pascal Frossard
cs.CL
In this paper, we propose an optimization-based adversarial attack against Neural Machine Translation (NMT) models. First, we propose an optimization problem to generate adversarial examples that are semantically similar to the original sentences but destroy the translation generated by the target NMT model. This optim...
http://arxiv.org/pdf/2306.08492v1
2023-06-14T13:13:34
2023-06-14T13:13:34
{ "title": [ -0.031982422, 0.005077362, -0.006904602, -0.01838684, 0.01235199, 0.0024471283, -0.005180359, -0.0143966675, -0.049865723, 0.048217773, 0.023742676, 0.0041885376, -0.009841919, -0.024002075, -0.04788208, -0.007255554, -0.00046682358, ...
http://arxiv.org/abs/2306.10058v1
EM-Network: Oracle Guided Self-distillation for Sequence Learning
Ji Won Yoon, Sunghwan Ahn, Hyeonseung Lee, Minchan Kim, Seok Min Kim, Nam Soo Kim
cs.LG, cs.CL, eess.AS
We introduce EM-Network, a novel self-distillation approach that effectively leverages target information for supervised sequence-to-sequence (seq2seq) learning. In contrast to conventional methods, it is trained with oracle guidance, which is derived from the target sequence. Since the oracle guidance compactly repres...
http://arxiv.org/pdf/2306.10058v1
2023-06-14T12:24:55
2023-06-14T12:24:55
{ "title": [ -0.012008667, -0.016418457, -0.059783936, -0.07458496, -0.0032978058, -0.014419556, -0.050750732, 0.03289795, 0.0022697449, 0.038146973, -0.035217285, -0.00082683563, 0.0067863464, -0.011154175, -0.053955078, -0.012168884, 0.0119018555, ...
http://arxiv.org/abs/2306.08456v2
PoetryDiffusion: Towards Joint Semantic and Metrical Manipulation in Poetry Generation
Zhiyuan Hu, Chumin Liu, Yue Feng, Anh Tuan Luu, Bryan Hooi
cs.CL
Controllable text generation is a challenging and meaningful field in natural language generation (NLG). Especially, poetry generation is a typical one with well-defined and strict conditions for text generation which is an ideal playground for the assessment of current methodologies. While prior works succeeded in con...
http://arxiv.org/pdf/2306.08456v2
2023-08-15T16:21:47
2023-06-14T11:57:31
{ "title": [ -0.034851074, -0.021621704, -0.04675293, -0.05279541, -0.002412796, 0.00028181076, -0.034088135, 0.016098022, 0.015991211, 0.051818848, 0.009490967, 0.0012617111, 0.027938843, -0.012527466, -0.054992676, 0.014472961, -0.01828003, 0.0...
http://arxiv.org/abs/2306.08433v1
"Definition Modeling: To model definitions." Generating Definitions With Little to No Semantics
Vincent Segonne, Timothee Mickus
cs.CL
Definition Modeling, the task of generating definitions, was first proposed as a means to evaluate the semantic quality of word embeddings-a coherent lexical semantic representations of a word in context should contain all the information necessary to generate its definition. The relative novelty of this task entails t...
http://arxiv.org/pdf/2306.08433v1
2023-06-14T11:08:38
2023-06-14T11:08:38
{ "title": [ 0.0004682541, -0.000017106533, -0.01612854, -0.00024461746, -0.022399902, -0.009651184, -0.008628845, -0.038269043, -0.014884949, 0.010231018, 0.005039215, 0.008880615, -0.020599365, -0.048461914, 0.0050468445, -0.015022278, 0.011871338,...
http://arxiv.org/abs/2306.08403v1
Building a Corpus for Biomedical Relation Extraction of Species Mentions
Oumaima El Khettari, Solen Quiniou, Samuel Chaffron
cs.CL
We present a manually annotated corpus, Species-Species Interaction, for extracting meaningful binary relations between species, in biomedical texts, at sentence level, with a focus on the gut microbiota. The corpus leverages PubTator to annotate species in full-text articles after evaluating different Named Entity Rec...
http://arxiv.org/pdf/2306.08403v1
2023-06-14T09:56:32
2023-06-14T09:56:32
{ "title": [ -0.01789856, -0.013175964, -0.044677734, -0.052093506, -0.0037574768, 0.0031776428, -0.03012085, -0.005748749, 0.013160706, 0.014213562, -0.01928711, -0.004497528, -0.012924194, -0.017425537, -0.044158936, -0.006877899, -0.015716553, ...
http://arxiv.org/abs/2306.08401v1
LiveChat: A Large-Scale Personalized Dialogue Dataset Automatically Constructed from Live Streaming
Jingsheng Gao, Yixin Lian, Ziyi Zhou, Yuzhuo Fu, Baoyuan Wang
cs.CL
Open-domain dialogue systems have made promising progress in recent years. While the state-of-the-art dialogue agents are built upon large-scale text-based social media data and large pre-trained models, there is no guarantee these agents could also perform well in fast-growing scenarios, such as live streaming, due to...
http://arxiv.org/pdf/2306.08401v1
2023-06-14T09:50:06
2023-06-14T09:50:06
{ "title": [ 0.008087158, -0.007156372, -0.04095459, -0.089782715, 0.011360168, 0.00030636787, -0.04324341, 0.012901306, -0.000121712685, 0.054016113, 0.0029144287, 0.015960693, -0.0076789856, -0.011962891, -0.046783447, 0.017623901, 0.0021095276, ...
http://arxiv.org/abs/2306.08400v1
Simple Embodied Language Learning as a Byproduct of Meta-Reinforcement Learning
Evan Zheran Liu, Sahaana Suri, Tong Mu, Allan Zhou, Chelsea Finn
cs.CL, cs.AI, cs.LG
Whereas machine learning models typically learn language by directly training on language tasks (e.g., next-word prediction), language emerges in human children as a byproduct of solving non-language tasks (e.g., acquiring food). Motivated by this observation, we ask: can embodied reinforcement learning (RL) agents als...
http://arxiv.org/pdf/2306.08400v1
2023-06-14T09:48:48
2023-06-14T09:48:48
{ "title": [ -0.022857666, -0.0010061264, -0.020721436, -0.012939453, -0.016693115, -0.013885498, -0.0073547363, -0.02482605, -0.038513184, 0.035491943, -0.012779236, 0.012718201, -0.014602661, -0.02708435, -0.019454956, -0.0079574585, 0.0020446777, ...
http://arxiv.org/abs/2306.08374v1
SpeechGLUE: How Well Can Self-Supervised Speech Models Capture Linguistic Knowledge?
Takanori Ashihara, Takafumi Moriya, Kohei Matsuura, Tomohiro Tanaka, Yusuke Ijima, Taichi Asami, Marc Delcroix, Yukinori Honma
cs.CL, cs.SD, eess.AS
Self-supervised learning (SSL) for speech representation has been successfully applied in various downstream tasks, such as speech and speaker recognition. More recently, speech SSL models have also been shown to be beneficial in advancing spoken language understanding tasks, implying that the SSL models have the poten...
http://arxiv.org/pdf/2306.08374v1
2023-06-14T09:04:29
2023-06-14T09:04:29
{ "title": [ -0.022399902, 0.0013132095, -0.05810547, -0.054870605, -0.011985779, -0.018005371, -0.05569458, 0.0076026917, -0.024124146, 0.031829834, -0.016693115, 0.0022888184, 0.00844574, -0.025863647, -0.030273438, -0.0039634705, -0.0009446144, ...
http://arxiv.org/abs/2306.08373v1
A semantically enhanced dual encoder for aspect sentiment triplet extraction
Baoxing Jiang, Shehui Liang, Peiyu Liu, Kaifang Dong, Hongye Li
cs.CL, cs.AI
Aspect sentiment triplet extraction (ASTE) is a crucial subtask of aspect-based sentiment analysis (ABSA) that aims to comprehensively identify sentiment triplets. Previous research has focused on enhancing ASTE through innovative table-filling strategies. However, these approaches often overlook the multi-perspective ...
http://arxiv.org/pdf/2306.08373v1
2023-06-14T09:04:14
2023-06-14T09:04:14
{ "title": [ 0.014770508, 0.0027656555, -0.027313232, -0.08294678, 0.0076141357, -0.0064697266, -0.034820557, 0.0048294067, 0.050628662, 0.06347656, -0.011962891, 0.007850647, 0.009780884, 0.011253357, -0.09423828, 0.03253174, -0.000767231, 0.031...
http://arxiv.org/abs/2306.08368v1
T5-SR: A Unified Seq-to-Seq Decoding Strategy for Semantic Parsing
Yuntao Li, Zhenpeng Su, Yutian Li, Hanchu Zhang, Sirui Wang, Wei Wu, Yan Zhang
cs.CL, cs.AI
Translating natural language queries into SQLs in a seq2seq manner has attracted much attention recently. However, compared with abstract-syntactic-tree-based SQL generation, seq2seq semantic parsers face much more challenges, including poor quality on schematical information prediction and poor semantic coherence betw...
http://arxiv.org/pdf/2306.08368v1
2023-06-14T08:57:13
2023-06-14T08:57:13
{ "title": [ -0.028778076, -0.012710571, -0.05050659, -0.08880615, 0.0040779114, 0.007621765, -0.075927734, 0.025421143, -0.009750366, 0.029327393, -0.029846191, 0.030456543, 0.0020427704, -0.0079574585, -0.029418945, 0.011070251, -0.01398468, 0....
http://arxiv.org/abs/2306.08350v1
Multi-target Backdoor Attacks for Code Pre-trained Models
Yanzhou Li, Shangqing Liu, Kangjie Chen, Xiaofei Xie, Tianwei Zhang, Yang Liu
cs.CR, cs.AI, cs.CL
Backdoor attacks for neural code models have gained considerable attention due to the advancement of code intelligence. However, most existing works insert triggers into task-specific data for code-related downstream tasks, thereby limiting the scope of attacks. Moreover, the majority of attacks for pre-trained models ...
http://arxiv.org/pdf/2306.08350v1
2023-06-14T08:38:51
2023-06-14T08:38:51
{ "title": [ -0.015350342, -0.010597229, -0.020690918, -0.05102539, 0.0068130493, -0.012260437, -0.01008606, -0.003780365, -0.038970947, 0.021636963, 0.015319824, 0.020645142, -0.0289917, -0.03805542, -0.00818634, -0.00031852722, 0.00009089708, 0...
http://arxiv.org/abs/2306.08329v1
Research on an improved Conformer end-to-end Speech Recognition Model with R-Drop Structure
Weidong Ji, Shijie Zan, Guohui Zhou, Xu Wang
cs.SD, cs.CL, eess.AS
To address the issue of poor generalization ability in end-to-end speech recognition models within deep learning, this study proposes a new Conformer-based speech recognition model called "Conformer-R" that incorporates the R-drop structure. This model combines the Conformer model, which has shown promising results in ...
http://arxiv.org/pdf/2306.08329v1
2023-06-14T08:01:23
2023-06-14T08:01:23
{ "title": [ -0.03466797, 0.016448975, -0.05029297, -0.078430176, 0.014770508, 0.011047363, -0.0602417, 0.015289307, -0.029541016, 0.041503906, -0.0014762878, -0.010375977, 0.0053634644, -0.021850586, -0.036102295, -0.00011175871, -0.0113220215, ...
http://arxiv.org/abs/2306.08315v1
Research on Named Entity Recognition in Improved transformer with R-Drop structure
Weidong Ji, Yousheng Zhang, Guohui Zhou, Xu Wang
cs.CL, cs.AI
To enhance the generalization ability of the model and improve the effectiveness of the transformer for named entity recognition tasks, the XLNet-Transformer-R model is proposed in this paper. The XLNet pre-trained model and the Transformer encoder with relative positional encodings are combined to enhance the model's ...
http://arxiv.org/pdf/2306.08315v1
2023-06-14T07:34:27
2023-06-14T07:34:27
{ "title": [ -0.024978638, 0.013061523, -0.057128906, -0.06185913, 0.027542114, -0.00019586086, -0.032073975, -0.0010375977, -0.030090332, 0.028076172, -0.027801514, 0.008842468, 0.03378296, -0.020309448, -0.0017004013, -0.0010766983, -0.0010509491, ...
http://arxiv.org/abs/2306.08302v2
Unifying Large Language Models and Knowledge Graphs: A Roadmap
Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, Xindong Wu
cs.CL, cs.AI
Large language models (LLMs), such as ChatGPT and GPT4, are making new waves in the field of natural language processing and artificial intelligence, due to their emergent ability and generalizability. However, LLMs are black-box models, which often fall short of capturing and accessing factual knowledge. In contrast, ...
http://arxiv.org/pdf/2306.08302v2
2023-06-20T14:18:49
2023-06-14T07:15:26
{ "title": [ -0.033325195, -0.016143799, -0.05316162, -0.10058594, 0.0043296814, 0.024475098, -0.093933105, 0.023361206, 0.008140564, 0.039520264, -0.009101868, -0.002336502, 0.009132385, -0.005706787, -0.03970337, 0.005584717, -0.0014333725, 0.0...
http://arxiv.org/abs/2306.08290v1
Investigating the dynamics of hand and lips in French Cued Speech using attention mechanisms and CTC-based decoding
Sanjana Sankar, Denis Beautemps, Frédéric Elisei, Olivier Perrotin, Thomas Hueber
cs.CL
Hard of hearing or profoundly deaf people make use of cued speech (CS) as a communication tool to understand spoken language. By delivering cues that are relevant to the phonetic information, CS offers a way to enhance lipreading. In literature, there have been several studies on the dynamics between the hand and the l...
http://arxiv.org/pdf/2306.08290v1
2023-06-14T06:58:07
2023-06-14T06:58:07
{ "title": [ -0.016220093, -0.003206253, -0.014854431, -0.009757996, -0.044708252, 0.0018615723, -0.032409668, -0.011070251, -0.008590698, 0.049713135, -0.030303955, 0.0036830902, 0.024627686, -0.023773193, -0.043884277, -0.010643005, 0.0023269653, ...
http://arxiv.org/abs/2306.10056v1
Generate to Understand for Representation
Changshang Xue, Xiande Zhong, Xiaoqing Liu
cs.CL, cs.IR, 68T50 (Primary) 03B65, 91F20(Secondary), I.7
In recent years, a significant number of high-quality pretrained models have emerged, greatly impacting Natural Language Understanding (NLU), Natural Language Generation (NLG), and Text Representation tasks. Traditionally, these models are pretrained on custom domain corpora and finetuned for specific tasks, resulting ...
http://arxiv.org/pdf/2306.10056v1
2023-06-14T06:00:18
2023-06-14T06:00:18
{ "title": [ 0.04437256, -0.018692017, 0.0005450249, -0.08477783, -0.03668213, -0.010444641, -0.06384277, -0.0034122467, 0.0024909973, 0.016143799, -0.0073013306, -0.044311523, -0.011558533, -0.035003662, 0.00907135, -0.03933716, 0.04345703, -0.0...
http://arxiv.org/abs/2306.08221v1
Contrastive Loss is All You Need to Recover Analogies as Parallel Lines
Narutatsu Ri, Fei-Tzin Lee, Nakul Verma
cs.CL
While static word embedding models are known to represent linguistic analogies as parallel lines in high-dimensional space, the underlying mechanism as to why they result in such geometric structures remains obscure. We find that an elementary contrastive-style method employed over distributional information performs c...
http://arxiv.org/pdf/2306.08221v1
2023-06-14T03:22:10
2023-06-14T03:22:10
{ "title": [ 0.02746582, -0.01424408, -0.0015258789, -0.009750366, -0.035736084, -0.016403198, 0.016464233, -0.014007568, 0.012138367, 0.060577393, -0.008239746, -0.01084137, 0.04525757, -0.028564453, -0.02670288, 0.009384155, 0.0087509155, 0.027...
http://arxiv.org/abs/2306.08193v2
Operationalising Representation in Natural Language Processing
Jacqueline Harding
cs.CL, cs.AI, cs.LG
Despite its centrality in the philosophy of cognitive science, there has been little prior philosophical work engaging with the notion of representation in contemporary NLP practice. This paper attempts to fill that lacuna: drawing on ideas from cognitive science, I introduce a framework for evaluating the representati...
http://arxiv.org/pdf/2306.08193v2
2023-10-08T01:49:38
2023-06-14T01:34:16
{ "title": [ -0.0033740997, -0.0035381317, -0.020889282, -0.06121826, 0.02015686, 0.00831604, -0.020355225, -0.018417358, 0.01600647, 0.026412964, -0.014587402, -0.009895325, -0.0022888184, -0.016616821, -0.05065918, 0.008476257, -0.014564514, 0....
http://arxiv.org/abs/2306.08190v1
Assessing the Effectiveness of GPT-3 in Detecting False Political Statements: A Case Study on the LIAR Dataset
Mars Gokturk Buchholz
cs.CL, I.2.7
The detection of political fake statements is crucial for maintaining information integrity and preventing the spread of misinformation in society. Historically, state-of-the-art machine learning models employed various methods for detecting deceptive statements. These methods include the use of metadata (W. Wang et al...
http://arxiv.org/pdf/2306.08190v1
2023-06-14T01:16:49
2023-06-14T01:16:49
{ "title": [ 0.037017822, 0.024169922, -0.03024292, -0.026428223, 0.027450562, -0.040283203, 0.010726929, -0.040100098, -0.04598999, 0.031707764, -0.016693115, 0.021377563, -0.013076782, -0.014022827, -0.031951904, -0.0088272095, -0.03164673, 0.0...
http://arxiv.org/abs/2306.08189v1
Language models are not naysayers: An analysis of language models on negation benchmarks
Thinh Hung Truong, Timothy Baldwin, Karin Verspoor, Trevor Cohn
cs.CL
Negation has been shown to be a major bottleneck for masked language models, such as BERT. However, whether this finding still holds for larger-sized auto-regressive language models (``LLMs'') has not been studied comprehensively. With the ever-increasing volume of research and applications of LLMs, we take a step back...
http://arxiv.org/pdf/2306.08189v1
2023-06-14T01:16:37
2023-06-14T01:16:37
{ "title": [ -0.010757446, -0.008270264, -0.0001552105, -0.058135986, -0.010467529, 0.012504578, -0.057525635, -0.017562866, 0.011726379, 0.03237915, -0.032592773, 0.0013504028, 0.010635376, -0.0044555664, -0.062347412, 0.006641388, -0.024032593, ...
http://arxiv.org/abs/2306.08162v1
INT2.1: Towards Fine-Tunable Quantized Large Language Models with Error Correction through Low-Rank Adaptation
Yuji Chai, John Gkountouras, Glenn G. Ko, David Brooks, Gu-Yeon Wei
cs.CL, cs.AI, cs.LG
We introduce a method that dramatically reduces fine-tuning VRAM requirements and rectifies quantization errors in quantized Large Language Models. First, we develop an extremely memory-efficient fine-tuning (EMEF) method for quantized models using Low-Rank Adaptation (LoRA), and drawing upon it, we construct an error-...
http://arxiv.org/pdf/2306.08162v1
2023-06-13T22:25:35
2023-06-13T22:25:35
{ "title": [ -0.043518066, -0.0072250366, -0.041870117, -0.044067383, -0.017181396, 0.002368927, -0.075927734, -0.0067214966, -0.0035495758, 0.04916382, -0.0058555603, -0.010444641, 0.012939453, -0.011436462, -0.06213379, 0.0046691895, -0.0036239624,...
http://arxiv.org/abs/2306.08161v2
h2oGPT: Democratizing Large Language Models
Arno Candel, Jon McKinney, Philipp Singer, Pascal Pfeiffer, Maximilian Jeblick, Prithvi Prabhu, Jeff Gambera, Mark Landry, Shivam Bansal, Ryan Chesler, Chun Ming Lee, Marcos V. Conde, Pasha Stetsenko, Olivier Grellier, SriSatish Ambati
cs.CL, cs.AI, cs.HC, cs.IR, cs.LG
Applications built on top of Large Language Models (LLMs) such as GPT-4 represent a revolution in AI due to their human-level capabilities in natural language processing. However, they also pose many significant risks such as the presence of biased, private, or harmful text, and the unauthorized inclusion of copyrighte...
http://arxiv.org/pdf/2306.08161v2
2023-06-16T17:48:22
2023-06-13T22:19:53
{ "title": [ -0.02659607, -0.023391724, -0.043640137, -0.081604004, 0.0072784424, 0.017578125, -0.059265137, 0.011817932, 0.008110046, 0.029907227, -0.006965637, -0.0002092123, 0.010772705, -0.016723633, -0.043762207, 0.022567749, -0.014434814, 0...
http://arxiv.org/abs/2306.08158v3
Survey on Sociodemographic Bias in Natural Language Processing
Vipul Gupta, Pranav Narayanan Venkit, Shomir Wilson, Rebecca J. Passonneau
cs.CL, cs.AI, cs.LG
Deep neural networks often learn unintended bias during training, which might have harmful effects when deployed in real-world settings. This work surveys 214 papers related to sociodemographic bias in natural language processing (NLP). In this study, we aim to provide a more comprehensive understanding of the similari...
http://arxiv.org/pdf/2306.08158v3
2023-08-22T00:55:21
2023-06-13T22:07:54
{ "title": [ -0.00016868114, 0.0036945343, -0.029815674, -0.036590576, -0.0058403015, -0.023132324, -0.012290955, -0.023086548, -0.0070724487, 0.017837524, -0.044311523, -0.003068924, 0.0020923615, -0.016342163, -0.008735657, -0.010894775, -0.0271301...
http://arxiv.org/abs/2306.08133v2
Large-scale Language Model Rescoring on Long-form Data
Tongzhou Chen, Cyril Allauzen, Yinghui Huang, Daniel Park, David Rybach, W. Ronny Huang, Rodrigo Cabrera, Kartik Audhkhasi, Bhuvana Ramabhadran, Pedro J. Moreno, Michael Riley
eess.AS, cs.CL
In this work, we study the impact of Large-scale Language Models (LLM) on Automated Speech Recognition (ASR) of YouTube videos, which we use as a source for long-form ASR. We demonstrate up to 8\% relative reduction in Word Error Eate (WER) on US English (en-us) and code-switched Indian English (en-in) long-form ASR te...
http://arxiv.org/pdf/2306.08133v2
2023-09-05T20:50:24
2023-06-13T20:54:12
{ "title": [ -0.029464722, -0.008575439, -0.04144287, -0.07940674, -0.009437561, 0.0010995865, -0.057800293, 0.025726318, 0.01651001, 0.04888916, -0.022598267, -0.002084732, 0.0039405823, -0.008903503, -0.048858643, 0.010047913, -0.0059013367, 0....
http://arxiv.org/abs/2306.08129v3
AVIS: Autonomous Visual Information Seeking with Large Language Model Agent
Ziniu Hu, Ahmet Iscen, Chen Sun, Kai-Wei Chang, Yizhou Sun, David A Ross, Cordelia Schmid, Alireza Fathi
cs.CV, cs.AI, cs.CL
In this paper, we propose an autonomous information seeking visual question answering framework, AVIS. Our method leverages a Large Language Model (LLM) to dynamically strategize the utilization of external tools and to investigate their outputs, thereby acquiring the indispensable knowledge needed to provide answers t...
http://arxiv.org/pdf/2306.08129v3
2023-11-02T07:54:37
2023-06-13T20:50:22
{ "title": [ -0.01637268, -0.018066406, 0.0012950897, -0.022369385, -0.012237549, 0.0062065125, -0.0021533966, -0.02897644, 0.004383087, 0.035858154, -0.049926758, -0.0032978058, 0.0046157837, 0.00085163116, -0.052886963, -0.013534546, 0.0040664673, ...
http://arxiv.org/abs/2306.08126v1
PersonaPKT: Building Personalized Dialogue Agents via Parameter-efficient Knowledge Transfer
Xu Han, Bin Guo, Yoon Jung, Benjamin Yao, Yu Zhang, Xiaohu Liu, Chenlei Guo
cs.CL, cs.AI
Personalized dialogue agents (DAs) powered by large pre-trained language models (PLMs) often rely on explicit persona descriptions to maintain personality consistency. However, such descriptions may not always be available or may pose privacy concerns. To tackle this bottleneck, we introduce PersonaPKT, a lightweight t...
http://arxiv.org/pdf/2306.08126v1
2023-06-13T20:47:29
2023-06-13T20:47:29
{ "title": [ -0.01210022, -0.020874023, -0.040222168, -0.076293945, 0.0063476562, 0.003255844, -0.03060913, 0.016647339, 0.0003232956, 0.026992798, 0.00894165, 0.02268982, -0.005508423, -0.021148682, -0.03692627, 0.0075950623, -0.016464233, 0.033...
http://arxiv.org/abs/2306.08122v1
Beyond Black Box AI-Generated Plagiarism Detection: From Sentence to Document Level
Mujahid Ali Quidwai, Chunhui Li, Parijat Dube
cs.CL, cs.AI, cs.LG
The increasing reliance on large language models (LLMs) in academic writing has led to a rise in plagiarism. Existing AI-generated text classifiers have limited accuracy and often produce false positives. We propose a novel approach using natural language processing (NLP) techniques, offering quantifiable metrics at bo...
http://arxiv.org/pdf/2306.08122v1
2023-06-13T20:34:55
2023-06-13T20:34:55
{ "title": [ -0.018737793, -0.002363205, -0.024658203, -0.021087646, -0.008590698, -0.010765076, 0.00007015467, -0.032409668, -0.007835388, 0.029220581, -0.032714844, 0.021209717, 0.015380859, -0.00554657, -0.027435303, -0.007545471, -0.02708435, ...
http://arxiv.org/abs/2306.08116v1
CipherSniffer: Classifying Cipher Types
Brendan Artley, Greg Mehdiyev
cs.CL, cs.CR, cs.LG
Ciphers are a powerful tool for encrypting communication. There are many different cipher types, which makes it computationally expensive to solve a cipher using brute force. In this paper, we frame the decryption task as a classification problem. We first create a dataset of transpositions, substitutions, text reversa...
http://arxiv.org/pdf/2306.08116v1
2023-06-13T20:18:24
2023-06-13T20:18:24
{ "title": [ -0.052124023, -0.0070877075, -0.04324341, 0.012428284, -0.029815674, -0.019805908, -0.020065308, -0.024261475, -0.017974854, 0.03540039, 0.00843811, -0.018371582, 0.030807495, -0.048675537, -0.005001068, 0.01071167, 0.0066223145, 0.0...
http://arxiv.org/abs/2306.08107v2
AutoML in the Age of Large Language Models: Current Challenges, Future Opportunities and Risks
Alexander Tornede, Difan Deng, Theresa Eimer, Joseph Giovanelli, Aditya Mohan, Tim Ruhkopf, Sarah Segel, Daphne Theodorakopoulos, Tanja Tornede, Henning Wachsmuth, Marius Lindauer
cs.LG, cs.CL
The fields of both Natural Language Processing (NLP) and Automated Machine Learning (AutoML) have achieved remarkable results over the past years. In NLP, especially Large Language Models (LLMs) have experienced a rapid series of breakthroughs very recently. We envision that the two fields can radically push the bounda...
http://arxiv.org/pdf/2306.08107v2
2023-08-21T15:01:46
2023-06-13T19:51:22
{ "title": [ -0.030029297, -0.018981934, -0.039123535, -0.07495117, -0.003414154, 0.017913818, -0.072021484, 0.0024433136, -0.015945435, 0.027877808, 0.0015859604, 0.007835388, 0.018310547, -0.033172607, -0.016418457, 0.02267456, -0.016159058, 0....
http://arxiv.org/abs/2306.08042v1
FLamE: Few-shot Learning from Natural Language Explanations
Yangqiaoyu Zhou, Yiming Zhang, Chenhao Tan
cs.CL, cs.AI
Natural language explanations have the potential to provide rich information that in principle guides model reasoning. Yet, recent work by Lampinen et al. (2022) has shown limited utility of natural language explanations in improving classification. To effectively learn from explanations, we present FLamE, a two-stage ...
http://arxiv.org/pdf/2306.08042v1
2023-06-13T18:01:46
2023-06-13T18:01:46
{ "title": [ -0.0071487427, -0.038085938, -0.014205933, -0.058776855, -0.020523071, 0.0044822693, -0.027572632, 0.011482239, -0.005256653, 0.0015087128, -0.029251099, 0.02494812, -0.028198242, -0.011398315, -0.019714355, -0.015060425, -0.003156662, ...
http://arxiv.org/abs/2306.07968v1
arXiVeri: Automatic table verification with GPT
Gyungin Shin, Weidi Xie, Samuel Albanie
cs.CL, cs.AI
Without accurate transcription of numerical data in scientific documents, a scientist cannot draw accurate conclusions. Unfortunately, the process of copying numerical data from one paper to another is prone to human error. In this paper, we propose to meet this challenge through the novel task of automatic table verif...
http://arxiv.org/pdf/2306.07968v1
2023-06-13T17:59:57
2023-06-13T17:59:57
{ "title": [ -0.013671875, 0.010139465, -0.04623413, -0.07611084, 0.025390625, 0.012046814, -0.06222534, -0.02218628, 0.010231018, 0.034973145, -0.016693115, 0.009483337, 0.0045814514, -0.013801575, -0.044281006, 0.013389587, -0.013023376, 0.0369...
http://arxiv.org/abs/2306.07952v2
MOFI: Learning Image Representations from Noisy Entity Annotated Images
Wentao Wu, Aleksei Timofeev, Chen Chen, Bowen Zhang, Kun Duan, Shuangning Liu, Yantao Zheng, Jon Shlens, Xianzhi Du, Zhe Gan, Yinfei Yang
cs.CV, cs.CL, cs.LG
We present MOFI, a new vision foundation model designed to learn image representations from noisy entity annotated images. MOFI differs from previous work in two key aspects: ($i$) pre-training data, and ($ii$) training recipe. Regarding data, we introduce a new approach to automatically assign entity labels to images ...
http://arxiv.org/pdf/2306.07952v2
2023-06-24T19:16:28
2023-06-13T17:51:18
{ "title": [ -0.0053977966, 0.008636475, -0.018325806, -0.03982544, 0.018371582, -0.011672974, -0.006931305, -0.021728516, 0.03842163, 0.027572632, -0.03387451, 0.02835083, 0.00182724, -0.010345459, -0.03488159, 0.021713257, 0.012245178, 0.034088...
http://arxiv.org/abs/2306.07951v2
Questioning the Survey Responses of Large Language Models
Ricardo Dominguez-Olmedo, Moritz Hardt, Celestine Mendler-Dünner
cs.CL
As large language models increase in capability, researchers have started to conduct surveys of all kinds on these models with varying scientific motivations. In this work, we examine what we can learn from language models' survey responses on the basis of the well-established American Community Survey (ACS) by the U.S...
http://arxiv.org/pdf/2306.07951v2
2023-10-12T17:34:12
2023-06-13T17:48:27
{ "title": [ -0.03753662, -0.0079956055, -0.026657104, -0.006072998, -0.034332275, -0.017211914, -0.0054092407, -0.01247406, -0.0060806274, 0.01134491, -0.04385376, -0.0209198, -0.041778564, -0.039916992, -0.035461426, -0.031433105, -0.0018434525, ...
http://arxiv.org/abs/2306.07934v1
BoardgameQA: A Dataset for Natural Language Reasoning with Contradictory Information
Mehran Kazemi, Quan Yuan, Deepti Bhatia, Najoung Kim, Xin Xu, Vaiva Imbrasaite, Deepak Ramachandran
cs.CL, cs.AI, cs.LG
Automated reasoning with unstructured natural text is a key requirement for many potential applications of NLP and for developing robust AI systems. Recently, Language Models (LMs) have demonstrated complex reasoning capacities even without any finetuning. However, existing evaluation for automated reasoning assumes ac...
http://arxiv.org/pdf/2306.07934v1
2023-06-13T17:39:20
2023-06-13T17:39:20
{ "title": [ -0.02142334, -0.00409317, -0.041137695, -0.04562378, 0.0031051636, 0.003042221, -0.03579712, -0.011169434, 0.019134521, -0.009536743, -0.057739258, -0.0016679764, -0.01979065, -0.010467529, -0.022613525, -0.022781372, -0.019607544, 0...
http://arxiv.org/abs/2306.07906v1
WebGLM: Towards An Efficient Web-Enhanced Question Answering System with Human Preferences
Xiao Liu, Hanyu Lai, Hao Yu, Yifan Xu, Aohan Zeng, Zhengxiao Du, Peng Zhang, Yuxiao Dong, Jie Tang
cs.CL, cs.AI
We present WebGLM, a web-enhanced question-answering system based on the General Language Model (GLM). Its goal is to augment a pre-trained large language model (LLM) with web search and retrieval capabilities while being efficient for real-world deployments. To achieve this, we develop WebGLM with strategies for the L...
http://arxiv.org/pdf/2306.07906v1
2023-06-13T16:57:53
2023-06-13T16:57:53
{ "title": [ 0.0006532669, -0.0012292862, -0.038238525, -0.066345215, 0.012214661, -0.0027236938, -0.016723633, 0.011520386, 0.023132324, 0.028076172, -0.022872925, -0.00007480383, 0.006713867, -0.0017719269, -0.06866455, 0.0043525696, -0.009109497, ...
http://arxiv.org/abs/2306.07902v1
Massively Multilingual Corpus of Sentiment Datasets and Multi-faceted Sentiment Classification Benchmark
Łukasz Augustyniak, Szymon Woźniak, Marcin Gruza, Piotr Gramacki, Krzysztof Rajda, Mikołaj Morzy, Tomasz Kajdanowicz
cs.CL, cs.AI
Despite impressive advancements in multilingual corpora collection and model training, developing large-scale deployments of multilingual models still presents a significant challenge. This is particularly true for language tasks that are culture-dependent. One such example is the area of multilingual sentiment analysi...
http://arxiv.org/pdf/2306.07902v1
2023-06-13T16:54:13
2023-06-13T16:54:13
{ "title": [ -0.017074585, -0.006259918, -0.05117798, -0.060302734, 0.009170532, -0.008399963, -0.046295166, 0.00052690506, 0.023712158, 0.04937744, -0.024993896, 0.0016069412, 0.0036792755, -0.008323669, -0.050476074, 0.015853882, -0.004623413, ...
http://arxiv.org/abs/2306.07899v1
Artificial Artificial Artificial Intelligence: Crowd Workers Widely Use Large Language Models for Text Production Tasks
Veniamin Veselovsky, Manoel Horta Ribeiro, Robert West
cs.CL, cs.CY
Large language models (LLMs) are remarkable data annotators. They can be used to generate high-fidelity supervised training data, as well as survey and experimental data. With the widespread adoption of LLMs, human gold--standard annotations are key to understanding the capabilities of LLMs and the validity of their re...
http://arxiv.org/pdf/2306.07899v1
2023-06-13T16:46:24
2023-06-13T16:46:24
{ "title": [ -0.034698486, -0.022613525, -0.026504517, -0.01399231, -0.044769287, -0.0042686462, -0.028167725, 0.023468018, 0.010810852, 0.046691895, -0.03451538, -0.008857727, 0.039916992, -0.012138367, -0.00008201599, -0.014083862, 0.014640808, ...
http://arxiv.org/abs/2306.07875v1
ReadProbe: A Demo of Retrieval-Enhanced Large Language Models to Support Lateral Reading
Dake Zhang, Ronak Pradeep
cs.IR, cs.AI, cs.CL, cs.HC
With the rapid growth and spread of online misinformation, people need tools to help them evaluate the credibility and accuracy of online information. Lateral reading, a strategy that involves cross-referencing information with multiple sources, may be an effective approach to achieving this goal. In this paper, we pre...
http://arxiv.org/pdf/2306.07875v1
2023-06-13T16:10:10
2023-06-13T16:10:10
{ "title": [ -0.036743164, -0.014060974, -0.026351929, -0.041900635, -0.01651001, 0.01586914, -0.051513672, -0.013000488, 0.017669678, 0.047729492, -0.027618408, 0.006965637, 0.027954102, -0.002122879, -0.055267334, 0.02293396, -0.0045814514, 0.0...
http://arxiv.org/abs/2306.07848v8
GEmo-CLAP: Gender-Attribute-Enhanced Contrastive Language-Audio Pretraining for Accurate Speech Emotion Recognition
Yu Pan, Yanni Hu, Yuguang Yang, Wen Fei, Jixun Yao, Heng Lu, Lei Ma, Jianjun Zhao
cs.CL, cs.MM, cs.SD, eess.AS
Contrastive cross-modality pretraining has recently exhibited impressive success in diverse fields, whereas there is limited research on their merits in speech emotion recognition (SER). In this paper, we propose GEmo-CLAP, a kind of gender-attribute-enhanced contrastive language-audio pretraining (CLAP) method for SER...
http://arxiv.org/pdf/2306.07848v8
2023-09-13T04:48:23
2023-06-13T15:28:10
{ "title": [ 0.0025348663, -0.012626648, -0.03475952, -0.06378174, -0.0035362244, -0.016189575, -0.03704834, 0.0044403076, 0.008460999, 0.06100464, 0.0074882507, 0.00028514862, 0.062805176, -0.02458191, -0.046325684, 0.028839111, -0.020004272, 0....
http://arxiv.org/abs/2306.07845v1
Adversarial Capsule Networks for Romanian Satire Detection and Sentiment Analysis
Sebastian-Vasile Echim, Răzvan-Alexandru Smădu, Andrei-Marius Avram, Dumitru-Clementin Cercel, Florin Pop
cs.CL
Satire detection and sentiment analysis are intensively explored natural language processing (NLP) tasks that study the identification of the satirical tone from texts and extracting sentiments in relationship with their targets. In languages with fewer research resources, an alternative is to produce artificial exampl...
http://arxiv.org/pdf/2306.07845v1
2023-06-13T15:23:44
2023-06-13T15:23:44
{ "title": [ 0.0020389557, -0.008003235, -0.019500732, -0.007144928, 0.0021381378, -0.00592041, -0.009651184, -0.025558472, -0.009506226, 0.045837402, -0.009773254, -0.034332275, 0.024139404, -0.018707275, -0.066467285, -0.018798828, -0.019088745, ...
http://arxiv.org/abs/2306.08020v1
Curatr: A Platform for Semantic Analysis and Curation of Historical Literary Texts
Susan Leavy, Gerardine Meaney, Karen Wade, Derek Greene
cs.CL, cs.AI
The increasing availability of digital collections of historical and contemporary literature presents a wealth of possibilities for new research in the humanities. The scale and diversity of such collections however, presents particular challenges in identifying and extracting relevant content. This paper presents Cura...
http://arxiv.org/pdf/2306.08020v1
2023-06-13T15:15:31
2023-06-13T15:15:31
{ "title": [ -0.026779175, 0.0019359589, -0.04638672, -0.050933838, 0.018356323, -0.0013551712, -0.013404846, -0.005252838, 0.0132369995, 0.05053711, -0.009651184, 0.035003662, 0.03845215, -0.0053977966, -0.04824829, 0.023803711, -0.004234314, 0....
http://arxiv.org/abs/2306.08018v3
Mol-Instructions: A Large-Scale Biomolecular Instruction Dataset for Large Language Models
Yin Fang, Xiaozhuan Liang, Ningyu Zhang, Kangwei Liu, Rui Huang, Zhuo Chen, Xiaohui Fan, Huajun Chen
q-bio.QM, cs.AI, cs.CE, cs.CL, cs.IR, cs.LG
Large Language Models (LLMs), with their remarkable task-handling capabilities and innovative outputs, have catalyzed significant advancements across a spectrum of fields. However, their proficiency within specialized domains such as biomolecular studies remains limited. To address this challenge, we introduce Mol-Inst...
http://arxiv.org/pdf/2306.08018v3
2023-10-02T15:27:20
2023-06-13T14:35:34
{ "title": [ -0.06536865, -0.03286743, -0.06726074, -0.072387695, -0.013977051, 0.017807007, -0.091674805, 0.04119873, 0.025054932, 0.025131226, -0.0048179626, -0.00881958, 0.020889282, -0.027877808, -0.025970459, 0.034851074, -0.0005135536, 0.04...
http://arxiv.org/abs/2306.07799v1
ChatGPT vs Human-authored Text: Insights into Controllable Text Summarization and Sentence Style Transfer
Dongqi Pu, Vera Demberg
cs.CL, cs.AI, cs.LG
Large-scale language models, like ChatGPT, have garnered significant media attention and stunned the public with their remarkable capacity for generating coherent text from short natural language prompts. In this paper, we aim to conduct a systematic inspection of ChatGPT's performance in two controllable generation ta...
http://arxiv.org/pdf/2306.07799v1
2023-06-13T14:21:35
2023-06-13T14:21:35
{ "title": [ -0.014251709, -0.0048980713, -0.015655518, -0.011238098, -0.040039062, 0.00007998943, -0.034179688, -0.02319336, -0.03781128, 0.051879883, -0.00072050095, 0.053985596, 0.00283432, 0.0048980713, -0.0154800415, 0.022109985, -0.019210815, ...
http://arxiv.org/abs/2306.07797v2
Monolingual and Cross-Lingual Knowledge Transfer for Topic Classification
Dmitry Karpov, Mikhail Burtsev
cs.CL, cs.AI
This article investigates the knowledge transfer from the RuQTopics dataset. This Russian topical dataset combines a large sample number (361,560 single-label, 170,930 multi-label) with extensive class coverage (76 classes). We have prepared this dataset from the "Yandex Que" raw data. By evaluating the RuQTopics - tra...
http://arxiv.org/pdf/2306.07797v2
2023-07-04T14:07:39
2023-06-13T14:19:45
{ "title": [ -0.023742676, -0.007183075, -0.053863525, -0.038513184, 0.011512756, -0.018096924, -0.034362793, -0.022354126, 0.018341064, 0.026489258, -0.01134491, 0.028213501, -0.015052795, -0.0096588135, -0.025054932, 0.0017690659, 0.0014543533, ...
http://arxiv.org/abs/2306.07790v1
NoCoLA: The Norwegian Corpus of Linguistic Acceptability
Matias Jentoft, David Samuel
cs.CL, cs.AI
While there has been a surge of large language models for Norwegian in recent years, we lack any tool to evaluate their understanding of grammaticality. We present two new Norwegian datasets for this task. NoCoLA_class is a supervised binary classification task where the goal is to discriminate between acceptable and n...
http://arxiv.org/pdf/2306.07790v1
2023-06-13T14:11:19
2023-06-13T14:11:19
{ "title": [ -0.017349243, -0.0033836365, -0.034729004, -0.05166626, 0.01600647, 0.0032405853, -0.03048706, -0.010612488, 0.014587402, 0.030593872, -0.034851074, 0.000056803226, 0.030441284, -0.012252808, -0.037139893, 0.016342163, -0.016036987, ...
http://arxiv.org/abs/2306.07786v2
A Cloud-based Machine Learning Pipeline for the Efficient Extraction of Insights from Customer Reviews
Robert Lakatos, Gergo Bogacsovics, Balazs Harangi, Istvan Lakatos, Attila Tiba, Janos Toth, Marianna Szabo, Andras Hajdu
cs.CL, cs.AI
The efficiency of natural language processing has improved dramatically with the advent of machine learning models, particularly neural network-based solutions. However, some tasks are still challenging, especially when considering specific domains. In this paper, we present a cloud-based system that can extract insigh...
http://arxiv.org/pdf/2306.07786v2
2023-06-18T10:56:14
2023-06-13T14:07:52
{ "title": [ -0.026992798, -0.023513794, -0.04309082, -0.05429077, -0.020126343, -0.01966858, -0.03479004, 0.010726929, -0.027114868, 0.03515625, -0.058044434, 0.030914307, -0.0006070137, 0.007572174, -0.038269043, -0.020431519, 0.014961243, 0.01...
http://arxiv.org/abs/2306.07764v1
Tokenization with Factorized Subword Encoding
David Samuel, Lilja Øvrelid
cs.CL, cs.AI
In recent years, language models have become increasingly larger and more complex. However, the input representations for these models continue to rely on simple and greedy subword tokenization methods. In this paper, we propose a novel tokenization method that factorizes subwords onto discrete triplets using a VQ-VAE ...
http://arxiv.org/pdf/2306.07764v1
2023-06-13T13:27:34
2023-06-13T13:27:34
{ "title": [ -0.039001465, -0.011276245, -0.042144775, -0.054351807, -0.013832092, 0.0061416626, -0.04800415, -0.00012922287, -0.0154953, 0.043182373, -0.031280518, 0.005393982, 0.022155762, -0.028961182, -0.012252808, 0.010864258, -0.010284424, ...
http://arxiv.org/abs/2306.07763v1
NAVER LABS Europe's Multilingual Speech Translation Systems for the IWSLT 2023 Low-Resource Track
Edward Gow-Smith, Alexandre Berard, Marcely Zanon Boito, Ioan Calapodescu
cs.CL
This paper presents NAVER LABS Europe's systems for Tamasheq-French and Quechua-Spanish speech translation in the IWSLT 2023 Low-Resource track. Our work attempts to maximize translation quality in low-resource settings using multilingual parameter-efficient solutions that leverage strong pre-trained models. Our primar...
http://arxiv.org/pdf/2306.07763v1
2023-06-13T13:22:30
2023-06-13T13:22:30
{ "title": [ -0.030685425, -0.015457153, -0.02947998, -0.019042969, -0.040771484, 0.016296387, -0.047424316, 0.029281616, -0.027297974, 0.070617676, 0.015579224, 0.0032863617, 0.044158936, -0.033691406, 0.0053253174, -0.0028972626, 0.007282257, 0...
http://arxiv.org/abs/2306.08012v1
A Novel Scheme to classify Read and Spontaneous Speech
Sunil Kumar Kopparapu
cs.SD, cs.CL, eess.AS
The COVID-19 pandemic has led to an increased use of remote telephonic interviews, making it important to distinguish between scripted and spontaneous speech in audio recordings. In this paper, we propose a novel scheme for identifying read and spontaneous speech. Our approach uses a pre-trained DeepSpeech audio-to-alp...
http://arxiv.org/pdf/2306.08012v1
2023-06-13T11:16:52
2023-06-13T11:16:52
{ "title": [ -0.018692017, 0.0056648254, -0.042938232, -0.052490234, -0.0024204254, -0.009216309, -0.05419922, 0.0023155212, -0.016601562, 0.025878906, -0.0043792725, 0.0042495728, -0.029632568, -0.02027893, -0.040374756, -0.006336212, -0.014556885, ...
http://arxiv.org/abs/2306.07691v1
StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models
Yinghao Aaron Li, Cong Han, Vinay S. Raghavan, Gavin Mischler, Nima Mesgarani
eess.AS, cs.AI, cs.CL, cs.LG, cs.SD
In this paper, we present StyleTTS 2, a text-to-speech (TTS) model that leverages style diffusion and adversarial training with large speech language models (SLMs) to achieve human-level TTS synthesis. StyleTTS 2 differs from its predecessor by modeling styles as a latent random variable through diffusion models to gen...
http://arxiv.org/pdf/2306.07691v1
2023-06-13T11:04:43
2023-06-13T11:04:43
{ "title": [ -0.014709473, 0.0016365051, -0.045013428, -0.044525146, 0.0052108765, -0.0043411255, -0.043060303, 0.002117157, -0.014007568, 0.041503906, 0.0016078949, 0.019805908, 0.025314331, -0.02607727, -0.010261536, 0.013130188, -0.009033203, ...
http://arxiv.org/abs/2306.07664v1
Rethink the Effectiveness of Text Data Augmentation: An Empirical Analysis
Zhengxiang Shi, Aldo Lipani
cs.CL, cs.AI, cs.LG
In recent years, language models (LMs) have made remarkable progress in advancing the field of natural language processing (NLP). However, the impact of data augmentation (DA) techniques on the fine-tuning (FT) performance of these LMs has been a topic of ongoing debate. In this study, we evaluate the effectiveness of ...
http://arxiv.org/pdf/2306.07664v1
2023-06-13T10:14:58
2023-06-13T10:14:58
{ "title": [ 0.005405426, -0.005153656, -0.038116455, -0.035491943, 0.01651001, -0.015472412, -0.008003235, -0.022033691, -0.006942749, 0.02961731, -0.031829834, 0.038208008, -0.008956909, -0.00009846687, -0.040618896, -0.0015201569, -0.028060913, ...
http://arxiv.org/abs/2306.07656v1
Is Anisotropy Inherent to Transformers?
Nathan Godey, Éric de la Clergerie, Benoît Sagot
cs.CL
The representation degeneration problem is a phenomenon that is widely observed among self-supervised learning methods based on Transformers. In NLP, it takes the form of anisotropy, a singular property of hidden representations which makes them unexpectedly close to each other in terms of angular distance (cosine-simi...
http://arxiv.org/pdf/2306.07656v1
2023-06-13T09:54:01
2023-06-13T09:54:01
{ "title": [ 0.034423828, 0.009353638, 0.012229919, 0.02368164, -0.019699097, -0.0039100647, 0.014709473, -0.046844482, -0.019454956, 0.055541992, 0.0118255615, 0.0098724365, 0.014205933, -0.03149414, 0.008232117, 0.022994995, 0.026657104, 0.0603...
http://arxiv.org/abs/2306.07650v1
Modality Adaption or Regularization? A Case Study on End-to-End Speech Translation
Yuchen Han, Chen Xu, Tong Xiao, Jingbo Zhu
cs.CL, cs.SD, eess.AS
Pre-training and fine-tuning is a paradigm for alleviating the data scarcity problem in end-to-end speech translation (E2E ST). The commonplace "modality gap" between speech and text data often leads to inconsistent inputs between pre-training and fine-tuning. However, we observe that this gap occurs in the early stage...
http://arxiv.org/pdf/2306.07650v1
2023-06-13T09:42:48
2023-06-13T09:42:48
{ "title": [ -0.028900146, 0.004058838, -0.029632568, -0.048187256, 0.003540039, -0.0007119179, -0.0335083, 0.014167786, -0.025482178, 0.054992676, 0.016021729, 0.0034122467, 0.03074646, -0.036834717, -0.039245605, 0.023269653, -0.0045318604, 0.0...
http://arxiv.org/abs/2306.07636v1
Hybrid lemmatization in HuSpaCy
Péter Berkecz, György Orosz, Zsolt Szántó, Gergő Szabó, Richárd Farkas
cs.CL, 68T50, I.2.7
Lemmatization is still not a trivial task for morphologically rich languages. Previous studies showed that hybrid architectures usually work better for these languages and can yield great results. This paper presents a hybrid lemmatizer utilizing both a neural model, dictionaries and hand-crafted rules. We introduce a ...
http://arxiv.org/pdf/2306.07636v1
2023-06-13T09:15:40
2023-06-13T09:15:40
{ "title": [ 0.022964478, 0.010498047, -0.03274536, -0.049041748, -0.035858154, -0.008422852, -0.06011963, -0.013305664, -0.045532227, 0.06149292, -0.025115967, -0.017089844, 0.064697266, -0.028289795, 0.022415161, -0.0036354065, -0.017242432, 0....
http://arxiv.org/abs/2306.07629v2
SqueezeLLM: Dense-and-Sparse Quantization
Sehoon Kim, Coleman Hooper, Amir Gholami, Zhen Dong, Xiuyu Li, Sheng Shen, Michael W. Mahoney, Kurt Keutzer
cs.CL, cs.LG
Generative Large Language Models (LLMs) have demonstrated remarkable results for a wide range of tasks. However, deploying these models for inference has been a significant challenge due to their unprecedented resource requirements. This has forced existing deployment frameworks to use multi-GPU inference pipelines, wh...
http://arxiv.org/pdf/2306.07629v2
2023-10-04T22:40:01
2023-06-13T08:57:54
{ "title": [ -0.050872803, -0.00016474724, -0.043121338, -0.035491943, 0.012756348, 0.009895325, -0.02116394, -0.012939453, 0.013839722, 0.045959473, -0.0048332214, 0.0140686035, 0.01235199, -0.009765625, -0.041656494, 0.02268982, 0.0206604, -0.0...
http://arxiv.org/abs/2306.07622v2
Human-Like Intuitive Behavior and Reasoning Biases Emerged in Language Models -- and Disappeared in GPT-4
Thilo Hagendorff, Sarah Fabi
cs.CL, cs.AI, cs.LG
Large language models (LLMs) are currently at the forefront of intertwining AI systems with human communication and everyday life. Therefore, it is of great importance to evaluate their emerging abilities. In this study, we show that LLMs, most notably GPT-3, exhibit behavior that strikingly resembles human-like intuit...
http://arxiv.org/pdf/2306.07622v2
2023-08-18T17:33:15
2023-06-13T08:43:13
{ "title": [ 0.003534317, -0.009521484, -0.009773254, 0.034606934, -0.040893555, -0.04888916, 0.03768921, -0.014678955, 0.009307861, 0.033813477, -0.025436401, -0.018005371, 0.016525269, -0.007904053, -0.041534424, -0.02708435, -0.025756836, 0.04...
http://arxiv.org/abs/2306.07621v1
Rank-Aware Negative Training for Semi-Supervised Text Classification
Ahmed Murtadha, Shengfeng Pan, Wen Bo, Jianlin Su, Xinxin Cao, Wenze Zhang, Yunfeng Liu
cs.CL
Semi-supervised text classification-based paradigms (SSTC) typically employ the spirit of self-training. The key idea is to train a deep classifier on limited labeled texts and then iteratively predict the unlabeled texts as their pseudo-labels for further training. However, the performance is largely affected by the a...
http://arxiv.org/pdf/2306.07621v1
2023-06-13T08:41:36
2023-06-13T08:41:36
{ "title": [ 0.019119263, 0.0008840561, -0.015655518, -0.038391113, 0.015655518, -0.012573242, 0.006793976, -0.019424438, 0.019058228, 0.01940918, -0.02545166, 0.014602661, 0.008468628, -0.018249512, -0.042144775, -0.013755798, -0.019119263, 0.02...
http://arxiv.org/abs/2306.07610v1
Soft Language Clustering for Multilingual Model Pre-training
Jiali Zeng, Yufan Jiang, Yongjing Yin, Yi Jing, Fandong Meng, Binghuai Lin, Yunbo Cao, Jie Zhou
cs.CL
Multilingual pre-trained language models have demonstrated impressive (zero-shot) cross-lingual transfer abilities, however, their performance is hindered when the target language has distant typology from source languages or when pre-training data is limited in size. In this paper, we propose XLM-P, which contextually...
http://arxiv.org/pdf/2306.07610v1
2023-06-13T08:08:08
2023-06-13T08:08:08
{ "title": [ -0.029663086, -0.013900757, -0.027496338, -0.05078125, -0.0061416626, 0.0126953125, -0.055236816, 0.0020160675, -0.014785767, 0.04373169, -0.008964539, 0.005264282, 0.01512146, -0.017028809, -0.01739502, 0.00096797943, 0.005821228, 0...
http://arxiv.org/abs/2306.07597v1
Question Decomposition Tree for Answering Complex Questions over Knowledge Bases
Xiang Huang, Sitao Cheng, Yiheng Shu, Yuheng Bao, Yuzhong Qu
cs.CL
Knowledge base question answering (KBQA) has attracted a lot of interest in recent years, especially for complex questions which require multiple facts to answer. Question decomposition is a promising way to answer complex questions. Existing decomposition methods split the question into sub-questions according to a si...
http://arxiv.org/pdf/2306.07597v1
2023-06-13T07:44:29
2023-06-13T07:44:29
{ "title": [ -0.019882202, -0.008079529, -0.037353516, -0.07647705, -0.002313614, 0.020599365, -0.04458618, 0.012145996, 0.037902832, 0.022369385, -0.034820557, 0.00598526, 0.0051078796, -0.013717651, -0.04434204, 0.009086609, 0.0124435425, 0.009...
http://arxiv.org/abs/2306.07567v2
Large Language Models Sometimes Generate Purely Negatively-Reinforced Text
Fabien Roger
cs.LG, cs.CL
When using adversarial training, it is common practice to train against the most egregious failures. However, this might imply using examples with sensitive information (such as leaked passwords or security vulnerabilities) as training data. One might assume that language models trained with gradient descent never gene...
http://arxiv.org/pdf/2306.07567v2
2023-06-16T16:23:21
2023-06-13T06:40:37
{ "title": [ 0.02519226, -0.02381897, 0.015144348, -0.029663086, -0.053771973, 0.0053977966, -0.016174316, -0.013168335, 0.0012187958, 0.03692627, -0.056732178, -0.005393982, 0.023605347, -0.022994995, 0.015159607, -0.01852417, -0.02029419, 0.067...
http://arxiv.org/abs/2306.08000v1
Improving Zero-Shot Detection of Low Prevalence Chest Pathologies using Domain Pre-trained Language Models
Aakash Mishra, Rajat Mittal, Christy Jestin, Kostas Tingos, Pranav Rajpurkar
physics.med-ph, cs.CL, cs.CV, cs.LG, eess.IV
Recent advances in zero-shot learning have enabled the use of paired image-text data to replace structured labels, replacing the need for expert annotated datasets. Models such as CLIP-based CheXzero utilize these advancements in the domain of chest X-ray interpretation. We hypothesize that domain pre-trained models su...
http://arxiv.org/pdf/2306.08000v1
2023-06-13T06:26:54
2023-06-13T06:26:54
{ "title": [ 0.017959595, -0.005554199, -0.035186768, -0.07696533, -0.010734558, -0.009284973, -0.052856445, -0.00048613548, 0.009277344, 0.027038574, -0.021118164, 0.005783081, -0.007888794, -0.032592773, -0.03894043, -0.0042915344, 0.0064430237, ...
http://arxiv.org/abs/2306.07554v1
HAUSER: Towards Holistic and Automatic Evaluation of Simile Generation
Qianyu He, Yikai Zhang, Jiaqing Liang, Yuncheng Huang, Yanghua Xiao, Yunwen Chen
cs.CL
Similes play an imperative role in creative writing such as story and dialogue generation. Proper evaluation metrics are like a beacon guiding the research of simile generation (SG). However, it remains under-explored as to what criteria should be considered, how to quantify each criterion into metrics, and whether the...
http://arxiv.org/pdf/2306.07554v1
2023-06-13T06:06:01
2023-06-13T06:06:01
{ "title": [ -0.015129089, -0.008888245, -0.0209198, -0.062072754, 0.010040283, 0.016921997, -0.033203125, 0.001663208, 0.013725281, 0.050598145, -0.0025539398, 0.014434814, 0.019515991, -0.010650635, -0.04953003, 0.020614624, -0.03338623, 0.0679...
http://arxiv.org/abs/2306.14903v1
Detect Depression from Social Networks with Sentiment Knowledge Sharing
Yan Shi, Yao Tian, Chengwei Tong, Chunyan Zhu, Qianqian Li, Mengzhu Zhang, Wei Zhao, Yong Liao, Pengyuan Zhou
cs.CL, cs.AI, cs.SI
Social network plays an important role in propagating people's viewpoints, emotions, thoughts, and fears. Notably, following lockdown periods during the COVID-19 pandemic, the issue of depression has garnered increasing attention, with a significant portion of individuals resorting to social networks as an outlet for e...
http://arxiv.org/pdf/2306.14903v1
2023-06-13T05:16:18
2023-06-13T05:16:18
{ "title": [ 0.014862061, -0.0023479462, -0.017715454, -0.061035156, 0.010848999, 0.00097227097, -0.02760315, -0.0056266785, -0.0062828064, 0.04940796, -0.035614014, 0.0022125244, 0.011726379, -0.015144348, -0.05456543, -0.005126953, 0.01007843, ...
http://arxiv.org/abs/2306.07536v1
TART: A plug-and-play Transformer module for task-agnostic reasoning
Kush Bhatia, Avanika Narayan, Christopher De Sa, Christopher Ré
cs.LG, cs.AI, cs.CL
Large language models (LLMs) exhibit in-context learning abilities which enable the same model to perform several tasks without any task-specific training. In contrast, traditional adaptation approaches, such as fine-tuning, modify the underlying models for each specific task. In-context learning, however, consistently...
http://arxiv.org/pdf/2306.07536v1
2023-06-13T04:37:00
2023-06-13T04:37:00
{ "title": [ -0.01991272, -0.015312195, -0.05718994, -0.079711914, 0.0076942444, 0.00724411, -0.05328369, 0.017425537, -0.002784729, 0.009727478, -0.016052246, -0.0018386841, -0.010643005, -0.020996094, -0.021026611, -0.012527466, -0.0012664795, ...
http://arxiv.org/abs/2306.07512v1
Noisy Positive-Unlabeled Learning with Self-Training for Speculative Knowledge Graph Reasoning
Ruijie Wang, Baoyu Li, Yichen Lu, Dachun Sun, Jinning Li, Yuchen Yan, Shengzhong Liu, Hanghang Tong, Tarek F. Abdelzaher
cs.LG, cs.AI, cs.CL, cs.SI
This paper studies speculative reasoning task on real-world knowledge graphs (KG) that contain both \textit{false negative issue} (i.e., potential true facts being excluded) and \textit{false positive issue} (i.e., unreliable or outdated facts being included). State-of-the-art methods fall short in the speculative reas...
http://arxiv.org/pdf/2306.07512v1
2023-06-13T02:43:21
2023-06-13T02:43:21
{ "title": [ -0.022720337, 0.013000488, -0.047576904, -0.07470703, -0.0019626617, 0.0055274963, -0.0602417, 0.013366699, -0.017425537, 0.014205933, -0.012168884, 0.0088272095, -0.041046143, -0.026367188, -0.024551392, -0.021865845, 0.016998291, 0...
http://arxiv.org/abs/2306.07500v1
Adding guardrails to advanced chatbots
Yanchen Wang, Lisa Singh
cs.CY, cs.AI, cs.CL
Generative AI models continue to become more powerful. The launch of ChatGPT in November 2022 has ushered in a new era of AI. ChatGPT and other similar chatbots have a range of capabilities, from answering student homework questions to creating music and art. There are already concerns that humans may be replaced by ch...
http://arxiv.org/pdf/2306.07500v1
2023-06-13T02:23:04
2023-06-13T02:23:04
{ "title": [ 0.057525635, -0.028884888, 0.008239746, -0.057769775, -0.035888672, 0.0041236877, 0.012031555, 0.01448822, -0.046325684, 0.04925537, -0.03765869, 0.02961731, 0.018692017, -0.0152282715, 0.024169922, -0.015296936, 0.020568848, 0.01649...
http://arxiv.org/abs/2306.07499v1
Improving Opinion-based Question Answering Systems Through Label Error Detection and Overwrite
Xiao Yang, Ahmed K. Mohamed, Shashank Jain, Stanislav Peshterliev, Debojeet Chatterjee, Hanwen Zha, Nikita Bhalla, Gagan Aneja, Pranab Mohanty
cs.CL, cs.AI, cs.LG
Label error is a ubiquitous problem in annotated data. Large amounts of label error substantially degrades the quality of deep learning models. Existing methods to tackle the label error problem largely focus on the classification task, and either rely on task specific architecture or require non-trivial additional com...
http://arxiv.org/pdf/2306.07499v1
2023-06-13T02:20:58
2023-06-13T02:20:58
{ "title": [ 0.01965332, -0.012054443, -0.016235352, -0.0440979, 0.0076828003, -0.012756348, -0.0042762756, -0.0060272217, 0.02508545, 0.009269714, -0.032806396, -0.008140564, -0.00048041344, -0.01184082, -0.051452637, -0.012237549, -0.0137786865, ...
http://arxiv.org/abs/2306.07486v1
Knowledge-Prompted Estimator: A Novel Approach to Explainable Machine Translation Assessment
Hao Yang, Min Zhang, Shimin Tao, Minghan Wang, Daimeng Wei, Yanfei Jiang
cs.CL
Cross-lingual Machine Translation (MT) quality estimation plays a crucial role in evaluating translation performance. GEMBA, the first MT quality assessment metric based on Large Language Models (LLMs), employs one-step prompting to achieve state-of-the-art (SOTA) in system-level MT quality estimation; however, it lack...
http://arxiv.org/pdf/2306.07486v1
2023-06-13T01:18:32
2023-06-13T01:18:32
{ "title": [ -0.044433594, -0.008102417, 0.010238647, -0.047210693, -0.03173828, 0.017410278, -0.049468994, -0.0050354004, 0.0018424988, -0.0010757446, -0.029205322, 0.0014829636, 0.0014772415, -0.02645874, -0.033569336, -0.0064811707, -0.0036792755,...