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/2305.11038v3
Learning In-context Learning for Named Entity Recognition
Jiawei Chen, Yaojie Lu, Hongyu Lin, Jie Lou, Wei Jia, Dai Dai, Hua Wu, Boxi Cao, Xianpei Han, Le Sun
cs.CL
Named entity recognition in real-world applications suffers from the diversity of entity types, the emergence of new entity types, and the lack of high-quality annotations. To address the above problems, this paper proposes an in-context learning-based NER approach, which can effectively inject in-context NER ability i...
http://arxiv.org/pdf/2305.11038v3
2023-05-26T05:44:00
2023-05-18T15:31:34
{ "title": [ -0.0118637085, -0.0025691986, -0.048675537, -0.074035645, 0.0317688, -0.008277893, -0.009986877, 0.01537323, 0.012832642, 0.02696228, -0.049041748, 0.018753052, 0.008613586, -0.0006456375, -0.028320312, -0.0023326874, -0.008453369, 0...
http://arxiv.org/abs/2305.11034v1
Trading Syntax Trees for Wordpieces: Target-oriented Opinion Words Extraction with Wordpieces and Aspect Enhancement
Samuel Mensah, Kai Sun, Nikolaos Aletras
cs.CL
State-of-the-art target-oriented opinion word extraction (TOWE) models typically use BERT-based text encoders that operate on the word level, along with graph convolutional networks (GCNs) that incorporate syntactic information extracted from syntax trees. These methods achieve limited gains with GCNs and have difficul...
http://arxiv.org/pdf/2305.11034v1
2023-05-18T15:22:00
2023-05-18T15:22:00
{ "title": [ -0.01033783, -0.023651123, -0.031311035, -0.06518555, -0.020492554, 0.005531311, -0.035064697, 0.011016846, 0.027786255, 0.025436401, -0.03982544, -0.002609253, -0.016952515, 0.0019216537, -0.07800293, -0.018371582, -0.010627747, 0.0...
http://arxiv.org/abs/2305.11029v2
Uncertainty Guided Label Denoising for Document-level Distant Relation Extraction
Qi Sun, Kun Huang, Xiaocui Yang, Pengfei Hong, Kun Zhang, Soujanya Poria
cs.CL, cs.AI
Document-level relation extraction (DocRE) aims to infer complex semantic relations among entities in a document. Distant supervision (DS) is able to generate massive auto-labeled data, which can improve DocRE performance. Recent works leverage pseudo labels generated by the pre-denoising model to reduce noise in DS da...
http://arxiv.org/pdf/2305.11029v2
2023-05-26T08:23:43
2023-05-18T15:15:56
{ "title": [ -0.022659302, 0.015975952, -0.07196045, -0.06561279, 0.0037212372, -0.017654419, -0.05053711, -0.00031852722, 0.02708435, 0.001709938, -0.03604126, 0.003490448, -0.00017762184, -0.021133423, -0.03274536, -0.0184021, -0.003835678, 0.0...
http://arxiv.org/abs/2305.11023v1
Generalized Multiple Intent Conditioned Slot Filling
Harshil Shah, Arthur Wilcke, Marius Cobzarenco, Cristi Cobzarenco, Edward Challis, David Barber
cs.CL
Natural language understanding includes the tasks of intent detection (identifying a user's objectives) and slot filling (extracting the entities relevant to those objectives). Prior slot filling methods assume that each intent type cannot occur more than once within a message, however this is often not a valid assumpt...
http://arxiv.org/pdf/2305.11023v1
2023-05-18T15:04:52
2023-05-18T15:04:52
{ "title": [ 0.023208618, 0.01927185, -0.03414917, -0.054229736, 0.010543823, -0.007347107, 0.00009018183, -0.026443481, -0.02520752, 0.037506104, -0.03640747, 0.0131073, -0.0019779205, -0.02961731, -0.005279541, -0.02017212, -0.017074585, 0.0163...
http://arxiv.org/abs/2305.11016v1
Silver Syntax Pre-training for Cross-Domain Relation Extraction
Elisa Bassignana, Filip Ginter, Sampo Pyysalo, Rob van der Goot, Barbara Plank
cs.CL
Relation Extraction (RE) remains a challenging task, especially when considering realistic out-of-domain evaluations. One of the main reasons for this is the limited training size of current RE datasets: obtaining high-quality (manually annotated) data is extremely expensive and cannot realistically be repeated for eac...
http://arxiv.org/pdf/2305.11016v1
2023-05-18T14:49:19
2023-05-18T14:49:19
{ "title": [ 0.0064086914, -0.000626564, -0.06732178, -0.09777832, 0.033233643, -0.015319824, -0.040740967, 0.0007505417, 0.007545471, 0.0154953, 0.0023708344, 0.000078201294, 0.00062942505, -0.029022217, -0.024993896, 0.005718231, -0.010734558, ...
http://arxiv.org/abs/2305.11013v1
FunASR: A Fundamental End-to-End Speech Recognition Toolkit
Zhifu Gao, Zerui Li, Jiaming Wang, Haoneng Luo, Xian Shi, Mengzhe Chen, Yabin Li, Lingyun Zuo, Zhihao Du, Zhangyu Xiao, Shiliang Zhang
cs.SD, cs.CL, eess.AS
This paper introduces FunASR, an open-source speech recognition toolkit designed to bridge the gap between academic research and industrial applications. FunASR offers models trained on large-scale industrial corpora and the ability to deploy them in applications. The toolkit's flagship model, Paraformer, is a non-auto...
http://arxiv.org/pdf/2305.11013v1
2023-05-18T14:45:09
2023-05-18T14:45:09
{ "title": [ -0.019363403, -0.0075950623, -0.054016113, -0.08709717, 0.025527954, 0.007980347, -0.046173096, 0.02357483, -0.01121521, 0.036865234, -0.01096344, 0.0053253174, 0.0007429123, -0.01991272, -0.037841797, 0.00819397, -0.004512787, 0.045...
http://arxiv.org/abs/2305.11004v2
Taxonomy Completion with Probabilistic Scorer via Box Embedding
Wei Xue, Yongliang Shen, Wenqi Ren, Jietian Guo, Shiliang Pu, Weiming Lu
cs.CL
Taxonomy completion, a task aimed at automatically enriching an existing taxonomy with new concepts, has gained significant interest in recent years. Previous works have introduced complex modules, external information, and pseudo-leaves to enrich the representation and unify the matching process of attachment and inse...
http://arxiv.org/pdf/2305.11004v2
2023-05-19T07:48:49
2023-05-18T14:34:58
{ "title": [ 0.012588501, 0.0067710876, -0.05014038, -0.09710693, 0.023803711, 0.0031871796, -0.046417236, 0.005718231, 0.014762878, -0.005393982, -0.031097412, 0.01777649, -0.022949219, -0.01612854, -0.02659607, -0.015403748, -0.016174316, 0.030...
http://arxiv.org/abs/2305.11000v2
SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities
Dong Zhang, Shimin Li, Xin Zhang, Jun Zhan, Pengyu Wang, Yaqian Zhou, Xipeng Qiu
cs.CL
Multi-modal large language models are regarded as a crucial step towards Artificial General Intelligence (AGI) and have garnered significant interest with the emergence of ChatGPT. However, current speech-language models typically adopt the cascade paradigm, preventing inter-modal knowledge transfer. In this paper, we ...
http://arxiv.org/pdf/2305.11000v2
2023-05-19T14:41:16
2023-05-18T14:23:25
{ "title": [ -0.007843018, 0.0051345825, -0.050109863, -0.05505371, 0.011810303, -0.011871338, -0.058746338, -0.008041382, 0.006500244, 0.020050049, -0.017501831, 0.0073928833, 0.018066406, -0.019088745, -0.016204834, 0.011405945, -0.022079468, 0...
http://arxiv.org/abs/2305.10998v2
The Web Can Be Your Oyster for Improving Large Language Models
Junyi Li, Tianyi Tang, Wayne Xin Zhao, Jingyuan Wang, Jian-Yun Nie, Ji-Rong Wen
cs.CL
Large language models (LLMs) encode a large amount of world knowledge. However, as such knowledge is frozen at the time of model training, the models become static and limited by the training data at that time. In order to further improve the capacity of LLMs for knowledge-intensive tasks, we consider augmenting LLMs w...
http://arxiv.org/pdf/2305.10998v2
2023-05-24T09:35:39
2023-05-18T14:20:32
{ "title": [ 0.027999878, -0.02166748, -0.03173828, -0.09857178, -0.021957397, -0.022491455, -0.041503906, 0.04333496, -0.012313843, 0.013374329, -0.036376953, -0.00434494, 0.01991272, -0.035949707, 0.021331787, -0.011383057, 0.015594482, 0.03536...
http://arxiv.org/abs/2305.10992v1
How does the task complexity of masked pretraining objectives affect downstream performance?
Atsuki Yamaguchi, Hiroaki Ozaki, Terufumi Morishita, Gaku Morio, Yasuhiro Sogawa
cs.CL, cs.AI
Masked language modeling (MLM) is a widely used self-supervised pretraining objective, where a model needs to predict an original token that is replaced with a mask given contexts. Although simpler and computationally efficient pretraining objectives, e.g., predicting the first character of a masked token, have recentl...
http://arxiv.org/pdf/2305.10992v1
2023-05-18T14:11:57
2023-05-18T14:11:57
{ "title": [ 0.022033691, 0.027267456, -0.06048584, -0.007572174, 0.017532349, -0.040130615, 0.002840042, -0.012641907, -0.050842285, 0.027740479, 0.04776001, -0.013076782, 0.0027179718, -0.06060791, 0.028213501, -0.007843018, 0.016357422, 0.0224...
http://arxiv.org/abs/2305.10991v1
Less is More! A slim architecture for optimal language translation
Luca Herranz-Celotti, Ermal Rrapaj
cs.CL, cs.AI
The softmax attention mechanism has emerged as a noteworthy development in the field of Artificial Intelligence research, building on the successes of Transformer-based architectures. However, their ever increasing sizes necessitate ever increasing computational memory, that limits their usage. We propose KgV, a sigmoi...
http://arxiv.org/pdf/2305.10991v1
2023-05-18T14:09:52
2023-05-18T14:09:52
{ "title": [ -0.05645752, -0.021743774, -0.022598267, -0.050323486, -0.0011014938, 0.019821167, -0.043792725, 0.007873535, -0.009048462, 0.045684814, -0.0030651093, 0.013656616, 0.010757446, -0.018539429, -0.054351807, 0.012138367, 0.011650085, 0...
http://arxiv.org/abs/2305.10985v1
Multi-CrossRE A Multi-Lingual Multi-Domain Dataset for Relation Extraction
Elisa Bassignana, Filip Ginter, Sampo Pyysalo, Rob van der Goot, Barbara Plank
cs.CL
Most research in Relation Extraction (RE) involves the English language, mainly due to the lack of multi-lingual resources. We propose Multi-CrossRE, the broadest multi-lingual dataset for RE, including 26 languages in addition to English, and covering six text domains. Multi-CrossRE is a machine translated version of ...
http://arxiv.org/pdf/2305.10985v1
2023-05-18T14:01:33
2023-05-18T14:01:33
{ "title": [ -0.00655365, -0.01171875, -0.07299805, -0.074645996, 0.021865845, -0.020553589, -0.03717041, 0.010238647, 0.017349243, 0.030593872, -0.006061554, 0.010467529, 0.00894928, -0.015975952, -0.025970459, 0.021377563, -0.01272583, 0.064331...
http://arxiv.org/abs/2305.10971v2
NollySenti: Leveraging Transfer Learning and Machine Translation for Nigerian Movie Sentiment Classification
Iyanuoluwa Shode, David Ifeoluwa Adelani, Jing Peng, Anna Feldman
cs.CL
Africa has over 2000 indigenous languages but they are under-represented in NLP research due to lack of datasets. In recent years, there have been progress in developing labeled corpora for African languages. However, they are often available in a single domain and may not generalize to other domains. In this paper, we...
http://arxiv.org/pdf/2305.10971v2
2023-08-22T07:25:43
2023-05-18T13:38:36
{ "title": [ -0.013557434, -0.015403748, -0.015640259, -0.004180908, -0.018875122, -0.008834839, -0.0066986084, -0.02418518, -0.00970459, 0.046173096, -0.022003174, 0.0005607605, 0.018463135, -0.005428314, -0.049621582, -0.015365601, 0.0010280609, ...
http://arxiv.org/abs/2305.10951v2
Making More of Little Data: Improving Low-Resource Automatic Speech Recognition Using Data Augmentation
Martijn Bartelds, Nay San, Bradley McDonnell, Dan Jurafsky, Martijn Wieling
cs.CL, eess.AS
The performance of automatic speech recognition (ASR) systems has advanced substantially in recent years, particularly for languages for which a large amount of transcribed speech is available. Unfortunately, for low-resource languages, such as minority languages, regional languages or dialects, ASR performance general...
http://arxiv.org/pdf/2305.10951v2
2023-05-19T02:15:41
2023-05-18T13:20:38
{ "title": [ 0.0020217896, -0.0036888123, -0.06604004, -0.081604004, 0.01133728, -0.015686035, -0.039398193, 0.01939392, -0.01550293, 0.035705566, -0.042785645, 0.028640747, -0.0013608932, -0.0031204224, -0.039245605, -0.0035190582, 0.0038585663, ...
http://arxiv.org/abs/2305.10930v3
On the Off-Target Problem of Zero-Shot Multilingual Neural Machine Translation
Liang Chen, Shuming Ma, Dongdong Zhang, Furu Wei, Baobao Chang
cs.CL, cs.AI
While multilingual neural machine translation has achieved great success, it suffers from the off-target issue, where the translation is in the wrong language. This problem is more pronounced on zero-shot translation tasks. In this work, we find that failing in encoding discriminative target language signal will lead t...
http://arxiv.org/pdf/2305.10930v3
2023-06-02T02:58:03
2023-05-18T12:43:31
{ "title": [ -0.015655518, -0.010559082, 0.008644104, -0.02394104, -0.009307861, 0.011741638, -0.016403198, 0.000094890594, -0.008201599, 0.060394287, 0.018844604, 0.00027132034, 0.015396118, -0.016830444, -0.0446167, 0.028396606, -0.010116577, 0...
http://arxiv.org/abs/2305.10928v1
Multilingual Event Extraction from Historical Newspaper Adverts
Nadav Borenstein, Natalia da Silva Perez, Isabelle Augenstein
cs.CL, cs.LG
NLP methods can aid historians in analyzing textual materials in greater volumes than manually feasible. Developing such methods poses substantial challenges though. First, acquiring large, annotated historical datasets is difficult, as only domain experts can reliably label them. Second, most available off-the-shelf N...
http://arxiv.org/pdf/2305.10928v1
2023-05-18T12:40:41
2023-05-18T12:40:41
{ "title": [ 0.015914917, 0.0140686035, -0.025756836, -0.062561035, 0.041534424, -0.009819031, -0.016555786, -0.014503479, 0.019760132, 0.032226562, 0.00762558, 0.026763916, -0.001405716, -0.010360718, -0.041107178, 0.031219482, -0.012489319, 0.0...
http://arxiv.org/abs/2305.10927v3
Causal Document-Grounded Dialogue Pre-training
Yingxiu Zhao, Bowen Yu, Haiyang Yu, Bowen Li, Jinyang Li, Chao Wang, Fei Huang, Yongbin Li, Nevin L. Zhang
cs.CL
The goal of document-grounded dialogue (DocGD) is to generate a response by grounding the evidence in a supporting document in accordance with the dialogue context. This process involves four variables that are causally connected. Recently, task-specific pre-training has greatly boosted performances on many downstream ...
http://arxiv.org/pdf/2305.10927v3
2023-11-05T15:26:49
2023-05-18T12:39:25
{ "title": [ 0.009277344, 0.0020694733, -0.06665039, -0.08874512, -0.01777649, -0.017715454, -0.06390381, 0.018814087, -0.029449463, 0.017868042, -0.009132385, -0.011756897, 0.028625488, -0.05496216, 0.04473877, -0.027679443, -0.0039901733, 0.051...
http://arxiv.org/abs/2305.10923v1
Query Performance Prediction: From Ad-hoc to Conversational Search
Chuan Meng, Negar Arabzadeh, Mohammad Aliannejadi, Maarten de Rijke
cs.IR, cs.CL, cs.LG, H.3.3
Query performance prediction (QPP) is a core task in information retrieval. The QPP task is to predict the retrieval quality of a search system for a query without relevance judgments. Research has shown the effectiveness and usefulness of QPP for ad-hoc search. Recent years have witnessed considerable progress in conv...
http://arxiv.org/pdf/2305.10923v1
2023-05-18T12:37:01
2023-05-18T12:37:01
{ "title": [ 0.002298355, -0.0031108856, -0.021835327, -0.05267334, 0.018463135, -0.0027179718, -0.01826477, -0.0104522705, 0.0206604, 0.04248047, -0.02128601, -0.010757446, -0.0007109642, -0.009689331, -0.057159424, -0.008125305, -0.002494812, 0...
http://arxiv.org/abs/2305.10920v1
Emergent Communication with Attention
Ryokan Ri, Ryo Ueda, Jason Naradowsky
cs.CL, cs.AI
To develop computational agents that better communicate using their own emergent language, we endow the agents with an ability to focus their attention on particular concepts in the environment. Humans often understand an object or scene as a composite of concepts and those concepts are further mapped onto words. We im...
http://arxiv.org/pdf/2305.10920v1
2023-05-18T12:31:45
2023-05-18T12:31:45
{ "title": [ -0.007537842, 0.009483337, -0.01424408, -0.027297974, -0.00088119507, -0.007247925, -0.007835388, -0.020584106, 0.015151978, 0.02973938, -0.021759033, -0.025543213, 0.0046806335, -0.02633667, -0.043121338, 0.00042271614, -0.00032830238, ...
http://arxiv.org/abs/2305.10913v2
Weakly-Supervised Visual-Textual Grounding with Semantic Prior Refinement
Davide Rigoni, Luca Parolari, Luciano Serafini, Alessandro Sperduti, Lamberto Ballan
cs.CV, cs.AI, cs.CL, cs.LG, cs.MM
Using only image-sentence pairs, weakly-supervised visual-textual grounding aims to learn region-phrase correspondences of the respective entity mentions. Compared to the supervised approach, learning is more difficult since bounding boxes and textual phrases correspondences are unavailable. In light of this, we propos...
http://arxiv.org/pdf/2305.10913v2
2023-09-26T09:29:26
2023-05-18T12:25:07
{ "title": [ -0.013206482, 0.018051147, -0.020248413, -0.06732178, -0.0006918907, 0.014213562, -0.05215454, -0.023406982, 0.023361206, 0.030563354, -0.041625977, 0.01876831, 0.009254456, -0.0057296753, -0.04852295, -0.011253357, 0.0027713776, 0.0...
http://arxiv.org/abs/2305.10907v1
Take a Break in the Middle: Investigating Subgoals towards Hierarchical Script Generation
Xinze Li, Yixin Cao, Muhao Chen, Aixin Sun
cs.CL
Goal-oriented Script Generation is a new task of generating a list of steps that can fulfill the given goal. In this paper, we propose to extend the task from the perspective of cognitive theory. Instead of a simple flat structure, the steps are typically organized hierarchically - Human often decompose a complex task ...
http://arxiv.org/pdf/2305.10907v1
2023-05-18T12:10:06
2023-05-18T12:10:06
{ "title": [ -0.037872314, -0.016418457, -0.05053711, -0.07873535, -0.0058784485, 0.007259369, -0.06652832, 0.008728027, -0.025344849, 0.036254883, -0.013648987, 0.025268555, -0.012161255, -0.015029907, -0.02633667, -0.017532349, -0.007980347, -0...
http://arxiv.org/abs/2305.10892v1
EventNet-ITA: Italian Frame Parsing for Events
Marco Rovera
cs.CL
This paper introduces EventNet-ITA, a large, multi-domain corpus annotated with event frames for Italian, and presents an efficient approach for multi-label Frame Parsing. The approach is then evaluated on the dataset. Covering a wide range of individual, social and historical phenomena, the main contribution of EventN...
http://arxiv.org/pdf/2305.10892v1
2023-05-18T11:41:56
2023-05-18T11:41:56
{ "title": [ -0.012992859, -0.014846802, -0.04257202, -0.080078125, 0.0013589859, 0.005340576, -0.07891846, -0.008361816, 0.023925781, 0.051971436, 0.0069503784, -0.0395813, 0.08477783, -0.042663574, -0.04272461, 0.02784729, -0.002210617, 0.02632...
http://arxiv.org/abs/2305.10866v1
TEPrompt: Task Enlightenment Prompt Learning for Implicit Discourse Relation Recognition
Wei Xiang, Chao Liang, Bang Wang
cs.CL
Implicit Discourse Relation Recognition (IDRR) aims at classifying the relation sense between two arguments without an explicit connective. Recently, the ConnPrompt~\cite{Wei.X:et.al:2022:COLING} has leveraged the powerful prompt learning for IDRR based on the fusion of multi-prompt decisions from three different yet m...
http://arxiv.org/pdf/2305.10866v1
2023-05-18T10:38:06
2023-05-18T10:38:06
{ "title": [ 0.012535095, 0.002029419, -0.061828613, -0.08135986, 0.0051612854, -0.01689148, -0.059265137, 0.02130127, 0.045684814, 0.0032081604, -0.028656006, -0.007724762, -0.0028915405, -0.021759033, -0.040863037, -0.008270264, -0.017593384, 0...
http://arxiv.org/abs/2305.10865v2
Semantically Aligned Task Decomposition in Multi-Agent Reinforcement Learning
Wenhao Li, Dan Qiao, Baoxiang Wang, Xiangfeng Wang, Bo Jin, Hongyuan Zha
cs.LG, cs.AI, cs.CL, cs.MA
The difficulty of appropriately assigning credit is particularly heightened in cooperative MARL with sparse reward, due to the concurrent time and structural scales involved. Automatic subgoal generation (ASG) has recently emerged as a viable MARL approach inspired by utilizing subgoals in intrinsically motivated reinf...
http://arxiv.org/pdf/2305.10865v2
2023-09-30T08:27:28
2023-05-18T10:37:54
{ "title": [ -0.020004272, 0.016204834, -0.047851562, -0.011627197, 0.021987915, -0.020629883, 0.0066375732, -0.014511108, -0.034576416, 0.020050049, -0.005645752, 0.031463623, -0.04336548, -0.008895874, -0.007926941, -0.01940918, -0.007873535, 0...
http://arxiv.org/abs/2305.10848v1
Advancing Full-Text Search Lemmatization Techniques with Paradigm Retrieval from OpenCorpora
Dmitriy Kalugin-Balashov
cs.CL
In this paper, we unveil a groundbreaking method to amplify full-text search lemmatization, utilizing the OpenCorpora dataset and a bespoke paradigm retrieval algorithm. Our primary aim is to streamline the extraction of a word's primary form or lemma - a crucial factor in full-text search. Additionally, we propose a c...
http://arxiv.org/pdf/2305.10848v1
2023-05-18T10:07:50
2023-05-18T10:07:50
{ "title": [ -0.00447464, 0.0044937134, -0.042175293, -0.08392334, 0.00077676773, 0.00085639954, -0.035247803, 0.0053482056, -0.016677856, 0.04067993, -0.016082764, 0.0070991516, 0.014907837, -0.019989014, -0.024642944, -0.0044784546, -0.011810303, ...
http://arxiv.org/abs/2305.10847v4
Large Language Models can be Guided to Evade AI-Generated Text Detection
Ning Lu, Shengcai Liu, Rui He, Qi Wang, Ke Tang
cs.CL, cs.AI
Large Language Models (LLMs) have demonstrated exceptional performance in a variety of tasks, including essay writing and question answering. However, it is crucial to address the potential misuse of these models, which can lead to detrimental outcomes such as plagiarism and spamming. Recently, several detectors have b...
http://arxiv.org/pdf/2305.10847v4
2023-06-17T03:48:41
2023-05-18T10:03:25
{ "title": [ 0.011367798, -0.025299072, 0.008323669, 0.010986328, -0.05807495, -0.0178833, 0.01966858, -0.014328003, -0.011817932, 0.025817871, -0.032958984, -0.012565613, 0.017028809, -0.037109375, 0.013511658, -0.0256958, -0.012825012, 0.052001...
http://arxiv.org/abs/2305.10845v1
TAPIR: Learning Adaptive Revision for Incremental Natural Language Understanding with a Two-Pass Model
Patrick Kahardipraja, Brielen Madureira, David Schlangen
cs.CL
Language is by its very nature incremental in how it is produced and processed. This property can be exploited by NLP systems to produce fast responses, which has been shown to be beneficial for real-time interactive applications. Recent neural network-based approaches for incremental processing mainly use RNNs or Tran...
http://arxiv.org/pdf/2305.10845v1
2023-05-18T09:58:19
2023-05-18T09:58:19
{ "title": [ -0.037994385, -0.020645142, -0.04220581, -0.08666992, -0.009300232, 0.008987427, -0.07324219, 0.026977539, 0.012390137, -0.011238098, -0.05444336, 0.0055656433, -0.029342651, 0.002937317, -0.0357666, -0.027923584, -0.010368347, 0.013...
http://arxiv.org/abs/2305.10839v1
A Lexical-aware Non-autoregressive Transformer-based ASR Model
Chong-En Lin, Kuan-Yu Chen
cs.CL, cs.SD, eess.AS
Non-autoregressive automatic speech recognition (ASR) has become a mainstream of ASR modeling because of its fast decoding speed and satisfactory result. To further boost the performance, relaxing the conditional independence assumption and cascading large-scaled pre-trained models are two active research directions. I...
http://arxiv.org/pdf/2305.10839v1
2023-05-18T09:50:47
2023-05-18T09:50:47
{ "title": [ -0.0067596436, 0.020126343, -0.052246094, -0.078430176, 0.002922058, 0.003873825, -0.07458496, 0.020141602, 0.00034952164, 0.046539307, -0.03100586, -0.019943237, 0.042633057, -0.005504608, -0.053863525, 0.00063991547, -0.0077400208, ...
http://arxiv.org/abs/2305.10835v1
Ahead-of-Time P-Tuning
Daniil Gavrilov, Nikita Balagansky
cs.LG, cs.CL
In this paper, we propose Ahead-of-Time (AoT) P-Tuning, a novel parameter-efficient fine-tuning method for pre-trained Language Models (LMs) that adds input-dependent bias before each Transformer layer. We evaluate AoT P-Tuning on GLUE and SuperGLUE benchmarking datasets using RoBERTa and DeBERTa models, showing that i...
http://arxiv.org/pdf/2305.10835v1
2023-05-18T09:24:53
2023-05-18T09:24:53
{ "title": [ 0.039276123, -0.018569946, -0.010864258, -0.06604004, -0.053100586, -0.0231781, -0.070007324, -0.0079956055, -0.039245605, 0.028060913, 0.0013513565, -0.016220093, 0.022277832, -0.047424316, 0.055480957, -0.015556335, 0.009864807, 0....
http://arxiv.org/abs/2305.10834v1
AIwriting: Relations Between Image Generation and Digital Writing
Scott Rettberg, Talan Memmott, Jill Walker Rettberg, Jason Nelson, Patrick Lichty
cs.AI, cs.CL, cs.HC, cs.MM, J.5
During 2022, both transformer-based AI text generation sys-tems such as GPT-3 and AI text-to-image generation systems such as DALL-E 2 and Stable Diffusion made exponential leaps forward and are unquestionably altering the fields of digital art and electronic literature. In this panel a group of electronic literature a...
http://arxiv.org/pdf/2305.10834v1
2023-05-18T09:23:05
2023-05-18T09:23:05
{ "title": [ -0.035308838, -0.009376526, -0.016113281, 0.010101318, -0.028961182, -0.0093688965, 0.01008606, -0.03286743, 0.0062561035, 0.02796936, -0.040618896, 0.028121948, 0.0012216568, -0.008262634, -0.035888672, -0.005718231, -0.013557434, 0...
http://arxiv.org/abs/2305.10833v3
Deep Learning Methods for Extracting Metaphorical Names of Flowers and Plants
Amal Haddad Haddad, Damith Premasiri, Tharindu Ranasinghe, Ruslan Mitkov
cs.CL
The domain of Botany is rich with metaphorical terms. Those terms play an important role in the description and identification of flowers and plants. However, the identification of such terms in discourse is an arduous task. This leads in some cases to committing errors during translation processes and lexicographic ta...
http://arxiv.org/pdf/2305.10833v3
2023-06-01T13:59:23
2023-05-18T09:22:29
{ "title": [ -0.041534424, -0.0040283203, -0.05633545, -0.010498047, -0.019546509, -0.027923584, -0.0040626526, -0.002729416, 0.0077934265, 0.038513184, -0.010925293, 0.0010318756, 0.028305054, -0.021133423, -0.02949524, 0.011436462, 0.0018701553, ...
http://arxiv.org/abs/2305.10819v2
CLEME: Debiasing Multi-reference Evaluation for Grammatical Error Correction
Jingheng Ye, Yinghui Li, Qingyu Zhou, Yangning Li, Shirong Ma, Hai-Tao Zheng, Ying Shen
cs.CL
Evaluating the performance of Grammatical Error Correction (GEC) systems is a challenging task due to its subjectivity. Designing an evaluation metric that is as objective as possible is crucial to the development of GEC task. However, mainstream evaluation metrics, i.e., reference-based metrics, introduce bias into th...
http://arxiv.org/pdf/2305.10819v2
2023-10-17T04:56:57
2023-05-18T08:57:17
{ "title": [ -0.01423645, 0.005870819, -0.033172607, -0.05834961, 0.016418457, 0.008163452, -0.04107666, -0.018753052, 0.010818481, 0.0362854, -0.011077881, 0.0003643036, 0.024627686, -0.012809753, -0.054351807, 0.008232117, -0.048736572, 0.07110...
http://arxiv.org/abs/2305.10818v1
Democratized Diffusion Language Model
Nikita Balagansky, Daniil Gavrilov
cs.LG, cs.CL
Despite the potential benefits of Diffusion Models for NLP applications, publicly available implementations, trained models, or reproducible training procedures currently need to be publicly available. We present the Democratized Diffusion Language Model (DDLM), based on the Continuous Diffusion for Categorical Data (C...
http://arxiv.org/pdf/2305.10818v1
2023-05-18T08:56:05
2023-05-18T08:56:05
{ "title": [ -0.015167236, 0.014183044, -0.061309814, -0.03250122, -0.029953003, -0.023025513, -0.045562744, 0.0073432922, -0.020126343, 0.047546387, 0.010810852, 0.0016727448, 0.020736694, -0.053741455, 0.0034656525, 0.018127441, -0.012962341, 0...
http://arxiv.org/abs/2305.10788v1
Whisper-KDQ: A Lightweight Whisper via Guided Knowledge Distillation and Quantization for Efficient ASR
Hang Shao, Wei Wang, Bei Liu, Xun Gong, Haoyu Wang, Yanmin Qian
cs.SD, cs.CL, eess.AS
Due to the rapid development of computing hardware resources and the dramatic growth of data, pre-trained models in speech recognition, such as Whisper, have significantly improved the performance of speech recognition tasks. However, these models usually have a high computational overhead, making it difficult to execu...
http://arxiv.org/pdf/2305.10788v1
2023-05-18T08:00:09
2023-05-18T08:00:09
{ "title": [ -0.026397705, -0.033111572, -0.06384277, -0.109680176, -0.020019531, 0.0046653748, -0.07922363, 0.04248047, -0.0011034012, 0.047180176, -0.0093688965, -0.002040863, 0.034362793, -0.022644043, -0.021972656, 0.009223938, 0.019973755, 0...
http://arxiv.org/abs/2305.10786v2
Ditto: A Simple and Efficient Approach to Improve Sentence Embeddings
Qian Chen, Wen Wang, Qinglin Zhang, Siqi Zheng, Chong Deng, Hai Yu, Jiaqing Liu, Yukun Ma, Chong Zhang
cs.CL
Prior studies diagnose the anisotropy problem in sentence representations from pre-trained language models, e.g., BERT, without fine-tuning. Our analysis reveals that the sentence embeddings from BERT suffer from a bias towards uninformative words, limiting the performance in semantic textual similarity (STS) tasks. To...
http://arxiv.org/pdf/2305.10786v2
2023-10-23T06:34:50
2023-05-18T07:56:40
{ "title": [ 0.0077934265, -0.0073776245, -0.033050537, -0.06604004, 0.020248413, 0.0006585121, -0.035614014, -0.010360718, 0.0340271, 0.016326904, -0.02178955, 0.013389587, 0.003829956, -0.009475708, -0.03768921, 0.017028809, -0.012687683, 0.050...
http://arxiv.org/abs/2305.10736v1
Counterfactual Debiasing for Generating Factually Consistent Text Summaries
Chenhe Dong, Yuexiang Xie, Yaliang Li, Ying Shen
cs.CL, cs.AI
Despite substantial progress in abstractive text summarization to generate fluent and informative texts, the factual inconsistency in the generated summaries remains an important yet challenging problem to be solved. In this paper, we construct causal graphs for abstractive text summarization and identify the intrinsic...
http://arxiv.org/pdf/2305.10736v1
2023-05-18T06:15:45
2023-05-18T06:15:45
{ "title": [ 0.0019054413, 0.018707275, -0.033081055, -0.05593872, 0.0015459061, -0.0018825531, -0.05050659, -0.042266846, 0.006591797, 0.036621094, -0.018966675, 0.019241333, 0.005657196, -0.0016403198, -0.061523438, 0.0014305115, -0.02053833, 0...
http://arxiv.org/abs/2305.10734v1
Diffusion-Based Speech Enhancement with Joint Generative and Predictive Decoders
Hao Shi, Kazuki Shimada, Masato Hirano, Takashi Shibuya, Yuichiro Koyama, Zhi Zhong, Shusuke Takahashi, Tatsuya Kawahara, Yuki Mitsufuji
cs.SD, cs.CL, eess.AS
Diffusion-based speech enhancement (SE) has been investigated recently, but its decoding is very time-consuming. One solution is to initialize the decoding process with the enhanced feature estimated by a predictive SE system. However, this two-stage method ignores the complementarity between predictive and diffusion S...
http://arxiv.org/pdf/2305.10734v1
2023-05-18T06:10:49
2023-05-18T06:10:49
{ "title": [ -0.014801025, 0.011146545, -0.06573486, -0.05130005, 0.026504517, -0.003686905, -0.009460449, 0.003967285, 0.014152527, 0.053985596, -0.00434494, -0.0096588135, 0.03564453, -0.035736084, -0.04827881, 0.016921997, -0.012298584, 0.0516...
http://arxiv.org/abs/2305.10731v2
Analyzing Norm Violations in Live-Stream Chat
Jihyung Moon, Dong-Ho Lee, Hyundong Cho, Woojeong Jin, Chan Young Park, Minwoo Kim, Jonathan May, Jay Pujara, Sungjoon Park
cs.CL
Toxic language, such as hate speech, can deter users from participating in online communities and enjoying popular platforms. Previous approaches to detecting toxic language and norm violations have been primarily concerned with conversations from online forums and social media, such as Reddit and Twitter. These approa...
http://arxiv.org/pdf/2305.10731v2
2023-10-08T03:57:31
2023-05-18T05:58:27
{ "title": [ 0.02053833, 0.0053749084, -0.022659302, -0.04827881, 0.011650085, -0.021026611, -0.005718231, 0.002626419, 0.0008406639, 0.06173706, -0.027450562, 0.009346008, 0.018463135, -0.0044517517, -0.055847168, 0.00274086, -0.010414124, 0.018...
http://arxiv.org/abs/2305.10713v2
Flatness-Aware Prompt Selection Improves Accuracy and Sample Efficiency
Lingfeng Shen, Weiting Tan, Boyuan Zheng, Daniel Khashabi
cs.CL, cs.LG
With growing capabilities of large language models, prompting them has become the dominant way to access them. This has motivated the development of strategies for automatically selecting effective language prompts. In this paper, we introduce prompt flatness, a new metric to quantify the expected utility of a language...
http://arxiv.org/pdf/2305.10713v2
2023-10-23T01:22:44
2023-05-18T05:17:57
{ "title": [ -0.010643005, -0.008911133, -0.033081055, -0.047576904, -0.024658203, -0.0022010803, -0.060668945, -0.022201538, 0.0209198, 0.026535034, -0.026657104, -0.024475098, 0.008506775, -0.03869629, -0.0357666, -0.01210022, 0.014205933, -0.0...
http://arxiv.org/abs/2305.10709v1
NoisywikiHow: A Benchmark for Learning with Real-world Noisy Labels in Natural Language Processing
Tingting Wu, Xiao Ding, Minji Tang, Hao Zhang, Bing Qin, Ting Liu
cs.CL, cs.AI
Large-scale datasets in the real world inevitably involve label noise. Deep models can gradually overfit noisy labels and thus degrade model generalization. To mitigate the effects of label noise, learning with noisy labels (LNL) methods are designed to achieve better generalization performance. Due to the lack of suit...
http://arxiv.org/pdf/2305.10709v1
2023-05-18T05:01:04
2023-05-18T05:01:04
{ "title": [ -0.024276733, -0.0025291443, -0.016204834, -0.03894043, -0.013786316, -0.0063934326, -0.01725769, -0.01197052, 0.006603241, 0.021652222, -0.051483154, 0.029693604, -0.008804321, -0.005458832, -0.01928711, -0.009170532, 0.0063705444, ...
http://arxiv.org/abs/2305.10703v1
ReGen: Zero-Shot Text Classification via Training Data Generation with Progressive Dense Retrieval
Yue Yu, Yuchen Zhuang, Rongzhi Zhang, Yu Meng, Jiaming Shen, Chao Zhang
cs.CL, cs.IR, cs.LG
With the development of large language models (LLMs), zero-shot learning has attracted much attention for various NLP tasks. Different from prior works that generate training data with billion-scale natural language generation (NLG) models, we propose a retrieval-enhanced framework to create training data from a genera...
http://arxiv.org/pdf/2305.10703v1
2023-05-18T04:30:09
2023-05-18T04:30:09
{ "title": [ -0.01033783, -0.0058021545, -0.04562378, -0.048217773, -0.006450653, -0.012580872, -0.012893677, 0.0009841919, 0.026397705, 0.029434204, -0.021057129, 0.0131073, -0.0043945312, -0.0038890839, -0.05706787, 0.0043182373, -0.003850937, ...
http://arxiv.org/abs/2305.10688v2
MolXPT: Wrapping Molecules with Text for Generative Pre-training
Zequn Liu, Wei Zhang, Yingce Xia, Lijun Wu, Shufang Xie, Tao Qin, Ming Zhang, Tie-Yan Liu
cs.CL
Generative pre-trained Transformer (GPT) has demonstrates its great success in natural language processing and related techniques have been adapted into molecular modeling. Considering that text is the most important record for scientific discovery, in this paper, we propose MolXPT, a unified language model of text and...
http://arxiv.org/pdf/2305.10688v2
2023-05-26T04:35:46
2023-05-18T03:58:19
{ "title": [ -0.045135498, -0.034057617, -0.046569824, -0.068847656, -0.029785156, 0.014320374, -0.0703125, 0.013748169, -0.018463135, 0.013221741, -0.01638794, 0.02960205, -0.018630981, -0.025360107, 0.0010938644, -0.0013628006, 0.0016717911, 0....
http://arxiv.org/abs/2305.10686v1
RMSSinger: Realistic-Music-Score based Singing Voice Synthesis
Jinzheng He, Jinglin Liu, Zhenhui Ye, Rongjie Huang, Chenye Cui, Huadai Liu, Zhou Zhao
cs.SD, cs.CL, eess.AS
We are interested in a challenging task, Realistic-Music-Score based Singing Voice Synthesis (RMS-SVS). RMS-SVS aims to generate high-quality singing voices given realistic music scores with different note types (grace, slur, rest, etc.). Though significant progress has been achieved, recent singing voice synthesis (SV...
http://arxiv.org/pdf/2305.10686v1
2023-05-18T03:57:51
2023-05-18T03:57:51
{ "title": [ -0.025512695, -0.018356323, -0.056549072, -0.05935669, -0.017227173, -0.015014648, -0.030014038, 0.024124146, -0.010635376, 0.05178833, -0.02168274, 0.008872986, 0.037597656, -0.027893066, -0.031036377, 0.004814148, -0.01739502, 0.04...
http://arxiv.org/abs/2305.10683v4
Paxion: Patching Action Knowledge in Video-Language Foundation Models
Zhenhailong Wang, Ansel Blume, Sha Li, Genglin Liu, Jaemin Cho, Zineng Tang, Mohit Bansal, Heng Ji
cs.CV, cs.CL
Action knowledge involves the understanding of textual, visual, and temporal aspects of actions. We introduce the Action Dynamics Benchmark (ActionBench) containing two carefully designed probing tasks: Action Antonym and Video Reversal, which targets multimodal alignment capabilities and temporal understanding skills ...
http://arxiv.org/pdf/2305.10683v4
2023-10-21T16:34:03
2023-05-18T03:53:59
{ "title": [ -0.042663574, -0.025741577, -0.059448242, -0.08148193, -0.015625, -0.0046920776, -0.0647583, 0.010734558, 0.0006489754, 0.029342651, -0.019897461, 0.01033783, 0.002046585, -0.012840271, -0.025939941, -0.01625061, -0.006385803, 0.0289...
http://arxiv.org/abs/2305.10680v2
Accurate and Reliable Confidence Estimation Based on Non-Autoregressive End-to-End Speech Recognition System
Xian Shi, Haoneng Luo, Zhifu Gao, Shiliang Zhang, Zhijie Yan
cs.SD, cs.CL, eess.AS
Estimating confidence scores for recognition results is a classic task in ASR field and of vital importance for kinds of downstream tasks and training strategies. Previous end-to-end~(E2E) based confidence estimation models (CEM) predict score sequences of equal length with input transcriptions, leading to unreliable e...
http://arxiv.org/pdf/2305.10680v2
2023-05-25T02:26:35
2023-05-18T03:34:50
{ "title": [ -0.025817871, 0.024887085, -0.06341553, -0.07598877, 0.02444458, -0.0026016235, -0.052337646, -0.005592346, -0.00023448467, 0.03250122, -0.03277588, -0.004676819, -0.0059318542, -0.02255249, -0.055664062, -0.01737976, -0.016433716, 0...
http://arxiv.org/abs/2305.10679v1
Think Outside the Code: Brainstorming Boosts Large Language Models in Code Generation
Xin-Ye Li, Jiang-Tian Xue, Zheng Xie, Ming Li
cs.AI, cs.CL, cs.SE
Code generation aims to automatically generate source code from high-level task specifications, which can significantly increase productivity of software engineering. Recently, approaches based on large language models (LLMs) have shown remarkable code generation abilities on simple tasks. However, generate code for mo...
http://arxiv.org/pdf/2305.10679v1
2023-05-18T03:32:54
2023-05-18T03:32:54
{ "title": [ -0.01638794, -0.04397583, -0.038635254, -0.04852295, -0.03692627, -0.016174316, -0.054779053, 0.02607727, -0.001868248, 0.012138367, -0.01838684, 0.017028809, -0.023452759, -0.021942139, -0.029937744, -0.019561768, -0.010269165, 0.03...
http://arxiv.org/abs/2305.10666v2
a unified front-end framework for english text-to-speech synthesis
Zelin Ying, Chen Li, Yu Dong, Qiuqiang Kong, Qiao Tian, Yuanyuan Huo, Yuxuan Wang
cs.CL, cs.SD, eess.AS
The front-end is a critical component of English text-to-speech (TTS) systems, responsible for extracting linguistic features that are essential for a text-to-speech model to synthesize speech, such as prosodies and phonemes. The English TTS front-end typically consists of a text normalization (TN) module, a prosody wo...
http://arxiv.org/pdf/2305.10666v2
2023-08-29T07:16:52
2023-05-18T02:57:54
{ "title": [ -0.021575928, -0.008972168, -0.036468506, -0.08892822, 0.010231018, 0.015457153, -0.064331055, 0.037902832, -0.0045166016, 0.051849365, -0.021896362, 0.010650635, 0.021362305, -0.010147095, -0.05206299, 0.009277344, -0.01574707, 0.02...
http://arxiv.org/abs/2305.10652v3
Speech Separation based on Contrastive Learning and Deep Modularization
Peter Ochieng
cs.SD, cs.CL, eess.AS
The current monaural state of the art tools for speech separation relies on supervised learning. This means that they must deal with permutation problem, they are impacted by the mismatch on the number of speakers used in training and inference. Moreover, their performance heavily relies on the presence of high-quality...
http://arxiv.org/pdf/2305.10652v3
2023-09-12T06:47:52
2023-05-18T02:19:05
{ "title": [ 0.0023918152, 0.024169922, -0.072509766, -0.08666992, 0.03173828, -0.01424408, -0.0385437, 0.0129470825, -0.017608643, 0.0501709, -0.0008621216, 0.03677368, 0.0031642914, -0.029632568, -0.021362305, 0.017196655, -0.013259888, 0.04010...
http://arxiv.org/abs/2305.10649v1
ZeroPrompt: Streaming Acoustic Encoders are Zero-Shot Masked LMs
Xingchen Song, Di Wu, Binbin Zhang, Zhendong Peng, Bo Dang, Fuping Pan, Zhiyong Wu
cs.SD, cs.CL, eess.AS, I.2.7
In this paper, we present ZeroPrompt (Figure 1-(a)) and the corresponding Prompt-and-Refine strategy (Figure 3), two simple but effective \textbf{training-free} methods to decrease the Token Display Time (TDT) of streaming ASR models \textbf{without any accuracy loss}. The core idea of ZeroPrompt is to append zeroed co...
http://arxiv.org/pdf/2305.10649v1
2023-05-18T02:08:33
2023-05-18T02:08:33
{ "title": [ 0.009338379, -0.008636475, -0.03994751, -0.034576416, -0.025466919, -0.021911621, -0.033416748, -0.0082092285, 0.004776001, 0.02268982, -0.027618408, 0.0014686584, 0.037109375, -0.049438477, 0.031234741, 0.00579834, -0.005268097, 0.0...
http://arxiv.org/abs/2305.10647v1
BioAug: Conditional Generation based Data Augmentation for Low-Resource Biomedical NER
Sreyan Ghosh, Utkarsh Tyagi, Sonal Kumar, Dinesh Manocha
cs.CL, cs.AI, cs.IR
Biomedical Named Entity Recognition (BioNER) is the fundamental task of identifying named entities from biomedical text. However, BioNER suffers from severe data scarcity and lacks high-quality labeled data due to the highly specialized and expert knowledge required for annotation. Though data augmentation has shown to...
http://arxiv.org/pdf/2305.10647v1
2023-05-18T02:04:38
2023-05-18T02:04:38
{ "title": [ -0.0024280548, 0.0019321442, -0.050598145, -0.09954834, 0.021133423, 0.022598267, -0.06921387, 0.013206482, 0.0004620552, 0.024002075, -0.025268555, 0.002248764, -0.004180908, -0.033447266, -0.041809082, -0.0048103333, 0.01197052, 0....
http://arxiv.org/abs/2305.10645v2
Are Large Language Models Fit For Guided Reading?
Peter Ochieng
cs.CL
This paper looks at the ability of large language models to participate in educational guided reading. We specifically, evaluate their ability to generate meaningful questions from the input text, generate diverse questions both in terms of content coverage and difficulty of the questions and evaluate their ability to ...
http://arxiv.org/pdf/2305.10645v2
2023-05-19T10:47:56
2023-05-18T02:03:55
{ "title": [ -0.068115234, -0.01651001, -0.020233154, -0.026535034, -0.05505371, 0.006336212, -0.06866455, -0.0018081665, 0.003435135, 0.031951904, -0.022460938, -0.011802673, 0.0046653748, -0.02229309, -0.032562256, -0.00932312, 0.005897522, 0.0...
http://arxiv.org/abs/2305.10626v3
Language Models Meet World Models: Embodied Experiences Enhance Language Models
Jiannan Xiang, Tianhua Tao, Yi Gu, Tianmin Shu, Zirui Wang, Zichao Yang, Zhiting Hu
cs.CL, cs.AI, cs.LG
While large language models (LMs) have shown remarkable capabilities across numerous tasks, they often struggle with simple reasoning and planning in physical environments, such as understanding object permanence or planning household activities. The limitation arises from the fact that LMs are trained only on written ...
http://arxiv.org/pdf/2305.10626v3
2023-10-28T17:55:44
2023-05-18T00:35:38
{ "title": [ -0.018432617, -0.011314392, -0.010437012, -0.04598999, -0.029296875, 0.005012512, -0.0435791, -0.010032654, 0.000003993511, 0.021621704, -0.020950317, -0.0015668869, 0.02204895, -0.0262146, -0.03414917, 0.016159058, -0.013465881, 0.0...
http://arxiv.org/abs/2305.10615v2
ML-SUPERB: Multilingual Speech Universal PERformance Benchmark
Jiatong Shi, Dan Berrebbi, William Chen, Ho-Lam Chung, En-Pei Hu, Wei Ping Huang, Xuankai Chang, Shang-Wen Li, Abdelrahman Mohamed, Hung-yi Lee, Shinji Watanabe
cs.SD, cs.CL, eess.AS
Speech processing Universal PERformance Benchmark (SUPERB) is a leaderboard to benchmark the performance of Self-Supervised Learning (SSL) models on various speech processing tasks. However, SUPERB largely considers English speech in its evaluation. This paper presents multilingual SUPERB (ML-SUPERB), covering 143 lang...
http://arxiv.org/pdf/2305.10615v2
2023-08-11T17:39:21
2023-05-18T00:01:27
{ "title": [ -0.032562256, -0.013763428, -0.042175293, -0.076049805, 0.0013017654, 0.006187439, -0.06964111, 0.011489868, -0.016342163, 0.055389404, 0.016311646, 0.009284973, 0.01864624, -0.026275635, -0.034851074, 0.028045654, -0.009010315, 0.04...
http://arxiv.org/abs/2305.10614v2
Token-wise Decomposition of Autoregressive Language Model Hidden States for Analyzing Model Predictions
Byung-Doh Oh, William Schuler
cs.CL, cs.AI
While there is much recent interest in studying why Transformer-based large language models make predictions the way they do, the complex computations performed within each layer have made their behavior somewhat opaque. To mitigate this opacity, this work presents a linear decomposition of final hidden states from aut...
http://arxiv.org/pdf/2305.10614v2
2023-06-02T22:50:07
2023-05-17T23:55:32
{ "title": [ -0.027404785, -0.0099105835, -0.033294678, -0.051330566, -0.028793335, -0.015853882, -0.05368042, 0.004623413, -0.033843994, 0.0345459, -0.01084137, 0.0155181885, 0.01828003, -0.036346436, -0.006916046, -0.00566864, 0.000008761883, 0...
http://arxiv.org/abs/2305.10613v3
Temporal Knowledge Graph Forecasting Without Knowledge Using In-Context Learning
Dong-Ho Lee, Kian Ahrabian, Woojeong Jin, Fred Morstatter, Jay Pujara
cs.CL
Temporal knowledge graph (TKG) forecasting benchmarks challenge models to predict future facts using knowledge of past facts. In this paper, we apply large language models (LLMs) to these benchmarks using in-context learning (ICL). We investigate whether and to what extent LLMs can be used for TKG forecasting, especial...
http://arxiv.org/pdf/2305.10613v3
2023-10-20T04:46:44
2023-05-17T23:50:28
{ "title": [ 0.0033550262, -0.015716553, -0.07702637, -0.09301758, -0.00217247, -0.024032593, -0.043762207, 0.023605347, 0.014465332, 0.038726807, -0.025970459, 0.0013952255, -0.026885986, -0.0035877228, -0.05770874, -0.022994995, 0.025360107, 0....
http://arxiv.org/abs/2305.10610v1
Solving Cosine Similarity Underestimation between High Frequency Words by L2 Norm Discounting
Saeth Wannasuphoprasit, Yi Zhou, Danushka Bollegala
cs.CL
Cosine similarity between two words, computed using their contextualised token embeddings obtained from masked language models (MLMs) such as BERT has shown to underestimate the actual similarity between those words (Zhou et al., 2022). This similarity underestimation problem is particularly severe for highly frequent ...
http://arxiv.org/pdf/2305.10610v1
2023-05-17T23:41:30
2023-05-17T23:41:30
{ "title": [ -0.030258179, 0.0026741028, -0.010467529, 0.01084137, -0.025405884, -0.019882202, -0.009063721, -0.033050537, 0.015296936, 0.02456665, -0.044555664, 0.030197144, 0.02658081, -0.01687622, 0.000813961, -0.012542725, -0.00057172775, 0.0...
http://arxiv.org/abs/2305.10601v1
Tree of Thoughts: Deliberate Problem Solving with Large Language Models
Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, Karthik Narasimhan
cs.CL, cs.AI, cs.LG
Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions pla...
http://arxiv.org/pdf/2305.10601v1
2023-05-17T23:16:17
2023-05-17T23:16:17
{ "title": [ -0.03062439, -0.032226562, -0.024383545, -0.056274414, -0.0126571655, 0.010719299, -0.04647827, -0.001996994, 0.013343811, 0.006702423, -0.018218994, -0.01576233, 0.0026855469, -0.029052734, -0.031799316, -0.007980347, -0.014862061, ...
http://arxiv.org/abs/2306.05539v1
Instruction Tuned Models are Quick Learners
Himanshu Gupta, Saurabh Arjun Sawant, Swaroop Mishra, Mutsumi Nakamura, Arindam Mitra, Santosh Mashetty, Chitta Baral
cs.CL
Instruction tuning of language models has demonstrated the ability to enhance model generalization to unseen tasks via in-context learning using a few examples. However, typical supervised learning still requires a plethora of downstream training data for finetuning. Often in real-world situations, there is a scarcity ...
http://arxiv.org/pdf/2306.05539v1
2023-05-17T22:30:01
2023-05-17T22:30:01
{ "title": [ 0.019317627, -0.024215698, 0.016357422, -0.012359619, -0.047973633, -0.010147095, -0.016708374, -0.020431519, -0.023773193, 0.031311035, -0.029922485, -0.022750854, -0.06549072, -0.01777649, -0.033447266, -0.04309082, 0.021377563, 0....
http://arxiv.org/abs/2305.10588v2
A Better Way to Do Masked Language Model Scoring
Carina Kauf, Anna Ivanova
cs.CL
Estimating the log-likelihood of a given sentence under an autoregressive language model is straightforward: one can simply apply the chain rule and sum the log-likelihood values for each successive token. However, for masked language models (MLMs), there is no direct way to estimate the log-likelihood of a sentence. T...
http://arxiv.org/pdf/2305.10588v2
2023-05-23T15:12:33
2023-05-17T21:51:58
{ "title": [ -0.028793335, 0.011497498, -0.055725098, -0.034698486, -0.030853271, -0.01725769, -0.066101074, -0.00096416473, -0.017700195, 0.018157959, -0.021102905, -0.02558899, 0.009094238, -0.05697632, 0.0047416687, -0.002380371, -0.003376007, ...
http://arxiv.org/abs/2305.10568v2
From chocolate bunny to chocolate crocodile: Do Language Models Understand Noun Compounds?
Jordan Coil, Vered Shwartz
cs.CL
Noun compound interpretation is the task of expressing a noun compound (e.g. chocolate bunny) in a free-text paraphrase that makes the relationship between the constituent nouns explicit (e.g. bunny-shaped chocolate). We propose modifications to the data and evaluation setup of the standard task (Hendrickx et al., 2013...
http://arxiv.org/pdf/2305.10568v2
2023-05-24T21:30:11
2023-05-17T21:05:23
{ "title": [ -0.048706055, -0.0038013458, -0.024902344, -0.059539795, -0.028244019, 0.011276245, -0.093688965, -0.0072021484, 0.0017147064, 0.009681702, -0.030288696, -0.00055933, 0.019622803, -0.024154663, -0.039733887, -0.00049591064, -0.039794922,...
http://arxiv.org/abs/2305.10563v2
HaSa: Hardness and Structure-Aware Contrastive Knowledge Graph Embedding
Honggen Zhang, June Zhang, Igor Molybog
cs.AI, cs.CL
We consider a contrastive learning approach to knowledge graph embedding (KGE) via InfoNCE. For KGE, efficient learning relies on augmenting the training data with negative triples. However, most KGE works overlook the bias from generating the negative triples-false negative triples (factual triples missing from the kn...
http://arxiv.org/pdf/2305.10563v2
2023-10-15T02:15:41
2023-05-17T20:46:18
{ "title": [ -0.012619019, -0.014663696, -0.04373169, -0.07366943, -0.0033740997, 0.0006842613, -0.045196533, 0.000037372112, 0.031021118, 0.018249512, -0.021560669, 0.017456055, -0.018218994, -0.01335144, -0.04220581, -0.0003323555, -0.002161026, ...
http://arxiv.org/abs/2305.11186v2
Compress, Then Prompt: Improving Accuracy-Efficiency Trade-off of LLM Inference with Transferable Prompt
Zhaozhuo Xu, Zirui Liu, Beidi Chen, Yuxin Tang, Jue Wang, Kaixiong Zhou, Xia Hu, Anshumali Shrivastava
cs.CL, cs.LG
While the numerous parameters in Large Language Models (LLMs) contribute to their superior performance, this massive scale makes them inefficient and memory-hungry. Thus, they are hard to deploy on commodity hardware, such as one single GPU. Given the memory and power constraints of such devices, model compression meth...
http://arxiv.org/pdf/2305.11186v2
2023-10-10T04:01:30
2023-05-17T20:45:13
{ "title": [ -0.078430176, -0.029403687, -0.06854248, -0.034698486, -0.04119873, -0.004337311, -0.08251953, 0.011711121, 0.007144928, -0.009361267, -0.039123535, 0.018920898, -0.026504517, -0.027999878, 0.006790161, -0.014099121, -0.000012218952, ...
http://arxiv.org/abs/2305.10561v1
Massively Multi-Lingual Event Understanding: Extraction, Visualization, and Search
Chris Jenkins, Shantanu Agarwal, Joel Barry, Steven Fincke, Elizabeth Boschee
cs.CL
In this paper, we present ISI-Clear, a state-of-the-art, cross-lingual, zero-shot event extraction system and accompanying user interface for event visualization & search. Using only English training data, ISI-Clear makes global events available on-demand, processing user-supplied text in 100 languages ranging from Afr...
http://arxiv.org/pdf/2305.10561v1
2023-05-17T20:41:51
2023-05-17T20:41:51
{ "title": [ -0.002784729, -0.007118225, -0.039245605, -0.100097656, 0.026351929, 0.01966858, -0.066345215, -0.0018262863, 0.021530151, 0.030685425, 0.0077171326, -0.0066070557, 0.0059661865, -0.023590088, -0.03778076, 0.026000977, 0.0041656494, ...
http://arxiv.org/abs/2305.10557v2
Bring More Attention to Syntactic Symmetry for Automatic Postediting of High-Quality Machine Translations
Baikjin Jung, Myungji Lee, Jong-Hyeok Lee, Yunsu Kim
cs.CL
Automatic postediting (APE) is an automated process to refine a given machine translation (MT). Recent findings present that existing APE systems are not good at handling high-quality MTs even for a language pair with abundant data resources, English-to-German: the better the given MT is, the harder it is to decide wha...
http://arxiv.org/pdf/2305.10557v2
2023-06-17T09:58:02
2023-05-17T20:25:19
{ "title": [ 0.00018954277, -0.0042037964, -0.009918213, -0.018188477, -0.022140503, -0.0068740845, -0.00793457, -0.013328552, -0.0043945312, 0.08380127, 0.005744934, 0.03125, 0.0769043, -0.023452759, -0.007980347, 0.050689697, 0.005256653, 0.018...
http://arxiv.org/abs/2305.10528v1
Scalable and Safe Remediation of Defective Actions in Self-Learning Conversational Systems
Sarthak Ahuja, Mohammad Kachuee, Fateme Sheikholeslami, Weiqing Liu, Jaeyoung Do
cs.AI, cs.CL, cs.LG
Off-Policy reinforcement learning has been a driving force for the state-of-the-art conversational AIs leading to more natural humanagent interactions and improving the user satisfaction for goal-oriented agents. However, in large-scale commercial settings, it is often challenging to balance between policy improvements...
http://arxiv.org/pdf/2305.10528v1
2023-05-17T19:22:24
2023-05-17T19:22:24
{ "title": [ 0.021209717, 0.016098022, -0.029571533, -0.084350586, 0.013946533, -0.006126404, -0.047027588, 0.0010604858, -0.0041275024, 0.04837036, -0.019424438, 0.02168274, 0.007797241, -0.015716553, -0.036499023, -0.0045051575, -0.005302429, 0...
http://arxiv.org/abs/2305.10519v2
Statistical Knowledge Assessment for Large Language Models
Qingxiu Dong, Jingjing Xu, Lingpeng Kong, Zhifang Sui, Lei Li
cs.CL, cs.LG
Given varying prompts regarding a factoid question, can a large language model (LLM) reliably generate factually correct answers? Existing LLMs may generate distinct responses for different prompts. In this paper, we study the problem of quantifying knowledge contained in an LLM regarding a given set of facts. We propo...
http://arxiv.org/pdf/2305.10519v2
2023-10-28T07:58:04
2023-05-17T18:54:37
{ "title": [ -0.053985596, -0.010643005, -0.071899414, -0.05996704, 0.000050842762, -0.005706787, -0.06793213, 0.021713257, 0.020889282, 0.02180481, -0.023727417, -0.012107849, 0.006454468, -0.018188477, -0.05822754, 0.0052948, 0.0022640228, 0.02...
http://arxiv.org/abs/2305.10512v1
IMAD: IMage-Augmented multi-modal Dialogue
Moskvoretskii Viktor, Frolov Anton, Kuznetsov Denis
cs.CL, cs.HC, I.4.10; I.7.m
Currently, dialogue systems have achieved high performance in processing text-based communication. However, they have not yet effectively incorporated visual information, which poses a significant challenge. Furthermore, existing models that incorporate images in dialogue generation focus on discussing the image itself...
http://arxiv.org/pdf/2305.10512v1
2023-05-17T18:38:10
2023-05-17T18:38:10
{ "title": [ 0.0068244934, 0.0052490234, -0.031341553, -0.046813965, 0.011878967, -0.009735107, -0.020446777, -0.033843994, -0.0007338524, 0.024169922, -0.0011863708, 0.003604889, -0.014091492, -0.03237915, -0.03414917, -0.010665894, -0.010688782, ...
http://arxiv.org/abs/2305.10510v1
ChatGPT Perpetuates Gender Bias in Machine Translation and Ignores Non-Gendered Pronouns: Findings across Bengali and Five other Low-Resource Languages
Sourojit Ghosh, Aylin Caliskan
cs.CY, cs.AI, cs.CL
In this multicultural age, language translation is one of the most performed tasks, and it is becoming increasingly AI-moderated and automated. As a novel AI system, ChatGPT claims to be proficient in such translation tasks and in this paper, we put that claim to the test. Specifically, we examine ChatGPT's accuracy in...
http://arxiv.org/pdf/2305.10510v1
2023-05-17T18:30:05
2023-05-17T18:30:05
{ "title": [ -0.024780273, -0.024414062, -0.011436462, 0.013206482, -0.053955078, -0.020065308, 0.0004925728, 0.00056934357, -0.006614685, 0.055419922, -0.024963379, 0.004928589, 0.05307007, -0.010749817, -0.015686035, -0.0021381378, -0.051696777, ...
http://arxiv.org/abs/2305.10496v1
Incorporating Attribution Importance for Improving Faithfulness Metrics
Zhixue Zhao, Nikolaos Aletras
cs.CL, cs.AI
Feature attribution methods (FAs) are popular approaches for providing insights into the model reasoning process of making predictions. The more faithful a FA is, the more accurately it reflects which parts of the input are more important for the prediction. Widely used faithfulness metrics, such as sufficiency and com...
http://arxiv.org/pdf/2305.10496v1
2023-05-17T18:05:49
2023-05-17T18:05:49
{ "title": [ -0.00042319298, -0.0019989014, -0.030792236, -0.051330566, -0.0015640259, -0.0033721924, -0.05029297, -0.02999878, -0.02949524, 0.04550171, 0.01096344, -0.00491333, 0.02255249, -0.043945312, -0.012817383, 0.008720398, -0.012672424, 0...
http://arxiv.org/abs/2305.10429v3
DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining
Sang Michael Xie, Hieu Pham, Xuanyi Dong, Nan Du, Hanxiao Liu, Yifeng Lu, Percy Liang, Quoc V. Le, Tengyu Ma, Adams Wei Yu
cs.CL, cs.LG
The mixture proportions of pretraining data domains (e.g., Wikipedia, books, web text) greatly affect language model (LM) performance. In this paper, we propose Domain Reweighting with Minimax Optimization (DoReMi), which first trains a small proxy model using group distributionally robust optimization (Group DRO) over...
http://arxiv.org/pdf/2305.10429v3
2023-10-24T23:16:15
2023-05-17T17:58:13
{ "title": [ -0.014518738, 0.0010328293, -0.06677246, -0.05621338, -0.02192688, -0.020019531, -0.07446289, -0.006919861, -0.046539307, 0.039367676, -0.018417358, 0.043640137, 0.026870728, -0.028747559, 0.014022827, 0.000726223, -0.011047363, 0.04...
http://arxiv.org/abs/2305.10427v1
Accelerating Transformer Inference for Translation via Parallel Decoding
Andrea Santilli, Silvio Severino, Emilian Postolache, Valentino Maiorca, Michele Mancusi, Riccardo Marin, Emanuele Rodolà
cs.CL, cs.AI, cs.LG
Autoregressive decoding limits the efficiency of transformers for Machine Translation (MT). The community proposed specific network architectures and learning-based methods to solve this issue, which are expensive and require changes to the MT model, trading inference speed at the cost of the translation quality. In th...
http://arxiv.org/pdf/2305.10427v1
2023-05-17T17:57:34
2023-05-17T17:57:34
{ "title": [ -0.06201172, -0.0027122498, -0.031799316, -0.058288574, -0.021270752, 0.020553589, -0.08111572, -0.0038223267, -0.014984131, 0.050476074, -0.0071144104, 0.014579773, 0.04748535, -0.025405884, -0.008148193, 0.022750854, 0.015327454, 0...
http://arxiv.org/abs/2305.10425v1
SLiC-HF: Sequence Likelihood Calibration with Human Feedback
Yao Zhao, Rishabh Joshi, Tianqi Liu, Misha Khalman, Mohammad Saleh, Peter J. Liu
cs.CL, cs.AI
Learning from human feedback has been shown to be effective at aligning language models with human preferences. Past work has often relied on Reinforcement Learning from Human Feedback (RLHF), which optimizes the language model using reward scores assigned from a reward model trained on human preference data. In this w...
http://arxiv.org/pdf/2305.10425v1
2023-05-17T17:57:10
2023-05-17T17:57:10
{ "title": [ -0.060516357, 0.00016736984, -0.04977417, -0.07946777, -0.000374794, 0.008422852, -0.09411621, -0.022766113, -0.014892578, 0.01965332, -0.019989014, -0.00083732605, 0.026062012, -0.033569336, -0.018554688, -0.0021266937, 0.00907135, ...
http://arxiv.org/abs/2305.10407v1
BAD: BiAs Detection for Large Language Models in the context of candidate screening
Nam Ho Koh, Joseph Plata, Joyce Chai
cs.CL, I.2, I.2.7, F.2.2, I.2.7
Application Tracking Systems (ATS) have allowed talent managers, recruiters, and college admissions committees to process large volumes of potential candidate applications efficiently. Traditionally, this screening process was conducted manually, creating major bottlenecks due to the quantity of applications and introd...
http://arxiv.org/pdf/2305.10407v1
2023-05-17T17:47:31
2023-05-17T17:47:31
{ "title": [ -0.00856781, -0.011886597, -0.017547607, -0.0340271, -0.0027770996, -0.0033607483, -0.03439331, -0.037597656, 0.0072669983, 0.012878418, -0.03527832, -0.0018234253, 0.007194519, -0.028579712, -0.010475159, -0.013587952, -0.031463623, ...
http://arxiv.org/abs/2305.10403v3
PaLM 2 Technical Report
Rohan Anil, Andrew M. Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, Eric Chu, Jonathan H. Clark, Laurent El Shafey, Yanping Huang, Kathy Meier-Hellstern, Gaurav Mishra, Erica Moreira, Mark Omernick, Kevin Robinson, Sebastian Ruder, Yi Ta...
cs.CL, cs.AI
We introduce PaLM 2, a new state-of-the-art language model that has better multilingual and reasoning capabilities and is more compute-efficient than its predecessor PaLM. PaLM 2 is a Transformer-based model trained using a mixture of objectives. Through extensive evaluations on English and multilingual language, and r...
http://arxiv.org/pdf/2305.10403v3
2023-09-13T20:35:45
2023-05-17T17:46:53
{ "title": [ -0.03543091, 0.00018894672, -0.03125, 0.000688076, -0.04598999, -0.010810852, -0.04953003, -0.020690918, -0.08654785, 0.053527832, 0.024337769, -0.004951477, 0.0138549805, -0.0357666, 0.037139893, -0.023849487, 0.012588501, 0.0060424...
http://arxiv.org/abs/2305.10400v3
What You See is What You Read? Improving Text-Image Alignment Evaluation
Michal Yarom, Yonatan Bitton, Soravit Changpinyo, Roee Aharoni, Jonathan Herzig, Oran Lang, Eran Ofek, Idan Szpektor
cs.CL, cs.CV
Automatically determining whether a text and a corresponding image are semantically aligned is a significant challenge for vision-language models, with applications in generative text-to-image and image-to-text tasks. In this work, we study methods for automatic text-image alignment evaluation. We first introduce SeeTR...
http://arxiv.org/pdf/2305.10400v3
2023-07-13T07:49:57
2023-05-17T17:43:38
{ "title": [ -0.027923584, 0.011238098, -0.028747559, -0.010284424, 0.012001038, -0.01864624, -0.0116119385, -0.039001465, 0.023880005, 0.04916382, -0.032714844, 0.01574707, 0.008895874, 0.005836487, -0.06506348, 0.003917694, -0.0020828247, 0.011...
http://arxiv.org/abs/2305.10387v3
Elaborative Simplification as Implicit Questions Under Discussion
Yating Wu, William Sheffield, Kyle Mahowald, Junyi Jessy Li
cs.CL
Automated text simplification, a technique useful for making text more accessible to people such as children and emergent bilinguals, is often thought of as a monolingual translation task from complex sentences to simplified sentences using encoder-decoder models. This view fails to account for elaborative simplificati...
http://arxiv.org/pdf/2305.10387v3
2023-10-24T14:00:22
2023-05-17T17:26:16
{ "title": [ 0.0063171387, -0.002817154, -0.020401001, -0.066223145, -0.015090942, 0.0016384125, -0.041015625, -0.016860962, -0.015083313, 0.013214111, -0.017333984, 0.015419006, -0.05633545, -0.019714355, -0.009521484, -0.021133423, 0.018127441, ...
http://arxiv.org/abs/2305.10384v1
Logit-Based Ensemble Distribution Distillation for Robust Autoregressive Sequence Uncertainties
Yassir Fathullah, Guoxuan Xia, Mark Gales
cs.LG, cs.CL
Efficiently and reliably estimating uncertainty is an important objective in deep learning. It is especially pertinent to autoregressive sequence tasks, where training and inference costs are typically very high. However, existing research has predominantly focused on tasks with static data such as image classification...
http://arxiv.org/pdf/2305.10384v1
2023-05-17T17:21:10
2023-05-17T17:21:10
{ "title": [ -0.0039482117, 0.03010559, -0.08001709, -0.07598877, 0.03729248, -0.0013275146, -0.0569458, -0.024383545, 0.034851074, 0.03112793, -0.03225708, 0.024627686, -0.015434265, -0.017852783, -0.065979004, 0.0053482056, -0.0033397675, -0.00...
http://arxiv.org/abs/2305.10383v2
Large-Scale Text Analysis Using Generative Language Models: A Case Study in Discovering Public Value Expressions in AI Patents
Sergio Pelaez, Gaurav Verma, Barbara Ribeiro, Philip Shapira
cs.CL, cs.IR
Labeling data is essential for training text classifiers but is often difficult to accomplish accurately, especially for complex and abstract concepts. Seeking an improved method, this paper employs a novel approach using a generative language model (GPT-4) to produce labels and rationales for large-scale text analysis...
http://arxiv.org/pdf/2305.10383v2
2023-05-18T12:34:47
2023-05-17T17:18:26
{ "title": [ -0.021026611, 0.0065994263, -0.045715332, -0.015129089, -0.0041885376, -0.014442444, -0.014808655, -0.016525269, 0.011566162, 0.026489258, -0.03894043, 0.0064201355, -0.009315491, -0.006942749, -0.059814453, -0.018875122, -0.02645874, ...
http://arxiv.org/abs/2305.10355v3
Evaluating Object Hallucination in Large Vision-Language Models
Yifan Li, Yifan Du, Kun Zhou, Jinpeng Wang, Wayne Xin Zhao, Ji-Rong Wen
cs.CV, cs.CL, cs.MM
Inspired by the superior language abilities of large language models (LLM), large vision-language models (LVLM) have been recently explored by integrating powerful LLMs for improving the performance on complex multimodal tasks. Despite the promising progress on LVLMs, we find that LVLMs suffer from the hallucination pr...
http://arxiv.org/pdf/2305.10355v3
2023-10-26T02:52:40
2023-05-17T16:34:01
{ "title": [ -0.008354187, 0.010673523, -0.024551392, -0.031555176, -0.0055885315, -0.024093628, -0.02243042, -0.028244019, 0.01928711, 0.02558899, -0.022766113, -0.0090408325, -0.0052261353, -0.013694763, -0.061584473, -0.009651184, 0.006954193, ...
http://arxiv.org/abs/2305.10349v1
Interactive Learning of Hierarchical Tasks from Dialog with GPT
Lane Lawley, Christopher J. MacLellan
cs.HC, cs.AI, cs.CL
We present a system for interpretable, symbolic, interactive task learning from dialog using a GPT model as a conversational front-end. The learned tasks are represented as hierarchical decompositions of predicate-argument structures with scoped variable arguments. By using a GPT model to convert interactive dialog int...
http://arxiv.org/pdf/2305.10349v1
2023-05-17T16:32:40
2023-05-17T16:32:40
{ "title": [ 0.015960693, 0.011810303, -0.036193848, -0.060394287, 0.0009417534, -0.018737793, -0.044189453, -0.022369385, -0.0019330978, 0.025970459, -0.02708435, 0.015808105, -0.023376465, -0.011230469, -0.046844482, -0.009475708, -0.021224976, ...
http://arxiv.org/abs/2305.10321v2
Controllable Speaking Styles Using a Large Language Model
Atli Thor Sigurgeirsson, Simon King
cs.CL, cs.SD, eess.AS
Reference-based Text-to-Speech (TTS) models can generate multiple, prosodically-different renditions of the same target text. Such models jointly learn a latent acoustic space during training, which can be sampled from during inference. Controlling these models during inference typically requires finding an appropriate...
http://arxiv.org/pdf/2305.10321v2
2023-09-19T16:35:57
2023-05-17T16:01:50
{ "title": [ -0.021896362, 0.0048980713, -0.060943604, -0.04043579, -0.00016903877, -0.019546509, -0.045837402, 0.0060157776, -0.01751709, 0.048919678, -0.0049057007, -0.0041923523, 0.015205383, -0.025543213, -0.025131226, 0.016235352, 0.00075006485,...
http://arxiv.org/abs/2305.10314v1
LeTI: Learning to Generate from Textual Interactions
Xingyao Wang, Hao Peng, Reyhaneh Jabbarvand, Heng Ji
cs.CL, cs.AI, cs.SE
Finetuning pre-trained language models (LMs) enhances the models' capabilities. Prior techniques fine-tune a pre-trained LM on input-output pairs (e.g., instruction fine-tuning), or with numerical rewards that gauge the quality of its outputs (e.g., reinforcement learning from human feedback). We explore LMs' potential...
http://arxiv.org/pdf/2305.10314v1
2023-05-17T15:53:31
2023-05-17T15:53:31
{ "title": [ -0.011268616, -0.026168823, -0.019714355, -0.06112671, -0.0019311905, 0.014892578, -0.05441284, -0.01121521, 0.0048103333, 0.013824463, -0.019927979, 0.0137786865, -0.021560669, -0.01096344, -0.045898438, -0.0038051605, -0.021728516, ...
http://arxiv.org/abs/2305.10307v4
FACE: Evaluating Natural Language Generation with Fourier Analysis of Cross-Entropy
Zuhao Yang, Yingfang Yuan, Yang Xu, Shuo Zhan, Huajun Bai, Kefan Chen
cs.CL
Measuring the distance between machine-produced and human language is a critical open problem. Inspired by empirical findings from psycholinguistics on the periodicity of entropy in language, we propose FACE, a set of metrics based on Fourier Analysis of the estimated Cross-Entropy of language, for measuring the simila...
http://arxiv.org/pdf/2305.10307v4
2023-10-25T13:32:58
2023-05-17T15:44:57
{ "title": [ -0.042114258, -0.01335907, -0.043304443, -0.036499023, -0.0066452026, 0.0028305054, -0.04638672, -0.0038909912, 0.00894928, 0.027313232, -0.019424438, 0.026885986, 0.008911133, -0.000054001808, -0.048553467, 0.01626587, -0.018478394, ...
http://arxiv.org/abs/2305.10306v3
UniEX: An Effective and Efficient Framework for Unified Information Extraction via a Span-extractive Perspective
Ping Yang, Junyu Lu, Ruyi Gan, Junjie Wang, Yuxiang Zhang, Jiaxing Zhang, Pingjian Zhang
cs.CL, cs.AI
We propose a new paradigm for universal information extraction (IE) that is compatible with any schema format and applicable to a list of IE tasks, such as named entity recognition, relation extraction, event extraction and sentiment analysis. Our approach converts the text-based IE tasks as the token-pair problem, whi...
http://arxiv.org/pdf/2305.10306v3
2023-05-22T05:25:33
2023-05-17T15:44:12
{ "title": [ -0.029891968, -0.013954163, -0.056365967, -0.08544922, 0.015991211, 0.012428284, -0.058624268, -0.0074653625, 0.011940002, 0.026916504, -0.019348145, 0.02532959, 0.0039138794, -0.010688782, -0.021759033, 0.009811401, -0.018585205, 0....
http://arxiv.org/abs/2305.10284v1
Towards More Robust NLP System Evaluation: Handling Missing Scores in Benchmarks
Anas Himmi, Ekhine Irurozki, Nathan Noiry, Stephan Clemencon, Pierre Colombo
cs.CL, cs.AI
The evaluation of natural language processing (NLP) systems is crucial for advancing the field, but current benchmarking approaches often assume that all systems have scores available for all tasks, which is not always practical. In reality, several factors such as the cost of running baseline, private systems, computa...
http://arxiv.org/pdf/2305.10284v1
2023-05-17T15:20:31
2023-05-17T15:20:31
{ "title": [ -0.0340271, -0.0053977966, -0.02204895, -0.038208008, -0.007045746, 0.0018787384, -0.03152466, -0.020095825, 0.013786316, 0.009490967, -0.054016113, 0.011077881, -0.032165527, 0.0014619827, -0.06640625, -0.032226562, -0.024993896, 0....
http://arxiv.org/abs/2305.10276v6
Chain-of-Symbol Prompting Elicits Planning in Large Langauge Models
Hanxu Hu, Hongyuan Lu, Huajian Zhang, Yun-Ze Song, Wai Lam, Yue Zhang
cs.CL
In this paper, we take the initiative to investigate the performance of LLMs on complex planning tasks that require LLMs to understand a virtual spatial environment simulated via natural language and act correspondingly in text. We propose a benchmark named Natural Language Planning and Action (Natala) composed of a se...
http://arxiv.org/pdf/2305.10276v6
2023-10-04T02:03:45
2023-05-17T15:07:50
{ "title": [ 0.007827759, -0.0073013306, -0.010665894, -0.025421143, -0.028808594, -0.017501831, -0.0008816719, -0.018341064, -0.044769287, 0.0056991577, -0.027801514, 0.022521973, -0.026397705, -0.03338623, 0.0075187683, -0.038757324, 0.015960693, ...
http://arxiv.org/abs/2305.10270v3
Boosting Local Spectro-Temporal Features for Speech Analysis
Michael Guerzhoy
cs.CL, cs.SD, eess.AS
We introduce the problem of phone classification in the context of speech recognition, and explore several sets of local spectro-temporal features that can be used for phone classification. In particular, we present some preliminary results for phone classification using two sets of features that are commonly used for ...
http://arxiv.org/pdf/2305.10270v3
2023-08-04T22:58:39
2023-05-17T15:02:20
{ "title": [ 0.00674057, 0.012039185, -0.058807373, -0.08654785, 0.037872314, -0.006416321, -0.04837036, 0.0073547363, 0.013885498, 0.037963867, -0.0014429092, 0.004360199, -0.011428833, -0.01725769, -0.067871094, 0.002216339, -0.0009803772, 0.05...
http://arxiv.org/abs/2305.10266v1
Searching for Needles in a Haystack: On the Role of Incidental Bilingualism in PaLM's Translation Capability
Eleftheria Briakou, Colin Cherry, George Foster
cs.CL
Large, multilingual language models exhibit surprisingly good zero- or few-shot machine translation capabilities, despite having never seen the intentionally-included translation examples provided to typical neural translation systems. We investigate the role of incidental bilingualism -- the unintentional consumption ...
http://arxiv.org/pdf/2305.10266v1
2023-05-17T14:58:06
2023-05-17T14:58:06
{ "title": [ -0.054870605, -0.01763916, -0.037475586, 0.010467529, -0.049224854, -0.0055999756, -0.04067993, -0.016174316, -0.059661865, 0.046905518, -0.007675171, 0.022216797, -0.01259613, -0.023635864, -0.019302368, -0.032409668, -0.011291504, ...
http://arxiv.org/abs/2305.10263v2
M3KE: A Massive Multi-Level Multi-Subject Knowledge Evaluation Benchmark for Chinese Large Language Models
Chuang Liu, Renren Jin, Yuqi Ren, Linhao Yu, Tianyu Dong, Xiaohan Peng, Shuting Zhang, Jianxiang Peng, Peiyi Zhang, Qingqing Lyu, Xiaowen Su, Qun Liu, Deyi Xiong
cs.CL
Large language models have recently made tremendous progress in a variety of aspects, e.g., cross-task generalization, instruction following. Comprehensively evaluating the capability of large language models in multiple tasks is of great importance. In this paper, we propose M3KE, a Massive Multi-Level Multi-Subject K...
http://arxiv.org/pdf/2305.10263v2
2023-05-21T03:57:11
2023-05-17T14:56:31
{ "title": [ -0.05267334, -0.013160706, -0.04953003, -0.062469482, -0.012023926, 0.002866745, -0.05770874, 0.026626587, 0.023239136, 0.02923584, -0.0066375732, 0.018859863, -0.0011739731, -0.022018433, -0.050231934, 0.005420685, -0.02003479, 0.02...
http://arxiv.org/abs/2305.10250v3
MemoryBank: Enhancing Large Language Models with Long-Term Memory
Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, Yanlin Wang
cs.CL, cs.AI
Revolutionary advancements in Large Language Models have drastically reshaped our interactions with artificial intelligence systems. Despite this, a notable hindrance remains-the deficiency of a long-term memory mechanism within these models. This shortfall becomes increasingly evident in situations demanding sustained...
http://arxiv.org/pdf/2305.10250v3
2023-05-21T06:20:28
2023-05-17T14:40:29
{ "title": [ -0.007598877, -0.0064735413, -0.040374756, -0.07897949, 0.015419006, 0.0064849854, -0.049621582, 0.004512787, 0.021133423, 0.028198242, -0.013725281, 0.005996704, 0.010635376, -0.016159058, -0.043518066, 0.02458191, -0.010139465, 0.0...
http://arxiv.org/abs/2305.10236v1
A quantitative study of NLP approaches to question difficulty estimation
Luca Benedetto
cs.CL, cs.LG
Recent years witnessed an increase in the amount of research on the task of Question Difficulty Estimation from Text QDET with Natural Language Processing (NLP) techniques, with the goal of targeting the limitations of traditional approaches to question calibration. However, almost the entirety of previous research foc...
http://arxiv.org/pdf/2305.10236v1
2023-05-17T14:26:00
2023-05-17T14:26:00
{ "title": [ -0.008628845, -0.0069122314, -0.011802673, -0.025009155, -0.009414673, -0.006931305, -0.008804321, -0.023483276, 0.038360596, 0.001414299, -0.031463623, 0.0050849915, -0.027740479, -0.011787415, -0.061798096, -0.002111435, -0.021575928, ...
http://arxiv.org/abs/2305.15425v2
Language Model Tokenizers Introduce Unfairness Between Languages
Aleksandar Petrov, Emanuele La Malfa, Philip H. S. Torr, Adel Bibi
cs.CL, cs.LG
Recent language models have shown impressive multilingual performance, even when not explicitly trained for it. Despite this, there are concerns about the quality of their outputs across different languages. In this paper, we show how disparity in the treatment of different languages arises at the tokenization stage, w...
http://arxiv.org/pdf/2305.15425v2
2023-10-20T10:09:55
2023-05-17T14:17:57
{ "title": [ 0.011726379, -0.008674622, -0.010055542, -0.026306152, -0.05606079, -0.012741089, -0.029510498, 0.015792847, -0.04019165, 0.050628662, -0.020874023, -0.014259338, 0.015510559, -0.029968262, 0.005542755, -0.020355225, -0.0093688965, 0...
http://arxiv.org/abs/2305.10231v1
OpenSLU: A Unified, Modularized, and Extensible Toolkit for Spoken Language Understanding
Libo Qin, Qiguang Chen, Xiao Xu, Yunlong Feng, Wanxiang Che
cs.CL
Spoken Language Understanding (SLU) is one of the core components of a task-oriented dialogue system, which aims to extract the semantic meaning of user queries (e.g., intents and slots). In this work, we introduce OpenSLU, an open-source toolkit to provide a unified, modularized, and extensible toolkit for spoken lang...
http://arxiv.org/pdf/2305.10231v1
2023-05-17T14:12:29
2023-05-17T14:12:29
{ "title": [ -0.018127441, -0.018051147, -0.040283203, -0.08251953, 0.0034942627, 0.012283325, -0.0769043, 0.012817383, -0.009307861, 0.01953125, 0.00059986115, -0.00705719, 0.0028076172, -0.024658203, -0.026123047, 0.0032176971, -0.008003235, 0....
http://arxiv.org/abs/2305.10204v1
Shielded Representations: Protecting Sensitive Attributes Through Iterative Gradient-Based Projection
Shadi Iskander, Kira Radinsky, Yonatan Belinkov
cs.CL, cs.AI
Natural language processing models tend to learn and encode social biases present in the data. One popular approach for addressing such biases is to eliminate encoded information from the model's representations. However, current methods are restricted to removing only linearly encoded information. In this work, we pro...
http://arxiv.org/pdf/2305.10204v1
2023-05-17T13:26:57
2023-05-17T13:26:57
{ "title": [ -0.011642456, 0.03286743, -0.019226074, -0.05029297, 0.019821167, 0.009239197, -0.027145386, -0.049591064, -0.01626587, 0.03012085, 0.0041046143, -0.0090408325, 0.01739502, -0.031280518, -0.014373779, -0.004589081, -0.0039138794, 0.0...
http://arxiv.org/abs/2305.10196v1
A Survey on Zero Pronoun Translation
Longyue Wang, Siyou Liu, Mingzhou Xu, Linfeng Song, Shuming Shi, Zhaopeng Tu
cs.CL, cs.AI
Zero pronouns (ZPs) are frequently omitted in pro-drop languages (e.g. Chinese, Hungarian, and Hindi), but should be recalled in non-pro-drop languages (e.g. English). This phenomenon has been studied extensively in machine translation (MT), as it poses a significant challenge for MT systems due to the difficulty in de...
http://arxiv.org/pdf/2305.10196v1
2023-05-17T13:19:01
2023-05-17T13:19:01
{ "title": [ -0.023498535, -0.008285522, -0.03060913, -0.060180664, -0.008148193, 0.002336502, -0.028884888, 0.0071029663, 0.01109314, 0.061676025, 0.007843018, -0.015640259, 0.03036499, -0.029251099, -0.056396484, 0.029006958, -0.024337769, 0.04...
http://arxiv.org/abs/2305.10195v1
Boosting Distress Support Dialogue Responses with Motivational Interviewing Strategy
Anuradha Welivita, Pearl Pu
cs.CL
AI-driven chatbots have become an emerging solution to address psychological distress. Due to the lack of psychotherapeutic data, researchers use dialogues scraped from online peer support forums to train them. But since the responses in such platforms are not given by professionals, they contain both conforming and no...
http://arxiv.org/pdf/2305.10195v1
2023-05-17T13:18:28
2023-05-17T13:18:28
{ "title": [ 0.035461426, 0.010238647, -0.03338623, -0.07672119, 0.008010864, -0.012390137, -0.055603027, -0.01360321, -0.013191223, 0.016845703, 0.014457703, 0.011924744, 0.0074920654, -0.070617676, 0.021118164, 0.0026893616, -0.0036449432, 0.02...
http://arxiv.org/abs/2305.10190v1
Variable-length Neural Interlingua Representations for Zero-shot Neural Machine Translation
Zhuoyuan Mao, Haiyue Song, Raj Dabre, Chenhui Chu, Sadao Kurohashi
cs.CL
The language-independency of encoded representations within multilingual neural machine translation (MNMT) models is crucial for their generalization ability on zero-shot translation. Neural interlingua representations have been shown as an effective method for achieving this. However, fixed-length neural interlingua r...
http://arxiv.org/pdf/2305.10190v1
2023-05-17T13:15:10
2023-05-17T13:15:10
{ "title": [ -0.034454346, -0.009315491, -0.023544312, -0.044830322, -0.026657104, 0.020050049, -0.046569824, 0.015625, -0.0017337799, 0.05783081, -0.011604309, -0.0032367706, 0.00818634, -0.009536743, -0.047332764, 0.013000488, -0.010047913, 0.0...
http://arxiv.org/abs/2305.10172v1
Knowledge-enhanced Mixed-initiative Dialogue System for Emotional Support Conversations
Yang Deng, Wenxuan Zhang, Yifei Yuan, Wai Lam
cs.CL, cs.IR
Unlike empathetic dialogues, the system in emotional support conversations (ESC) is expected to not only convey empathy for comforting the help-seeker, but also proactively assist in exploring and addressing their problems during the conversation. In this work, we study the problem of mixed-initiative ESC where the use...
http://arxiv.org/pdf/2305.10172v1
2023-05-17T12:55:52
2023-05-17T12:55:52
{ "title": [ -0.021530151, -0.009170532, -0.05215454, -0.05239868, 0.013084412, -0.015930176, -0.020736694, 0.015266418, -0.00075387955, 0.04171753, -0.016448975, 0.0027256012, 0.015022278, -0.032104492, -0.04046631, 0.0021190643, -0.000107467175, ...
http://arxiv.org/abs/2305.10167v1
Pragmatic Reasoning in Structured Signaling Games
Emil Carlsson, Devdatt Dubhashi
cs.AI, cs.CL
In this work we introduce a structured signaling game, an extension of the classical signaling game with a similarity structure between meanings in the context, along with a variant of the Rational Speech Act (RSA) framework which we call structured-RSA (sRSA) for pragmatic reasoning in structured domains. We explore t...
http://arxiv.org/pdf/2305.10167v1
2023-05-17T12:43:29
2023-05-17T12:43:29
{ "title": [ 0.0035076141, -0.0026378632, 0.012504578, -0.036895752, -0.010620117, -0.0048599243, -0.0047340393, -0.028305054, -0.01637268, -0.020721436, -0.02420044, -0.0064315796, -0.072387695, -0.036346436, 0.002439499, -0.05657959, -0.0058555603,...
http://arxiv.org/abs/2305.10163v2
Qualifying Chinese Medical Licensing Examination with Knowledge Enhanced Generative Pre-training Model
Jiageng Wu, Xian Wu, Zhaopeng Qiu, Minghui Li, Yefeng Zheng, Jie Yang
cs.CL, cs.AI, cs.CY
Generative Pre-Training (GPT) models like ChatGPT have demonstrated exceptional performance in various Natural Language Processing (NLP) tasks. Although ChatGPT has been integrated into the overall workflow to boost efficiency in many domains, the lack of flexibility in the finetuning process hinders its applications i...
http://arxiv.org/pdf/2305.10163v2
2023-10-22T17:03:23
2023-05-17T12:31:26
{ "title": [ 0.01939392, 0.0021972656, -0.021255493, -0.058258057, 0.0017271042, -0.014595032, -0.018615723, -0.019378662, -0.013061523, 0.025497437, 0.001821518, -0.002374649, -0.00014436245, -0.049560547, -0.021835327, -0.00381279, -0.0051879883, ...