SPARTA employs sparse Transformer matching with inverted index for efficient open-domain QA, outperforming dense vector methods.

Tiancheng Zhao, Xiaopeng Lu, Kyusong Lee

Proposed ClariQ task with deep learning models for generating and ranking clarifying questions in open-domain dialogue.

Mohammad Aliannejadi, Julia Kiseleva, Aleksandr Chuklin et al.

This work adapts three large-scale controllable dialogue architectures to manage about 200 styles, improving style consistency and diversity.

Eric Michael Smith, Diana Gonzalez-Rico, Emily Dinan et al.

This study applies classical machine learning and Seq2Seq models (LSTM, Transformer) to Minangkabau sentiment analysis and machine translation, constructing the first relevant corpora.

Fajri Koto, Ikhwan Koto

GeDi uses smaller LMs as generative discriminators to guide large LMs, enhancing safety and control.

Ben Krause, Akhilesh Deepak Gotmare, Bryan McCann et al.

SpanBasedSP predicts span trees to improve compositional generalization, boosting accuracy from 61.0 to 88.9 on key datasets.

Jonathan Herzig, Jonathan Berant

Introduced IndoNLU benchmark with 12 tasks, trained IndoBERT models, outperforming multilingual models by significant margins.

Bryan Wilie, Karissa Vincentio, Genta Indra Winata et al.

KILT benchmark unifies Wikipedia-based knowledge tasks with dense retrieval and Seq2Seq models, outperforming task-specific approaches.

Fabio Petroni, Aleksandra Piktus, Angela Fan et al.

MTOP introduces a multilingual semantic parsing dataset with 100k utterances across 6 languages and 11 domains, achieving +6.3 Slot F1 with XLM-R models.

Haoran Li, Abhinav Arora, Shuohui Chen et al.

Survey of active learning for text classification using DNNs, enhancing performance or reducing labeled data.

Christopher Schröder, Andreas Niekler

Proposes InfoXLM, an information-theoretic framework utilizing mutual information maximization and contrastive learning to significantly improve cross-lingual transfer.

Zewen Chi, Li Dong, Furu Wei et al.

This paper introduces LaBSE, a multilingual sentence embedding model combining MLM, TLM, dual encoders, and additive margin softmax, achieving 83.7% accuracy across 112 languages.

Fangxiaoyu Feng, Yinfei Yang, Daniel Cer et al.

GShard combines conditional computation and automatic sharding to train a 600-billion-parameter multilingual Transformer efficiently, reducing costs and boosting performance.

Dmitry Lepikhin, HyoukJoong Lee, Yuanzhong Xu et al.

MARGE introduces a retrieval-based pretraining method for multilingual multi-document paraphrasing, achieving strong zero-shot performance on translation and summarization.

Mike Lewis, Marjan Ghazvininejad, Gargi Ghosh et al.

This survey reviews three categories of NLG evaluation metrics: human, automatic, and machine-learned, analyzing their strengths, weaknesses, and future directions.

Asli Celikyilmaz, Elizabeth Clark, Jianfeng Gao

Proposed neural models (Transformers and multilingual RNNs) with data augmentation achieved over 90% accuracy across 45 typologically diverse languages.

Ekaterina Vylomova, Jennifer White, Elizabeth Salesky et al.

wav2vec 2.0 achieves speech recognition via self-supervised learning, reaching 1.8/3.3 WER.

Alexei Baevski, Henry Zhou, Abdelrahman Mohamed et al.

CRISS iteratively mines pseudo-parallel sentences using pretrained sentence embeddings, boosting unsupervised translation BLEU by 2.4 on average.

Chau Tran, Yuqing Tang, Xian Li et al.

Pretrained FinBERT on 4.9B tokens, outperforming generic BERT in financial sentiment tasks with up to 15.6% accuracy gain.

Yi Yang, Mark Christopher Siy UY, Allen Huang

ETHOS dataset offers hate speech detection based on YouTube and Reddit comments, using binary and multi-label classification.

Ioannis Mollas, Zoe Chrysopoulou, Stamatis Karlos et al.