A quick guide to Amazon’s 45-plus NAACL papers

{"value":"Amazon’s 45-plus papers at the annual meeting of the North American chapter of the Association for Computational Linguistics, which ++[begins next week](https://www.amazon.science/conferences-and-events/naacl-2022)++, sorted by research area.\n\n#### **Continual learning**\n++[Lifelong pretraining: Continually adapting language models to emerging corpora](https://www.amazon.science/publications/lifelong-pretraining-continually-adapting-language-models-to-emerging-corpora)++\nXisen Jin, Dejiao Zhang, Henghui Zhu, Wei Xiao, Shang-Wen Li, Xiaokai Wei, Andrew O. Arnold, Xiang Ren\n\n++[Local-to-global learning for iterative training of production SLU models on new features](https://www.amazon.science/publications/local-to-global-learning-for-iterative-training-of-production-slu-models-on-new-features)++\nYulia Grishina, Daniil Sorokin\n\n++[Overcoming catastrophic forgetting during domain adaptation of seq2seq language generation](https://www.amazon.science/publications/overcoming-catastrophic-forgetting-during-domain-adaptation-of-seq2seq-language-generation)++\nDingcheng Li, Zheng Chen, Eunah Cho, Jie Hao, Xiaohu Liu, Xing Fan, Chenlei (Edward) Guo, Yang Liu\n\n\n\nIn \"[Overcoming catastrophic forgetting during domain adaptation of seq2seq language generation](https://www.amazon.science/publications/overcoming-catastrophic-forgetting-during-domain-adaptation-of-seq2seq-language-generation)\", Amazon researchers propose a method for estimating how much data representations shift when an existing model is trained on a new task (right).\n\n++[Temporal generalization for spoken language understanding](https://www.amazon.science/publications/temporal-generalization-for-spoken-language-understanding)++\nJudith Gaspers, Anoop Kumar, Greg Ver Steeg, Aram Galstyan\n\n#### **Data augmentation**\n\n++[Constraining word alignments with posterior regularization for label transfer](https://www.amazon.science/publications/constraining-word-alignments-with-posterior-regularization-for-label-transfer)++\nKevin Martin Jose, Thomas Gueudré\n\n\n\nAn example of the difficulty in using word alignment to transfer textual labels from one language to another. In English, the article \"the\" is assigned the label \"o\", for \"other\"; in French, the abbreviated article is combined with its noun, and both receive the same label (\"type\"). From \"++[Constraining word alignments with posterior regularization for label transfer](https://www.amazon.science/publications/constraining-word-alignments-with-posterior-regularization-for-label-transfer)++\".\n\n++[Controlled data generation via insertion operations for NLU](https://www.amazon.science/publications/controlled-data-generation-via-insertion-operations-for-nlu)++\nManoj Kumar, Haidar Khan, Yuval Merhav, Wael Hamza, Anna Rumshisky, Rahul Gupta\n\n++[Efficient semi supervised consistency training for natural language understanding](https://www.amazon.science/publications/efficient-semi-supervised-consistency-training-for-natural-language-understanding)++\nGeorge Leung, Joshua Tan\n\n++[Learning to generate examples for semantic processing tasks](https://www.amazon.science/publications/learning-to-generate-examples-for-semantic-processing-tasks)++\nDanilo Croce, Simone Filice, Giuseppe Castellucci, Roberto Basili\n\n#### **Dialogue**\n\n++[Learning dialogue representations from consecutive utterances](https://www.amazon.science/publications/learning-dialogue-representations-from-consecutive-utterances)++\nZhihan Zhou, Dejiao Zhang, Wei Xiao, Nicholas Dingwall, Xiaofei Ma, Andrew O. Arnold, Bing Xiang\n\n++[Massive-scale decoding for text generation using lattices](https://www.amazon.science/publications/massive-scale-decoding-for-text-generation-using-lattices)++\nJiacheng Xu, Siddhartha Reddy Jonnalagadda, Greg Durrett\n\n#### **Entity linking, resolution, and typing**\n\n++[Contrastive representation learning for cross-document coreference resolution of events and entities](https://www.amazon.science/publications/contrastive-representation-learning-for-cross-document-coreference-resolution-of-events-and-entities)++\nBenjamin Hsu, Graham Horwood\n\n++[Improving entity disambiguation by reasoning over a knowledge base](https://www.amazon.science/publications/improving-entity-disambiguation-by-reasoning-over-a-knowledge-base)++\nTom Ayoola, Joseph Fisher, Andrea Pierleoni\n\n++[ReFinED: An efficient zero-shot-capable approach to end-to-end entity linking](https://www.amazon.science/publications/refined-an-efficient-zero-shot-capable-approach-to-end-to-end-entity-linking)++\nTom Ayoola, Shubhi Tyagi, Joseph Fisher, Christos Christodoulopoulos, Andrea Pierleoni\n\n++[Instilling type knowledge in language models via multi-task QA](https://www.amazon.science/publications/instilling-type-knowledge-in-language-models-via-multi-task-qa)++\nShuyang Li, Mukund Sridhar, Chandana Satya Prakash, Jin Cao, Wael Hamza, Julian McAuley\n\n#### **Explainable AI**\n\n++[Entailment tree explanations via iterative retrieval-generation reasoner](https://www.amazon.science/publications/entailment-tree-explanations-via-iterative-retrieval-generation-reasoner)++\nDanilo Neves Ribeiro, Shen Wang, Xiaofei Ma, Rui Dong, Xiaokai Wei, Henry Zhu, Xinchi Chen, Zhiheng Huang, Peng Xu, Andrew O. Arnold, Dan Roth\n\n++[Locally aggregated feature attribution on natural language model understanding](https://www.amazon.science/publications/locally-aggregated-feature-attribution-on-natural-language-model-understanding)++\nSheng Zhang, Jin Wang, Haitao Jiang, Rui Song\n\nExtreme multilabel classificatio\n\n\n\nIn \"Entailment tree explanations via iterative retrieval-generation reasoner\", Amazon researchers propose a method for explaining the outputs of large language models by logically recombining premises extracted from supporting textual evidence.\n\n#### **Extreme multilabel classification**\n\n++[Augmenting training data for massive semantic matching models in low-traffic e-commerce stores](https://www.amazon.science/publications/augmenting-training-data-for-massive-semantic-matching-models-in-low-traffic-e-commerce-stores)++\nAshutosh Joshi, Shankar Vishwanath, Choon Hui Teo, Vaclav Petricek, Vishy Vishwanathan, Rahul Bhagat, Jonathan May\n\n++[Extreme zero shot learning for extreme text classification](https://www.amazon.science/publications/extreme-zero-shot-learning-for-extreme-text-classification)++\nYuanhao Xiong, Wei-Cheng Chang, Cho-Jui Hsieh, Hsiang-Fu Yu, Inderjit S. Dhillon\n\n#### **Extreme multilabel classification**\n\n++[Augmenting training data for massive semantic matching models in low-traffic e-commerce stores](https://www.amazon.science/publications/augmenting-training-data-for-massive-semantic-matching-models-in-low-traffic-e-commerce-stores)++\nAshutosh Joshi, Shankar Vishwanath, Choon Hui Teo, Vaclav Petricek, Vishy Vishwanathan, Rahul Bhagat, Jonathan May\n\n++[Extreme zero shot learning for extreme text classification](https://www.amazon.science/publications/extreme-zero-shot-learning-for-extreme-text-classification)++\nYuanhao Xiong, Wei-Cheng Chang, Cho-Jui Hsieh, Hsiang-Fu Yu, Inderjit S. Dhillon\n\n#### **Federated learning**\n\n++[Federated learning with noisy user feedback](https://www.amazon.science/publications/federated-learning-with-noisy-user-feedback)++\nRahul Sharma, Anil Ramakrishna, Ansel MacLaughlin, Anna Rumshisky, Jimit Majmudar, Clement Chung, Salman Avestimehr, Rahul Gupta\n\n#### **Keyword spotting**\n\n++[AB/BA analysis: A framework for estimating keyword spotting recall improvement while maintaining audio privacy](https://www.amazon.science/publications/ab-ba-analysis-a-framework-for-estimating-keyword-spotting-recall-improvement-while-maintaining-audio-privacy)++\nRaphael Petegrosso, Vasistakrishna Baderdinni, Thibaud Senechal, Benjamin L. Bullough\n\n#### **Machine translation**\n++[CoCoA-MT: A dataset and benchmark for contrastive controlled MT with application to formality](https://www.amazon.science/publications/cocoa-mt-a-dataset-and-benchmark-for-contrastive-controlled-mt-with-application-to-formality)++\nMaria Nadejde, Anna Currey, Benjamin Hsu, Xing Niu, Marcello Federico, Georgiana Dinu\n\n\n\nIn federated learning, distributed copies of a neural network are trained locally, and only their updates (red) are sent to a central model. \"++[Training mixed-domain translation models via federated learning](https://www.amazon.science/publications/training-mixed-domain-translation-models-via-federated-learning)++\" introduces a technique called dynamic pulling, in which distributed models with large shifts in parameter values between training rounds (lower left) see their parameters pulled into the central model separately from those of models with smaller shifts.\n\n++[The devil is in the details: On the pitfalls of vocabulary selection in neural machine translation](https://www.amazon.science/publications/the-devil-is-in-the-details-on-the-pitfalls-of-vocabulary-selection-in-neural-machine-translation)++\nTobias Domhan, Eva Hasler, Ke Tran, Sony Trenous, Bill Byrne, Felix Hieber\n\n++[Training mixed-domain translation models via federated learning](https://www.amazon.science/publications/training-mixed-domain-translation-models-via-federated-learning)++\nPeyman Passban, Tanya G. Roosta, Rahul Gupta, Ankit Chadha, Clement Chung\n\n#### **Multitask learning**\n++[Asynchronous convergence in multi-task learning via knowledge distillation from converged tasks](https://www.amazon.science/publications/asynchronous-convergence-in-multi-task-learning-via-knowledge-distillation-from-converged-tasks)++\nWeiyi Lu, Sunny Rajagopalan, Priyanka Nigam, Jaspreet Singh, Xiaodi Sun, Yi Xu, Belinda Zeng, Trishul Chilimbi\n\n++[Exploring the role of task transferability in large-scale multi-task learning](https://www.amazon.science/publications/exploring-the-role-of-task-transferability-in-large-scale-multi-task-learning)++\nVishakh Padmakumar, Leonard Lausen, Miguel Ballesteros, Sheng Zha, He He, George Karypis\n\n#### **Named-entity recognition**\n++[Dynamic gazetteer integration in multilingual models for cross-lingual and cross-domain named entity recognition](https://www.amazon.science/publications/dynamic-gazetteer-integration-in-multilingual-models-for-cross-lingual-and-cross-domain-named-entity-recognition)++\nBesnik Fetahu, Anjie Fang, Oleg Rokhlenko, Shervin Malmasi\n\n++[NER-MQMRC: Formulating named entity recognition as multi question machine reading comprehension](https://www.amazon.science/publications/ner-mqmrc-formulating-named-entity-recognition-as-multi-question-machine-reading-comprehension)++\nAnubhav Shrimal, Avi Jain, Kartik Mehta, Promod Yenigalla\n\n#### **Question answering**\n++[Answer consolidation: Formulation and benchmarking](https://www.amazon.science/publications/answer-consolidation-formulation-and-benchmarking)++\nWenxuan Zhou, Qiang Ning, Heba Elfardy, Kevin Small, Muhao Chen\n\n++[Paragraph-based transformer pre-training for multi-sentence inference](https://www.amazon.science/publications/paragraph-based-transformer-pre-training-for-multi-sentence-inference)++\nLuca Di Liello, Siddhant Garg, Luca Soldaini, Alessandro Moschitti\n\n++[PerKGQA: Question answering over personalized knowledge graphs](https://www.amazon.science/publications/perkgqa-question-answering-over-personalized-knowledge-graphs)++\nRitam Dutt, Kasturi Bhattacharjee, Rashmi Gangadharaiah, Dan Roth, Carolyn Penstein Rosé\n\n++[Product answer generation from heterogeneous sources: A new benchmark and best practices](https://www.amazon.science/publications/product-answer-generation-from-heterogeneous-sources-a-new-benchmark-and-best-practices)++\nXiaoyu Shen, Gianni Barlacchi, Marco Del Tredici, Weiwei Cheng, Adria de Gispert, Bill Byrne\n\n#### **Recommender systems**\n++[CERES: Pretraining of graph-conditioned transformer for semi-structured session data](https://www.amazon.science/publications/ceres-pretraining-of-graph-conditioned-transformer-for-semi-structured-session-data)++\nRui Feng, Chen Luo, Qingyu Yin, Bing Yin, Tuo Zhao, Chao Zhang\n\n#### **Self-learning**\n\n++[FPI: Failure point isolation in large-scale conversational assistants](https://www.amazon.science/publications/fpi-failure-point-isolation-in-large-scale-conversational-assistants)++\nRinat Khaziev, Usman Shahid, Tobias Röding, Rakesh Chada, Emir Kapanci, Pradeep Natarajan\n\n++[Scalable and robust self-learning for skill routing in large-scale conversational AI systems](https://www.amazon.science/publications/scalable-and-robust-self-learning-for-skill-routing-in-large-scale-conversational-ai-systems)++\nMohammad Kachuee, Jinseok Nam, Sarthak Ahuja, Jin-Myung Won, Sungjin Lee\n\n++[Self-aware feedback-based self-learning in large-scale conversational AI](https://www.amazon.science/publications/self-aware-feedback-based-self-learning-in-large-scale-conversational-ai)++\nPragaash Ponnusamy, Clint Solomon Mathialagan, Gustavo Aguilar, Chengyuan Ma, Chenlei (Edward) Guo\n\n\n\nIn \"FPI: Failure point isolation in large-scale conversational assistants\", Amazon researchers propose a method for deducing where in a conversational agent's processing pipeline an error has occurred.\n\n\nAn example of task-oriented semantic parsing, which converts natural language into a formal representation that an AI agent can act on. From \"[Compositional task-oriented parsing as abstractive question answering](https://www.amazon.science/publications/compositional-task-oriented-parsing-as-abstractive-question-answering)\".\n\n#### **Semantic parsing**\n++[Compositional task oriented parsing as abstractive question answering](https://www.amazon.science/publications/compositional-task-oriented-parsing-as-abstractive-question-answering)++\nWenting Zhao, Konstantine Arkoudas, Weiqi Sun, Claire Cardie\n\n++[SeqZero: Few-shot compositional semantic parsing with sequential prompts and zero-shot models](https://www.amazon.science/publications/seqzero-few-shot-compositional-semantic-parsing-with-sequential-prompts-and-zero-shot-models)++\nJingfeng Yang, Haoming Jiang, Qingyu Yin, Danqing Zhang, Bing Yin, Diyi Yang\n\n#### **Task adaptation**\n++[Attention fusion: A light yet efficient late fusion mechanism for task adaptation in NLU](https://www.amazon.science/publications/attention-fusion-a-light-yet-efficient-late-fusion-mechanism-for-task-adaptation-in-nlu)++\nJin Cao, Chandana Satya Prakash, Wael Hamza\n\n++[Empowering parameter-efficient transfer learning by recognizing the kernel structure in attention](https://www.amazon.science/publications/empowering-parameter-efficient-transfer-learning-by-recognizing-the-kernel-structure-in-attention)++\nYifan Chen, Devamanyu Hazarika, Mahdi Namazifar, Yang Liu, Di Jin, Dilek Hakkani-Tür\n\n#### **Text mining**\n++[Distantly supervised aspect clustering and naming for e-commerce reviews](https://www.amazon.science/publications/distantly-supervised-aspect-clustering-and-naming-for-e-commerce-reviews)++\nPrateek Sircar, Aniket Chakrabarti, Deepak Gupta, Anirban Majumdar\n\n++[Efficient few-shot fine-tuning for opinion summarization](https://www.amazon.science/publications/efficient-few-shot-fine-tuning-for-opinion-summarization)++\nArthur Bražinskas, Ramesh Nallapati, Mohit Bansal, Markus Dreyer\n\n++[FactGraph: Evaluating factuality in summarization with semantic graph representations](https://www.amazon.science/publications/factgraph-evaluating-factuality-in-summarization-with-semantic-graph-representations)++\nLeonardo F. R. Ribeiro, Mengwen Liu, Iryna Gurevych, Markus Dreyer, Mohit Bansal\n\n++[Enhanced knowledge selection for grounded dialogues via document semantic graphs](https://www.amazon.science/publications/enhanced-knowledge-selection-for-grounded-dialogues-via-document-semantic-graphs)++\nSha Li, Madhi Namazifar, Di Jin, Mohit Bansal, Heng Ji, Yang Liu, Dilek Hakkani-Tür\n\n++[Retrieval-augmented multilingual keyphrase generation with retriever-generator iterative training](https://www.amazon.science/publications/retrieval-augmented-multilingual-keyphrase-generation-with-retriever-generator-iterative-training)++\nYifan Gao, Qingyu Yin, Zheng Li, Rui Meng, Tong Zhao, Bing Yin, Irwin King, Michael R. Lyu\n\n++[What do users care about? Detecting actionable insights from user feedback](https://www.amazon.science/publications/what-do-users-care-about-detecting-actionable-insights-from-user-feedback)++\nKasturi Bhattacharjee, Rashmi Gangadharaiah, Kathleen McKeown, Dan Roth\n\n\n\nAn example of how a conversational agent might incorporate facts gleaned form online sources (white boxes) into its conversational replies (blue boxes). From \"++[Enhanced knowledge selection for grounded dialogues via document semantic graphs](https://www.amazon.science/publications/enhanced-knowledge-selection-for-grounded-dialogues-via-document-semantic-graphs)++\".\n\n#### **Text-to-speech**\n++[Empathic machines: using intermediate features as levers to emulate emotions in text-to-speech systems](https://www.amazon.science/publications/empathic-machines-using-intermediate-features-as-levers-to-emulate-emotions-in-text-to-speech-systems)++\nSaiteja Kosgi, Sarath Sivaprasad, Niranjan Pedanekar, Anil Nelakanti, Vineet Gandhi\n\nABOUT THE AUTHOR\n\n#### **Staff writer**","render":"<p>Amazon’s 45-plus papers at the annual meeting of the North American chapter of the Association for Computational Linguistics, which <ins><a href=\\"https://www.amazon.science/conferences-and-events/naacl-2022\\" target=\\"_blank\\">begins next week</a></ins>, sorted by research area.</p>\n<h4><a id=\\"Continual_learning_2\\"></a><strong>Continual learning</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/lifelong-pretraining-continually-adapting-language-models-to-emerging-corpora\\" target=\\"_blank\\">Lifelong pretraining: Continually adapting language models to emerging corpora</a></ins><br />\\nXisen Jin, Dejiao Zhang, Henghui Zhu, Wei Xiao, Shang-Wen Li, Xiaokai Wei, Andrew O. Arnold, Xiang Ren</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/local-to-global-learning-for-iterative-training-of-production-slu-models-on-new-features\\" target=\\"_blank\\">Local-to-global learning for iterative training of production SLU models on new features</a></ins><br />\\nYulia Grishina, Daniil Sorokin</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/overcoming-catastrophic-forgetting-during-domain-adaptation-of-seq2seq-language-generation\\" target=\\"_blank\\">Overcoming catastrophic forgetting during domain adaptation of seq2seq language generation</a></ins><br />\\nDingcheng Li, Zheng Chen, Eunah Cho, Jie Hao, Xiaohu Liu, Xing Fan, Chenlei (Edward) Guo, Yang Liu</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/c17bd9eff0864d8eb33b49a36826e3d1_%E4%B8%8B%E8%BD%BD.jpg\\" alt=\\"下载.jpg\\" /></p>\n<p>In “<a href=\\"https://www.amazon.science/publications/overcoming-catastrophic-forgetting-during-domain-adaptation-of-seq2seq-language-generation\\" target=\\"_blank\\">Overcoming catastrophic forgetting during domain adaptation of seq2seq language generation</a>”, Amazon researchers propose a method for estimating how much data representations shift when an existing model is trained on a new task (right).</p>\\n<p><ins><a href=\\"https://www.amazon.science/publications/temporal-generalization-for-spoken-language-understanding\\" target=\\"_blank\\">Temporal generalization for spoken language understanding</a></ins><br />\\nJudith Gaspers, Anoop Kumar, Greg Ver Steeg, Aram Galstyan</p>\n<h4><a id=\\"Data_augmentation_19\\"></a><strong>Data augmentation</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/constraining-word-alignments-with-posterior-regularization-for-label-transfer\\" target=\\"_blank\\">Constraining word alignments with posterior regularization for label transfer</a></ins><br />\\nKevin Martin Jose, Thomas Gueudré</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/69c1f3d5a9884de58d9a4750f20c9fbb_%E4%B8%8B%E8%BD%BD%20%281%29.jpg\\" alt=\\"下载 1.jpg\\" /></p>\n<p>An example of the difficulty in using word alignment to transfer textual labels from one language to another. In English, the article “the” is assigned the label “o”, for “other”; in French, the abbreviated article is combined with its noun, and both receive the same label (“type”). From “<ins><a href=\\"https://www.amazon.science/publications/constraining-word-alignments-with-posterior-regularization-for-label-transfer\\" target=\\"_blank\\">Constraining word alignments with posterior regularization for label transfer</a></ins>”.</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/controlled-data-generation-via-insertion-operations-for-nlu\\" target=\\"_blank\\">Controlled data generation via insertion operations for NLU</a></ins><br />\\nManoj Kumar, Haidar Khan, Yuval Merhav, Wael Hamza, Anna Rumshisky, Rahul Gupta</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/efficient-semi-supervised-consistency-training-for-natural-language-understanding\\" target=\\"_blank\\">Efficient semi supervised consistency training for natural language understanding</a></ins><br />\\nGeorge Leung, Joshua Tan</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/learning-to-generate-examples-for-semantic-processing-tasks\\" target=\\"_blank\\">Learning to generate examples for semantic processing tasks</a></ins><br />\\nDanilo Croce, Simone Filice, Giuseppe Castellucci, Roberto Basili</p>\n<h4><a id=\\"Dialogue_37\\"></a><strong>Dialogue</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/learning-dialogue-representations-from-consecutive-utterances\\" target=\\"_blank\\">Learning dialogue representations from consecutive utterances</a></ins><br />\\nZhihan Zhou, Dejiao Zhang, Wei Xiao, Nicholas Dingwall, Xiaofei Ma, Andrew O. Arnold, Bing Xiang</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/massive-scale-decoding-for-text-generation-using-lattices\\" target=\\"_blank\\">Massive-scale decoding for text generation using lattices</a></ins><br />\\nJiacheng Xu, Siddhartha Reddy Jonnalagadda, Greg Durrett</p>\n<h4><a id=\\"Entity_linking_resolution_and_typing_45\\"></a><strong>Entity linking, resolution, and typing</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/contrastive-representation-learning-for-cross-document-coreference-resolution-of-events-and-entities\\" target=\\"_blank\\">Contrastive representation learning for cross-document coreference resolution of events and entities</a></ins><br />\\nBenjamin Hsu, Graham Horwood</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/improving-entity-disambiguation-by-reasoning-over-a-knowledge-base\\" target=\\"_blank\\">Improving entity disambiguation by reasoning over a knowledge base</a></ins><br />\\nTom Ayoola, Joseph Fisher, Andrea Pierleoni</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/refined-an-efficient-zero-shot-capable-approach-to-end-to-end-entity-linking\\" target=\\"_blank\\">ReFinED: An efficient zero-shot-capable approach to end-to-end entity linking</a></ins><br />\\nTom Ayoola, Shubhi Tyagi, Joseph Fisher, Christos Christodoulopoulos, Andrea Pierleoni</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/instilling-type-knowledge-in-language-models-via-multi-task-qa\\" target=\\"_blank\\">Instilling type knowledge in language models via multi-task QA</a></ins><br />\\nShuyang Li, Mukund Sridhar, Chandana Satya Prakash, Jin Cao, Wael Hamza, Julian McAuley</p>\n<h4><a id=\\"Explainable_AI_59\\"></a><strong>Explainable AI</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/entailment-tree-explanations-via-iterative-retrieval-generation-reasoner\\" target=\\"_blank\\">Entailment tree explanations via iterative retrieval-generation reasoner</a></ins><br />\\nDanilo Neves Ribeiro, Shen Wang, Xiaofei Ma, Rui Dong, Xiaokai Wei, Henry Zhu, Xinchi Chen, Zhiheng Huang, Peng Xu, Andrew O. Arnold, Dan Roth</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/locally-aggregated-feature-attribution-on-natural-language-model-understanding\\" target=\\"_blank\\">Locally aggregated feature attribution on natural language model understanding</a></ins><br />\\nSheng Zhang, Jin Wang, Haitao Jiang, Rui Song</p>\n<p>Extreme multilabel classificatio</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/7ba03f19570546d9a6d2511fa3dc5e43_%E4%B8%8B%E8%BD%BD%20%282%29.jpg\\" alt=\\"下载 2.jpg\\" /></p>\n<p>In “Entailment tree explanations via iterative retrieval-generation reasoner”, Amazon researchers propose a method for explaining the outputs of large language models by logically recombining premises extracted from supporting textual evidence.</p>\n<h4><a id=\\"Extreme_multilabel_classification_73\\"></a><strong>Extreme multilabel classification</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/augmenting-training-data-for-massive-semantic-matching-models-in-low-traffic-e-commerce-stores\\" target=\\"_blank\\">Augmenting training data for massive semantic matching models in low-traffic e-commerce stores</a></ins><br />\\nAshutosh Joshi, Shankar Vishwanath, Choon Hui Teo, Vaclav Petricek, Vishy Vishwanathan, Rahul Bhagat, Jonathan May</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/extreme-zero-shot-learning-for-extreme-text-classification\\" target=\\"_blank\\">Extreme zero shot learning for extreme text classification</a></ins><br />\\nYuanhao Xiong, Wei-Cheng Chang, Cho-Jui Hsieh, Hsiang-Fu Yu, Inderjit S. Dhillon</p>\n<h4><a id=\\"Extreme_multilabel_classification_81\\"></a><strong>Extreme multilabel classification</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/augmenting-training-data-for-massive-semantic-matching-models-in-low-traffic-e-commerce-stores\\" target=\\"_blank\\">Augmenting training data for massive semantic matching models in low-traffic e-commerce stores</a></ins><br />\\nAshutosh Joshi, Shankar Vishwanath, Choon Hui Teo, Vaclav Petricek, Vishy Vishwanathan, Rahul Bhagat, Jonathan May</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/extreme-zero-shot-learning-for-extreme-text-classification\\" target=\\"_blank\\">Extreme zero shot learning for extreme text classification</a></ins><br />\\nYuanhao Xiong, Wei-Cheng Chang, Cho-Jui Hsieh, Hsiang-Fu Yu, Inderjit S. Dhillon</p>\n<h4><a id=\\"Federated_learning_89\\"></a><strong>Federated learning</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/federated-learning-with-noisy-user-feedback\\" target=\\"_blank\\">Federated learning with noisy user feedback</a></ins><br />\\nRahul Sharma, Anil Ramakrishna, Ansel MacLaughlin, Anna Rumshisky, Jimit Majmudar, Clement Chung, Salman Avestimehr, Rahul Gupta</p>\n<h4><a id=\\"Keyword_spotting_94\\"></a><strong>Keyword spotting</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/ab-ba-analysis-a-framework-for-estimating-keyword-spotting-recall-improvement-while-maintaining-audio-privacy\\" target=\\"_blank\\">AB/BA analysis: A framework for estimating keyword spotting recall improvement while maintaining audio privacy</a></ins><br />\\nRaphael Petegrosso, Vasistakrishna Baderdinni, Thibaud Senechal, Benjamin L. Bullough</p>\n<h4><a id=\\"Machine_translation_99\\"></a><strong>Machine translation</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/cocoa-mt-a-dataset-and-benchmark-for-contrastive-controlled-mt-with-application-to-formality\\" target=\\"_blank\\">CoCoA-MT: A dataset and benchmark for contrastive controlled MT with application to formality</a></ins><br />\\nMaria Nadejde, Anna Currey, Benjamin Hsu, Xing Niu, Marcello Federico, Georgiana Dinu</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/a586f48eb62c4cb99cddb42a3c058dd1_%E4%B8%8B%E8%BD%BD%20%283%29.jpg\\" alt=\\"下载 3.jpg\\" /></p>\n<p>In federated learning, distributed copies of a neural network are trained locally, and only their updates (red) are sent to a central model. “<ins><a href=\\"https://www.amazon.science/publications/training-mixed-domain-translation-models-via-federated-learning\\" target=\\"_blank\\">Training mixed-domain translation models via federated learning</a></ins>” introduces a technique called dynamic pulling, in which distributed models with large shifts in parameter values between training rounds (lower left) see their parameters pulled into the central model separately from those of models with smaller shifts.</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/the-devil-is-in-the-details-on-the-pitfalls-of-vocabulary-selection-in-neural-machine-translation\\" target=\\"_blank\\">The devil is in the details: On the pitfalls of vocabulary selection in neural machine translation</a></ins><br />\\nTobias Domhan, Eva Hasler, Ke Tran, Sony Trenous, Bill Byrne, Felix Hieber</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/training-mixed-domain-translation-models-via-federated-learning\\" target=\\"_blank\\">Training mixed-domain translation models via federated learning</a></ins><br />\\nPeyman Passban, Tanya G. Roosta, Rahul Gupta, Ankit Chadha, Clement Chung</p>\n<h4><a id=\\"Multitask_learning_113\\"></a><strong>Multitask learning</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/asynchronous-convergence-in-multi-task-learning-via-knowledge-distillation-from-converged-tasks\\" target=\\"_blank\\">Asynchronous convergence in multi-task learning via knowledge distillation from converged tasks</a></ins><br />\\nWeiyi Lu, Sunny Rajagopalan, Priyanka Nigam, Jaspreet Singh, Xiaodi Sun, Yi Xu, Belinda Zeng, Trishul Chilimbi</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/exploring-the-role-of-task-transferability-in-large-scale-multi-task-learning\\" target=\\"_blank\\">Exploring the role of task transferability in large-scale multi-task learning</a></ins><br />\\nVishakh Padmakumar, Leonard Lausen, Miguel Ballesteros, Sheng Zha, He He, George Karypis</p>\n<h4><a id=\\"Namedentity_recognition_120\\"></a><strong>Named-entity recognition</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/dynamic-gazetteer-integration-in-multilingual-models-for-cross-lingual-and-cross-domain-named-entity-recognition\\" target=\\"_blank\\">Dynamic gazetteer integration in multilingual models for cross-lingual and cross-domain named entity recognition</a></ins><br />\\nBesnik Fetahu, Anjie Fang, Oleg Rokhlenko, Shervin Malmasi</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/ner-mqmrc-formulating-named-entity-recognition-as-multi-question-machine-reading-comprehension\\" target=\\"_blank\\">NER-MQMRC: Formulating named entity recognition as multi question machine reading comprehension</a></ins><br />\\nAnubhav Shrimal, Avi Jain, Kartik Mehta, Promod Yenigalla</p>\n<h4><a id=\\"Question_answering_127\\"></a><strong>Question answering</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/answer-consolidation-formulation-and-benchmarking\\" target=\\"_blank\\">Answer consolidation: Formulation and benchmarking</a></ins><br />\\nWenxuan Zhou, Qiang Ning, Heba Elfardy, Kevin Small, Muhao Chen</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/paragraph-based-transformer-pre-training-for-multi-sentence-inference\\" target=\\"_blank\\">Paragraph-based transformer pre-training for multi-sentence inference</a></ins><br />\\nLuca Di Liello, Siddhant Garg, Luca Soldaini, Alessandro Moschitti</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/perkgqa-question-answering-over-personalized-knowledge-graphs\\" target=\\"_blank\\">PerKGQA: Question answering over personalized knowledge graphs</a></ins><br />\\nRitam Dutt, Kasturi Bhattacharjee, Rashmi Gangadharaiah, Dan Roth, Carolyn Penstein Rosé</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/product-answer-generation-from-heterogeneous-sources-a-new-benchmark-and-best-practices\\" target=\\"_blank\\">Product answer generation from heterogeneous sources: A new benchmark and best practices</a></ins><br />\\nXiaoyu Shen, Gianni Barlacchi, Marco Del Tredici, Weiwei Cheng, Adria de Gispert, Bill Byrne</p>\n<h4><a id=\\"Recommender_systems_140\\"></a><strong>Recommender systems</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/ceres-pretraining-of-graph-conditioned-transformer-for-semi-structured-session-data\\" target=\\"_blank\\">CERES: Pretraining of graph-conditioned transformer for semi-structured session data</a></ins><br />\\nRui Feng, Chen Luo, Qingyu Yin, Bing Yin, Tuo Zhao, Chao Zhang</p>\n<h4><a id=\\"Selflearning_144\\"></a><strong>Self-learning</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/fpi-failure-point-isolation-in-large-scale-conversational-assistants\\" target=\\"_blank\\">FPI: Failure point isolation in large-scale conversational assistants</a></ins><br />\\nRinat Khaziev, Usman Shahid, Tobias Röding, Rakesh Chada, Emir Kapanci, Pradeep Natarajan</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/scalable-and-robust-self-learning-for-skill-routing-in-large-scale-conversational-ai-systems\\" target=\\"_blank\\">Scalable and robust self-learning for skill routing in large-scale conversational AI systems</a></ins><br />\\nMohammad Kachuee, Jinseok Nam, Sarthak Ahuja, Jin-Myung Won, Sungjin Lee</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/self-aware-feedback-based-self-learning-in-large-scale-conversational-ai\\" target=\\"_blank\\">Self-aware feedback-based self-learning in large-scale conversational AI</a></ins><br />\\nPragaash Ponnusamy, Clint Solomon Mathialagan, Gustavo Aguilar, Chengyuan Ma, Chenlei (Edward) Guo</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/c82b352fba09411aa4bdfceab8f83f30_%E4%B8%8B%E8%BD%BD%20%284%29.jpg\\" alt=\\"下载 4.jpg\\" /></p>\n<p>In “FPI: Failure point isolation in large-scale conversational assistants”, Amazon researchers propose a method for deducing where in a conversational agent’s processing pipeline an error has occurred.</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/b4240172d9f04d25b9dc656b15af5e4f_%E4%B8%8B%E8%BD%BD%20%285%29.jpg\\" alt=\\"下载 5.jpg\\" /><br />\\nAn example of task-oriented semantic parsing, which converts natural language into a formal representation that an AI agent can act on. From “<a href=\\"https://www.amazon.science/publications/compositional-task-oriented-parsing-as-abstractive-question-answering\\" target=\\"_blank\\">Compositional task-oriented parsing as abstractive question answering</a>”.</p>\\n<h4><a id=\\"Semantic_parsing_162\\"></a><strong>Semantic parsing</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/compositional-task-oriented-parsing-as-abstractive-question-answering\\" target=\\"_blank\\">Compositional task oriented parsing as abstractive question answering</a></ins><br />\\nWenting Zhao, Konstantine Arkoudas, Weiqi Sun, Claire Cardie</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/seqzero-few-shot-compositional-semantic-parsing-with-sequential-prompts-and-zero-shot-models\\" target=\\"_blank\\">SeqZero: Few-shot compositional semantic parsing with sequential prompts and zero-shot models</a></ins><br />\\nJingfeng Yang, Haoming Jiang, Qingyu Yin, Danqing Zhang, Bing Yin, Diyi Yang</p>\n<h4><a id=\\"Task_adaptation_169\\"></a><strong>Task adaptation</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/attention-fusion-a-light-yet-efficient-late-fusion-mechanism-for-task-adaptation-in-nlu\\" target=\\"_blank\\">Attention fusion: A light yet efficient late fusion mechanism for task adaptation in NLU</a></ins><br />\\nJin Cao, Chandana Satya Prakash, Wael Hamza</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/empowering-parameter-efficient-transfer-learning-by-recognizing-the-kernel-structure-in-attention\\" target=\\"_blank\\">Empowering parameter-efficient transfer learning by recognizing the kernel structure in attention</a></ins><br />\\nYifan Chen, Devamanyu Hazarika, Mahdi Namazifar, Yang Liu, Di Jin, Dilek Hakkani-Tür</p>\n<h4><a id=\\"Text_mining_176\\"></a><strong>Text mining</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/distantly-supervised-aspect-clustering-and-naming-for-e-commerce-reviews\\" target=\\"_blank\\">Distantly supervised aspect clustering and naming for e-commerce reviews</a></ins><br />\\nPrateek Sircar, Aniket Chakrabarti, Deepak Gupta, Anirban Majumdar</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/efficient-few-shot-fine-tuning-for-opinion-summarization\\" target=\\"_blank\\">Efficient few-shot fine-tuning for opinion summarization</a></ins><br />\\nArthur Bražinskas, Ramesh Nallapati, Mohit Bansal, Markus Dreyer</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/factgraph-evaluating-factuality-in-summarization-with-semantic-graph-representations\\" target=\\"_blank\\">FactGraph: Evaluating factuality in summarization with semantic graph representations</a></ins><br />\\nLeonardo F. R. Ribeiro, Mengwen Liu, Iryna Gurevych, Markus Dreyer, Mohit Bansal</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/enhanced-knowledge-selection-for-grounded-dialogues-via-document-semantic-graphs\\" target=\\"_blank\\">Enhanced knowledge selection for grounded dialogues via document semantic graphs</a></ins><br />\\nSha Li, Madhi Namazifar, Di Jin, Mohit Bansal, Heng Ji, Yang Liu, Dilek Hakkani-Tür</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/retrieval-augmented-multilingual-keyphrase-generation-with-retriever-generator-iterative-training\\" target=\\"_blank\\">Retrieval-augmented multilingual keyphrase generation with retriever-generator iterative training</a></ins><br />\\nYifan Gao, Qingyu Yin, Zheng Li, Rui Meng, Tong Zhao, Bing Yin, Irwin King, Michael R. Lyu</p>\n<p><ins><a href=\\"https://www.amazon.science/publications/what-do-users-care-about-detecting-actionable-insights-from-user-feedback\\" target=\\"_blank\\">What do users care about? Detecting actionable insights from user feedback</a></ins><br />\\nKasturi Bhattacharjee, Rashmi Gangadharaiah, Kathleen McKeown, Dan Roth</p>\n<p><img src=\\"https://dev-media.amazoncloud.cn/973de8ce9d254edaab27906bf1334fef_%E4%B8%8B%E8%BD%BD%20%286%29.jpg\\" alt=\\"下载 6.jpg\\" /></p>\n<p>An example of how a conversational agent might incorporate facts gleaned form online sources (white boxes) into its conversational replies (blue boxes). From “<ins><a href=\\"https://www.amazon.science/publications/enhanced-knowledge-selection-for-grounded-dialogues-via-document-semantic-graphs\\" target=\\"_blank\\">Enhanced knowledge selection for grounded dialogues via document semantic graphs</a></ins>”.</p>\n<h4><a id=\\"Texttospeech_199\\"></a><strong>Text-to-speech</strong></h4>\\n<p><ins><a href=\\"https://www.amazon.science/publications/empathic-machines-using-intermediate-features-as-levers-to-emulate-emotions-in-text-to-speech-systems\\" target=\\"_blank\\">Empathic machines: using intermediate features as levers to emulate emotions in text-to-speech systems</a></ins><br />\\nSaiteja Kosgi, Sarath Sivaprasad, Niranjan Pedanekar, Anil Nelakanti, Vineet Gandhi</p>\n<p>ABOUT THE AUTHOR</p>\n<h4><a id=\\"Staff_writer_205\\"></a><strong>Staff writer</strong></h4>\n"}