Corrections: September 2024

data/xml/2020.findings.xml

@@ -330,7 +330,7 @@
     </paper>
     <paper id="23">
       <title>Minimize Exposure Bias of <fixed-case>S</fixed-case>eq2<fixed-case>S</fixed-case>eq Models in Joint Entity and Relation Extraction</title>
-      <author><first>Ranran Haoran</first><last>Zhang</last></author>
+      <author id="ranran-haoran-zhang"><first>Ranran Haoran</first><last>Zhang</last></author>
       <author><first>Qianying</first><last>Liu</last></author>
       <author><first>Aysa Xuemo</first><last>Fan</last></author>
       <author><first>Heng</first><last>Ji</last></author>

data/xml/2021.findings.xml

@@ -7954,17 +7954,19 @@
       <author><first>Li-Ming</first><last>Zhan</last></author>
       <author><first>Jiaxin</first><last>Chen</last></author>
       <author><first>Guangyuan</first><last>Shi</last></author>
-      <author><first>Xiao-Ming</first><last>Wu</last></author>
       <author><first>Albert Y.S.</first><last>Lam</last></author>
+      <author><first>Xiao-Ming</first><last>Wu</last></author>
       <pages>1114–1120</pages>
       <abstract>This paper investigates the effectiveness of pre-training for few-shot intent classification. While existing paradigms commonly further pre-train language models such as BERT on a vast amount of unlabeled corpus, we find it highly effective and efficient to simply fine-tune BERT with a small set of labeled utterances from public datasets. Specifically, fine-tuning BERT with roughly 1,000 labeled data yields a pre-trained model – IntentBERT, which can easily surpass the performance of existing pre-trained models for few-shot intent classification on novel domains with very different semantics. The high effectiveness of IntentBERT confirms the feasibility and practicality of few-shot intent detection, and its high generalization ability across different domains suggests that intent classification tasks may share a similar underlying structure, which can be efficiently learned from a small set of labeled data. The source code can be found at <url>https://github.com/hdzhang-code/IntentBERT</url>.</abstract>
-      <url hash="2ae86f5a">2021.findings-emnlp.96</url>
+      <url hash="297df895">2021.findings-emnlp.96</url>
       <bibkey>zhang-etal-2021-effectiveness-pre</bibkey>
       <doi>10.18653/v1/2021.findings-emnlp.96</doi>
       <video href="2021.findings-emnlp.96.mp4"/>
       <pwcdataset url="https://paperswithcode.com/dataset/banking77">BANKING77</pwcdataset>
       <pwcdataset url="https://paperswithcode.com/dataset/hint3">HINT3</pwcdataset>
       <pwcdataset url="https://paperswithcode.com/dataset/hwu64">HWU64</pwcdataset>
+      <revision id="1" href="2021.findings-emnlp.96v1" hash="2ae86f5a"/>
+      <revision id="2" href="2021.findings-emnlp.96v2" hash="297df895" date="2024-09-17">Changes the order of the authors.</revision>
     </paper>
     <paper id="97">
       <title>Improving Abstractive Dialogue Summarization with Hierarchical Pretraining and Topic Segment</title>

data/xml/2021.naacl.xml

@@ -7496,7 +7496,7 @@
       <author><first>Jiawei</first><last>Ma</last></author>
       <author><first>Jingxuan</first><last>Tu</last></author>
       <author><first>Ying</first><last>Lin</last></author>
-      <author><first>Ranran Haoran</first><last>Zhang</last></author>
+      <author id="ranran-haoran-zhang"><first>Ranran Haoran</first><last>Zhang</last></author>
       <author><first>Weili</first><last>Liu</last></author>
       <author><first>Aabhas</first><last>Chauhan</last></author>
       <author><first>Yingjun</first><last>Guan</last></author>

data/xml/2022.deeplo.xml

@@ -30,7 +30,7 @@
       <author><first>William</first><last>Chen</last></author>
       <author><first>Richard</first><last>Castro</last></author>
       <author><first>Núria</first><last>Bel</last></author>
-      <author><first>Cesar</first><last>Toshio</last></author>
+      <author><first>Cesar</first><last>Yoshikawa</last></author>
       <author><first>Renzo</first><last>Venturas</last></author>
       <author><first>Hilario</first><last>Aradiel</last></author>
       <author><first>Nelsi</first><last>Melgarejo</last></author>

data/xml/2022.naacl.xml

@@ -615,13 +615,13 @@
       <author><first>Haode</first><last>Zhang</last></author>
       <author><first>Haowen</first><last>Liang</last></author>
       <author><first>Yuwei</first><last>Zhang</last></author>
-      <author><first>Li-Ming</first><last>Zhan</last></author>
-      <author><first>Xiao-Ming</first><last>Wu</last></author>
+      <author><first>Liming</first><last>Zhan</last></author>
       <author><first>Xiaolei</first><last>Lu</last></author>
       <author><first>Albert</first><last>Lam</last></author>
+      <author><first>Xiao-Ming</first><last>Wu</last></author>
       <pages>532-542</pages>
       <abstract>It is challenging to train a good intent classifier for a task-oriented dialogue system with only a few annotations. Recent studies have shown that fine-tuning pre-trained language models with a small set of labeled utterances from public benchmarks in a supervised manner is extremely helpful. However, we find that supervised pre-training yields an anisotropic feature space, which may suppress the expressive power of the semantic representations. Inspired by recent research in isotropization, we propose to improve supervised pre-training by regularizing the feature space towards isotropy. We propose two regularizers based on contrastive learning and correlation matrix respectively, and demonstrate their effectiveness through extensive experiments. Our main finding is that it is promising to regularize supervised pre-training with isotropization to further improve the performance of few-shot intent detection. The source code can be found at <url>https://github.com/fanolabs/isoIntentBert-main</url>.</abstract>
-      <url hash="5e5f256f">2022.naacl-main.39</url>
+      <url hash="7283669b">2022.naacl-main.39</url>
       <attachment type="software" hash="f3d4a0c2">2022.naacl-main.39.software.zip</attachment>
       <bibkey>zhang-etal-2022-fine</bibkey>
       <doi>10.18653/v1/2022.naacl-main.39</doi>
@@ -630,6 +630,8 @@
       <pwcdataset url="https://paperswithcode.com/dataset/banking77">BANKING77</pwcdataset>
       <pwcdataset url="https://paperswithcode.com/dataset/hint3">HINT3</pwcdataset>
       <pwcdataset url="https://paperswithcode.com/dataset/hwu64">HWU64</pwcdataset>
+      <revision id="1" href="2022.naacl-main.39v1" hash="5e5f256f"/>
+      <revision id="2" href="2022.naacl-main.39v2" hash="7283669b" date="2024-09-17">Changes the order of the authors.</revision>
     </paper>
     <paper id="40">
       <title>Cross-document Misinformation Detection based on Event Graph Reasoning</title>

data/xml/2023.acl.xml

@@ -3128,10 +3128,12 @@
       <author><first>Anette</first><last>Frank</last><affiliation>Heidelberg University</affiliation></author>
       <pages>4032-4059</pages>
       <abstract>Vision and language models (VL) are known to exploit unrobust indicators in individual modalities (e.g., introduced by distributional biases) instead of focusing on relevant information in each modality. That a unimodal model achieves similar accuracy on a VL task to a multimodal one, indicates that so-called unimodal collapse occurred. However, accuracy-based tests fail to detect e.g., when the model prediction is wrong, while the model used relevant information from a modality. Instead, we propose MM-SHAP, a performance-agnostic multimodality score based on Shapley values that reliably quantifies in which proportions a multimodal model uses individual modalities. We apply MM-SHAP in two ways: (1) to compare models for their average degree of multimodality, and (2) to measure for individual models the contribution of individual modalities for different tasks and datasets. Experiments with six VL models – LXMERT, CLIP and four ALBEF variants – on four VL tasks highlight that unimodal collapse can occur to different degrees and in different directions, contradicting the wide-spread assumption that unimodal collapse is one-sided. Based on our results, we recommend MM-SHAP for analysing multimodal tasks, to diagnose and guide progress towards multimodal integration. Code available at <url>https://github.com/Heidelberg-NLP/MM-SHAP</url>.</abstract>
-      <url hash="86b9cb56">2023.acl-long.223</url>
+      <url hash="b1be1a8c">2023.acl-long.223</url>
       <bibkey>parcalabescu-frank-2023-mm</bibkey>
       <doi>10.18653/v1/2023.acl-long.223</doi>
       <video href="2023.acl-long.223.mp4"/>
+      <revision id="1" href="2023.acl-long.223v1" hash="86b9cb56"/>
+      <revision id="2" href="2023.acl-long.223v2" hash="b1be1a8c" date="2024-09-18">This revision includes mentions a sponsor in the Acknowledgments section and rectifies the line below Eq. (1).</revision>
     </paper>
     <paper id="224">
       <title>Towards Boosting the Open-Domain Chatbot with Human Feedback</title>

data/xml/2023.eacl.xml

@@ -1920,7 +1920,7 @@
     </paper>
     <paper id="142">
       <title><fixed-case>C</fixed-case>on<fixed-case>E</fixed-case>ntail: An Entailment-based Framework for Universal Zero and Few Shot Classification with Supervised Contrastive Pretraining</title>
-      <author><first>Ranran Haoran</first><last>Zhang</last><affiliation>The Pennsylvania State University</affiliation></author>
+      <author id="ranran-haoran-zhang"><first>Ranran Haoran</first><last>Zhang</last><affiliation>The Pennsylvania State University</affiliation></author>
       <author><first>Aysa Xuemo</first><last>Fan</last><affiliation>University of Illinois at Urbana-Champaign</affiliation></author>
       <author><first>Rui</first><last>Zhang</last><affiliation>Penn State University</affiliation></author>
       <pages>1941-1953</pages>

data/xml/2023.emnlp.xml

@@ -6067,7 +6067,7 @@
     <paper id="433">
       <title>Unified Low-Resource Sequence Labeling by Sample-Aware Dynamic Sparse Finetuning</title>
       <author><first>Sarkar Snigdha Sarathi</first><last>Das</last></author>
-      <author><first>Haoran</first><last>Zhang</last></author>
+      <author id="ranran-haoran-zhang"><first>Ranran Haoran</first><last>Zhang</last></author>
       <author><first>Peng</first><last>Shi</last></author>
       <author><first>Wenpeng</first><last>Yin</last></author>
       <author><first>Rui</first><last>Zhang</last></author>

data/xml/2023.findings.xml

@@ -3166,10 +3166,12 @@
       <author><first>Ryan</first><last>Cotterell</last><affiliation>ETH Zürich</affiliation></author>
       <pages>598-614</pages>
       <abstract>Byte-Pair Encoding (BPE) is a popular algorithm used for tokenizing data in NLP, despite being devised initially as a compression method.BPE appears to be a greedy algorithm at face value, but the underlying optimization problem that BPE seeks to solve has not yet been laid down. We formalize BPE as a combinatorial optimization problem. Via submodular functions, we prove that the iterative greedy version is a 1/sigma*(1-e(-sigma))-approximation of an optimal merge sequence, where sigma is the total backward curvature with respect to the optimal merge sequence. Empirically the lower bound of the approximation is approx0.37.We provide a faster implementation of BPE which improves the runtime complexity from O(NM) to O(N log M), where N is the sequence length and M is the merge count. Finally, we optimize the brute-force algorithm for optimal BPE using memoization.</abstract>
-      <url hash="40ed62e8">2023.findings-acl.38</url>
+      <url hash="8867d506">2023.findings-acl.38</url>
       <bibkey>zouhar-etal-2023-formal</bibkey>
       <doi>10.18653/v1/2023.findings-acl.38</doi>
       <video href="2023.findings-acl.38.mp4"/>
+      <revision id="1" href="2023.findings-acl.38v1" hash="40ed62e8"/>
+      <revision id="2" href="2023.findings-acl.38v2" hash="8867d506" date="2024-09-17">Fix typos in Proof of Theorem 4.2 and Algorithm 3 as well as the malformed rendering of Figure 3.</revision>
     </paper>
     <paper id="39">
       <title>Automatic Named Entity Obfuscation in Speech</title>
@@ -11991,14 +11993,16 @@
       <title>Revisit Few-shot Intent Classification with <fixed-case>PLM</fixed-case>s: Direct Fine-tuning vs. Continual Pre-training</title>
       <author><first>Haode</first><last>Zhang</last><affiliation>The Hong Kong Polytechnic University</affiliation></author>
       <author><first>Haowen</first><last>Liang</last><affiliation>The Hong Kong Polytechnic University</affiliation></author>
-      <author><first>Li-Ming</first><last>Zhan</last><affiliation>The Hong Kong Polytechnic University</affiliation></author>
-      <author><first>Xiao-Ming</first><last>Wu</last><affiliation>Hong Kong Polytechnic University</affiliation></author>
+      <author><first>Liming</first><last>Zhan</last><affiliation>The Hong Kong Polytechnic University</affiliation></author>
       <author><first>Albert Y.S.</first><last>Lam</last><affiliation>Fano Labs</affiliation></author>
+      <author><first>Xiao-Ming</first><last>Wu</last><affiliation>Hong Kong Polytechnic University</affiliation></author>
       <pages>11105-11121</pages>
       <abstract>We consider the task of few-shot intent detection, which involves training a deep learning model to classify utterances based on their underlying intents using only a small amount of labeled data. The current approach to address this problem is through continual pre-training, i.e., fine-tuning pre-trained language models (PLMs) on external resources (e.g., conversational corpora, public intent detection datasets, or natural language understanding datasets) before using them as utterance encoders for training an intent classifier. In this paper, we show that continual pre-training may not be essential, since the overfitting problem of PLMs on this task may not be as serious as expected. Specifically, we find that directly fine-tuning PLMs on only a handful of labeled examples already yields decent results compared to methods that employ continual pre-training, and the performance gap diminishes rapidly as the number of labeled data increases. To maximize the utilization of the limited available data, we propose a context augmentation method and leverage sequential self-distillation to boost performance. Comprehensive experiments on real-world benchmarks show that given only two or more labeled samples per class, direct fine-tuning outperforms many strong baselines that utilize external data sources for continual pre-training. The code can be found at <url>https://github.com/hdzhang-code/DFTPlus</url>.</abstract>
-      <url hash="811ea19b">2023.findings-acl.706</url>
+      <url hash="268745c5">2023.findings-acl.706</url>
       <bibkey>zhang-etal-2023-revisit</bibkey>
       <doi>10.18653/v1/2023.findings-acl.706</doi>
+      <revision id="1" href="2023.findings-acl.706v1" hash="811ea19b"/>
+      <revision id="2" href="2023.findings-acl.706v2" hash="268745c5" date="2024-09-17">Changes the order of the authors.</revision>
     </paper>
     <paper id="707">
       <title>Improving Contrastive Learning of Sentence Embeddings from <fixed-case>AI</fixed-case> Feedback</title>
@@ -21079,7 +21083,7 @@
     <paper id="496">
       <title>Exploring the Potential of Large Language Models in Generating Code-Tracing Questions for Introductory Programming Courses</title>
       <author><first>Aysa</first><last>Fan</last></author>
-      <author><first>Haoran</first><last>Zhang</last></author>
+      <author id="ranran-haoran-zhang"><first>Ranran Haoran</first><last>Zhang</last></author>
       <author><first>Luc</first><last>Paquette</last></author>
       <author><first>Rui</first><last>Zhang</last></author>
       <pages>7406-7421</pages>