]> 2021-03-01 2021-03-01T13:26:47Z Ernest ILISCA (0000-0002-3842-586X) - ORCID | Connecting Research and Researchers 2021-03-30T19:21:25Z 2021-03-30 > Weaviate is an API-based vector search engine with a graph data model that allows users to add data objects as graph nodes and (automatically or manually) add (machine learning) vectors to represent the nodes. Weaviate can be used for use cases ranging from similarity search to filtering out redundant information (i.e., deduplication) and from image search to enterprise NLP-based search. Weaviate Vector Search Engine | SeMI Technologies Renault group at Hugging Face 2021-03-18 2021-03-18T14:27:39Z 2021-03-09 BabelNet | Le plus grand dictionnaire encyclopédique et réseau sémantique BabelNet 5 classifies synsets (synonym sets) into concepts (e.g. university) and named entities (e.g. Sapienza University of Rome) with manual check of 100k synsets. 2021-03-09T08:11:17Z 2021-03-28 2021-03-28T16:16:09Z à l’époque d’Abraham la première religion monothéiste était loin d’être une entité cohérente… La plupart des Juifs ont, par exemple, longtemps adoré des dieux païens. Par ailleurs, nombre d’entre eux pensaient que Dieu avait une épouse qui était elle-même une idole vénérée Les secrets révélés de la Bible | ARTE 2021-03-16T10:36:10Z Israël dévoile un manuscrit biblique vieux de deux mille ans 2021-03-16 2021-03-04T08:19:59Z 2021-03-04 asahi417/tner: Language model finetuning on NER 2021-03-13T09:56:42Z Un an d’épidémie de Covid-19 : retrouver le sens du long terme 2021-03-13 > En France, deux défauts majeurs ont donné des longueurs d’avance à la pandémie : l’effacement de notre culture de la prévention et l’affaissement de notre science. Autant de signes tangibles d’un déclin inquiétant pour l’avenir et le rayonnement de notre pays. **Très inquiétant** pour un pays couvert de centrales nucléaires "Text is the API for humans" 2021-03-20T17:06:34Z 2021-03-20 SentenceTransformers Documentation 2021-03-25 2021-03-25T19:05:01Z 2010.12321 2020-10-23T11:57:33Z Moussa Kamal Eddine Moussa Kamal Eddine [2010.12321] BARThez: a Skilled Pretrained French Sequence-to-Sequence Model [On HuggingFace](doc:2021/03/barthez_transformers_4_5_0_de) ; [GitHub](https://github.com/moussaKam/BARThez) ([same author](doc:?uri=https%3A%2F%2Fwww2018.thewebconf.org%2Fprogram%2Ftutorials-track%2Ftutorial-213%2F)) 2021-03-31T23:00:47Z 2021-03-31 BARThez — transformers 4.5.0.dev0 documentation Inductive transfer learning has taken the entire NLP field by storm, with models such as BERT and BART setting new state of the art on countless NLU tasks. However, most of the available models and research have been conducted for English. In this work, we introduce BARThez, the first large-scale pretrained seq2seq model for French. Being based on BART, BARThez is particularly well-suited for generative tasks. We evaluate BARThez on five discriminative tasks from the FLUE benchmark and two generative tasks from a novel summarization dataset, OrangeSum, that we created for this research. We show BARThez to be very competitive with state-of-the-art BERT-based French language models such as CamemBERT and FlauBERT. We also continue the pretraining of a multilingual BART on BARThez' corpus, and show our resulting model, mBARThez, to significantly boost BARThez' generative performance. Code, data and models are publicly available. 2021-03-31 Antoine J. -P. Tixier 2021-03-31T19:08:05Z 2021-02-09T09:31:57Z BARThez: a Skilled Pretrained French Sequence-to-Sequence Model Michalis Vazirgiannis 2021-03-09T08:08:07Z Christopher Dengsø sur Twitter : "The moderation API now detects addresses in addition to other personal details." 2021-03-09 Roam vs Obsidian - Which one should you use? | Medium 2021-03-13 2021-03-13T11:27:36Z huggingface/awesome-papers: Papers & presentation materials from Hugging Face's internal science day 2021-03-26T01:58:09Z 2021-03-26 film by Ken Loach 2021-03-17T22:31:59Z 2021-03-17 The Angels' Share 2021-03-15T17:57:01Z 2021-03-15 New pipeline for zero-shot text classification - 🤗Transformers - Hugging Face Forums Same author: [Zero-shot classifier distillation at master · huggingface/transformers](doc:2021/02/zero_shot_classifier_distillati) 2021-03-26T01:49:42Z Rodrigo Nogueira 1901.04085 Rodrigo Nogueira sur Twitter : "Slides of our WSDM 2021 tutorial "Pretrained Transformers for Text Ranking: BERT and Beyond" 2021-03-09T08:09:28Z 2021-03-09 Recently, neural models pretrained on a language modeling task, such as ELMo (Peters et al., 2017), OpenAI GPT (Radford et al., 2018), and BERT (Devlin et al., 2018), have achieved impressive results on various natural language processing tasks such as question-answering and natural language inference. In this paper, we describe a simple re-implementation of BERT for query-based passage re-ranking. Our system is the state of the art on the TREC-CAR dataset and the top entry in the leaderboard of the MS MARCO passage retrieval task, outperforming the previous state of the art by 27% (relative) in MRR@10. The code to reproduce our results is available at https://github.com/nyu-dl/dl4marco-bert 2020-04-14T14:57:40Z 2019-01-13T23:27:58Z Rodrigo Nogueira Kyunghyun Cho Passage Re-ranking with BERT [1901.04085] Passage Re-ranking with BERT 2021-03-26 a simple re-implementation of BERT for query-based passage re-ranking ["Slides of our WSDM 2021 tutorial "Pretrained Transformers for Text Ranking: BERT and Beyond"](doc:2021/03/rodrigo_nogueira_sur_twitter_) Sur le catamaran Nomade des mers, l'ingénieur Corentin de Chatelperron voyage à travers le monde à la découverte des perspectives de la low-tech. 2021-03-14T17:55:53Z 2021-03-14 Nomade des mers, les escales de l'innovation Comment Bayer a fait pression sur le Mexique pour empêcher l’interdiction du glyphosate 2021-03-27 2021-03-27T21:10:08Z 2021-03-26T09:57:36Z Découverte d'un artefact datant de 24 000 ans à Vale da Pedra Furada, Piauí, Brésil | INSHS 2021-03-26 2021-03-02 2021-03-02T23:17:24Z Le Dit du Genji — Wikipédia Beliz Gunel [2010.02194] Self-training Improves Pre-training for Natural Language Understanding Vishrav Chaudhary Jingfei Du Edouard Grave Unsupervised pre-training has led to much recent progress in natural language understanding. In this paper, we study self-training as another way to leverage unlabeled data through semi-supervised learning. To obtain additional data for a specific task, we introduce SentAugment, a data augmentation method which computes task-specific query embeddings from labeled data to retrieve sentences from a bank of billions of unlabeled sentences crawled from the web. Unlike previous semi-supervised methods, our approach does not require in-domain unlabeled data and is therefore more generally applicable. Experiments show that self-training is complementary to strong RoBERTa baselines on a variety of tasks. Our augmentation approach leads to scalable and effective self-training with improvements of up to 2.6% on standard text classification benchmarks. Finally, we also show strong gains on knowledge-distillation and few-shot learning. 2021-03-12T06:17:22Z 2020-10-05T17:52:25Z Self-training Improves Pre-training for Natural Language Understanding Alexis Conneau 2021-03-12 2020-10-05T17:52:25Z Onur Celebi Michael Auli Jingfei Du 2010.02194 Ves Stoyanov Teddy Koker sur Twitter : "Torchsort, an implementation of "Fast Differentiable Sorting and Ranking" in PyTorch" 2021-03-25T17:24:25Z 2021-03-25 2021-03-19T13:32:54Z 2021-03-19 Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation | Frontiers in Computational Neuroscience cf. [Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation](doc:2021/03/equilibrium_propagation_bridgi) 2021-03-19 2021-03-19T13:36:13Z Julie Grollier sur Twitter : "EqSpike: spike-driven equilibrium propagation for neuromorphic implementations" 2021-03-12T18:44:46Z 2021-03-12 Hugging Face sur Twitter : "Fine-Tuning @facebookai's Wav2Vec2 for Speech Recognition is now possible in Transformers Not only for English but for 53 Languages L’hydrogène, une solution incertaine pour la mobilité 2021-03-27 2021-03-27T14:39:20Z Vaccins contre le Covid-19 : pourquoi la France accuse-t-elle un tel retard ? 2021-03-02T12:16:38Z 2021-03-02 Crédits publics en R&D pour la santé, de 2011 à 2018 : USA +8%, Allemagne +11% (->6 milliards), France -28% (-> un peu plus de 2milliards) (source OCDE) [video][Vaccins contre le Covid-19 : pourquoi la France accuse-t-elle un tel retard ?](doc:2021/03/vaccins_contre_le_covid_19_po) 2021-03-02 University-industry collaboration in R&D - World Economic Forum 2021-03-02T12:13:55Z 2021-03-30 2021-03-30T00:43:13Z Pulling Turtle RDF triples from the Google Knowledge Graph 2021-03-30T00:35:13Z 2021-03-23T22:53:09Z 2021-03-23T22:53:09Z Chenyan Xiong 2103.12876 [2103.12876] Complex Factoid Question Answering with a Free-Text Knowledge Graph Complex Factoid Question Answering with a Free-Text Knowledge Graph Chen Zhao > delft builds a free-text knowledge graph from Wikipedia, with entities as nodes and sentences in which entities co-occur as edges Xin Qian 2021-03-30 Jordan Boyd-Graber We introduce DELFT, a factoid question answering system which combines the nuance and depth of knowledge graph question answering approaches with the broader coverage of free-text. DELFT builds a free-text knowledge graph from Wikipedia, with entities as nodes and sentences in which entities co-occur as edges. For each question, DELFT finds the subgraph linking question entity nodes to candidates using text sentences as edges, creating a dense and high coverage semantic graph. A novel graph neural network reasons over the free-text graph-combining evidence on the nodes via information along edge sentences-to select a final answer. Experiments on three question answering datasets show DELFT can answer entity-rich questions better than machine reading based models, bert-based answer ranking and memory networks. DELFT's advantage comes from both the high coverage of its free-text knowledge graph-more than double that of dbpedia relations-and the novel graph neural network which reasons on the rich but noisy free-text evidence. Chen Zhao [Try It](doc:2021/03/wikidata_browser) 2021-03-06T16:15:21Z Wikidata browser 2021-03-06 [GitHub](doc:2021/03/ringgaard_sling_sling_a_natu) ringgaard/sling: SLING - A natural language frame semantics parser 2021-03-08T08:20:52Z 2021-03-08