]> How to do Unsupervised Clustering with Keras – Chengwei Zhang – Medium 2018-06-09T09:23:35Z 2018-06-09 Python Data Science Handbook (Jake VanderPlas) 2018-06-29T10:50:30Z 2018-06-29 2018-06-08 2018-06-08T15:23:26Z A Tri-Partite Neural Document Language Model for Semantic Information Retrieval (2018 - ESWC conference) from the abstract: Previous work in information retrieval have shown that using evidence, such as concepts and relations, from external knowledge sources could enhance the retrieval performance... This paper presents a new tri-partite neural document language framework that leverages explicit knowledge to jointly constrain word, concept, and document learning representations to tackle a number of issues including polysemy and granularity mismatch. "In @TensorFlow 1.9, it is much easier to use Keras with the Data API: just pass data iterators, specify the number of steps per epoch, and you're good to go! Plus it works in both graph mode and eager mode, kudos to the TF team!… https://t.co/EH3hY50N0o" 2018-06-10T09:18:12Z 2018-06-10 Aurélien Geron sur Twitter : "In @TensorFlow 1.9, it is much easier to use Keras with the Data API..." Extending NLP - Stardog 2018-06-14T13:21:28Z 2018-06-14 How to extend Stardog’s NLP pipeline 2018-06-12 Improving Language Understanding with Unsupervised Learning 2018-06-12T09:16:15Z > can we develop one model, train it in an unsupervised way on a large amount of data, and then fine-tune the model to achieve good performance on many different tasks? Our results indicate that this approach works surprisingly well; the same core model can be fine-tuned for very different tasks with minimal adaptation. a scalable, task-agnostic system based on a combination of two existing ideas: transformers and unsupervised pre-training. unsupervised generative pre-training of language models followed by discriminative fine-tunning. la sonde Parker Solar Probe partira de Floride cet été pour s’avancer au plus près de l’étoile Un ticket pour le Soleil | CNRS Le journal 2018-06-28T01:30:27Z 2018-06-28 > a framework for training classifiers in which an **annotator** provides a natural language explanation for each labeling decision. A semantic parser converts these explanations into programmatic labeling functions that generate noisy labels for an arbitrary amount of unlabeled data, which is used to train a classifier. On three relation extraction tasks, we find that users are able to train classifiers with comparable F1 scores from 5–100 faster by providing explanations instead of just labels 2018-06-23 2018-06-23T00:55:49Z Training Classifiers with Natural Language Explanations Yann LeCun Ruslan Salakhutdinov [1806.05662] GLoMo: Unsupervisedly Learned Relational Graphs as Transferable Representations William W. Cohen Jake Zhao Modern deep transfer learning approaches have mainly focused on learning generic feature vectors from one task that are transferable to other tasks, such as word embeddings in language and pretrained convolutional features in vision. However, these approaches usually transfer unary features and largely ignore more structured graphical representations. This work explores the possibility of learning generic latent relational graphs that capture dependencies between pairs of data units (e.g., words or pixels) from large-scale unlabeled data and transferring the graphs to downstream tasks. Our proposed transfer learning framework improves performance on various tasks including question answering, natural language inference, sentiment analysis, and image classification. We also show that the learned graphs are generic enough to be transferred to different embeddings on which the graphs have not been trained (including GloVe embeddings, ELMo embeddings, and task-specific RNN hidden unit), or embedding-free units such as image pixels. Modern deep transfer learning approaches have mainly focused on learning generic feature vectors from one task that are transferable to other tasks, such as word embeddings in language and pretrained convolutional features in vision. However, these approaches usually transfer unary features and largely ignore more structured graphical representations. This work explores the possibility of learning generic latent relational graphs that capture dependencies between pairs of data units (e.g., words or pixels) from large-scale unlabeled data and transferring the graphs to downstream tasks. 2018-06-14T17:41:19Z Zhilin Yang 2018-06-23T00:58:21Z Bhuwan Dhingra Zhilin Yang 2018-06-23 2018-07-02T20:24:33Z GLoMo: Unsupervisedly Learned Relational Graphs as Transferable Representations 1806.05662 Kaiming He 2018-06-08 2018-06-08T13:15:55Z Niger Islamic State hostage: 'They want to kill foreign soldiers' | The Guardian 2018-06-21 2018-06-21T13:12:11Z Taskonomy | Stanford 2018-06-07 2018-06-07T23:58:17Z 2017 Deloitte State of Cognitive Survey > We cast all tasks as question answering over a context. [arxiv](https://arxiv.org/abs/1806.08730) [slides](doc:2021/01/the_natural_language_decathlon_) The Natural Language Decathlon: Multitask Learning as Question Answering (2018) Salesforce research 2018-06-21 2018-06-21T12:55:41Z Télérama sur Twitter : "Bansky Paris Invasion ! Venu incognito comme toujours, le célèbre street artist a déjà laissé deux œuvres qui témoignent de son passage dans les 18e et 19e arrondissements. Elles livrent un message fort au gouvernement français. #Banksy #Paris #streetart https://t.co/AbiT6RfsEw… https://t.co/t302gOpZri" 2018-06-24T20:03:47Z 2018-06-24 "Bansky Paris Invasion !" 2018-06-17T12:27:17Z 2018-06-17 Why You Don’t Need Data Scientists – Kurt Cagle – Medium sebastianruder/NLP-progress: Repository to track the progress in Natural Language Processing (NLP), including the datasets and the current state-of-the-art for the most common NLP tasks. 2018-06-23 2018-06-23T01:04:30Z Markup for Autos - schema.org 2018-06-19T10:55:59Z 2018-06-19 Why do low-dimensional word vectors exist? > a text corpus is imagined as being generated by a random walk in a latent variable space, and the word production is via a loglinear distribution. This model is shown to imply several empirically discovered past methods for word embedding like word2vec, GloVe, PMI etc [Related paper](/doc/?uri=http%3A%2F%2Fwww.aclweb.org%2Fanthology%2FQ16-1028) 2018-06-10T15:07:37Z 2018-06-10 Sanjeev Arora on "A theoretical approach to semantic representations" - YouTube (2016) A Word Embedding Approach to Predicting the Compositionality of Multiword Expressions (2015) 2018-06-08T07:46:42Z 2018-06-08 > introduction to extremely simple ways of computing sentence embeddings, which on many standard tasks, beat many state-of-the-art deep learning methods. Related to [this paper](/doc/?uri=https%3A%2F%2Fopenreview.net%2Fforum%3Fid%3DSyK00v5xx) (BTW, contains a good intro to text embeddings) 2018-06-25 2018-06-25T21:00:24Z Deep-learning-free Text and Sentence Embedding, Part 1 – Off the convex path Evaluation of sentence embeddings in downstream and linguistic probing tasks 2018-06-27T11:48:33Z 2018-06-27 2018-06-19 2018-06-19T10:06:38Z Understanding the Working of Universal Language Model Fine Tuning (ULMFiT) – Let the Machines Learn Pl@ntNet identifiez une plante à partir d'une photo 2018-06-06 2018-06-06T22:17:42Z 2018-06-25 2018-06-25T21:04:28Z Deep-learning-free Text and Sentence Embedding, Part 2 – Off the convex path > Can we design a text embedding with the simplicity and transparency of SIF while also incorporating word order information? yes we can. 2018-06-04T17:58:18Z Matt Botvinick George Dahl Chris Dyer 1806.01261 Justin Gilmer Andrew Ballard Ashish Vaswani [1806.01261] Relational inductive biases, deep learning, and graph networks 2018-10-17T17:51:36Z Ryan Faulkner Victor Bapst Caglar Gulcehre Pushmeet Kohli Peter W. Battaglia Peter W. Battaglia Victoria Langston David Raposo Oriol Vinyals Mateusz Malinowski Yujia Li Artificial intelligence (AI) has undergone a renaissance recently, making major progress in key domains such as vision, language, control, and decision-making. This has been due, in part, to cheap data and cheap compute resources, which have fit the natural strengths of deep learning. However, many defining characteristics of human intelligence, which developed under much different pressures, remain out of reach for current approaches. In particular, generalizing beyond one's experiences--a hallmark of human intelligence from infancy--remains a formidable challenge for modern AI. The following is part position paper, part review, and part unification. We argue that combinatorial generalization must be a top priority for AI to achieve human-like abilities, and that structured representations and computations are key to realizing this objective. Just as biology uses nature and nurture cooperatively, we reject the false choice between "hand-engineering" and "end-to-end" learning, and instead advocate for an approach which benefits from their complementary strengths. We explore how using relational inductive biases within deep learning architectures can facilitate learning about entities, relations, and rules for composing them. We present a new building block for the AI toolkit with a strong relational inductive bias--the graph network--which generalizes and extends various approaches for neural networks that operate on graphs, and provides a straightforward interface for manipulating structured knowledge and producing structured behaviors. We discuss how graph networks can support relational reasoning and combinatorial generalization, laying the foundation for more sophisticated, interpretable, and flexible patterns of reasoning. As a companion to this paper, we have released an open-source software library for building graph networks, with demonstrations of how to use them in practice. Razvan Pascanu Kelsey Allen Alvaro Sanchez-Gonzalez Vinicius Zambaldi Nicolas Heess > generalizing beyond one's experiences--a hallmark of human intelligence from infancy--remains a formidable challenge for modern AI > A key signature of human intelligence is the ability to make infine use of finite means" (Humboldt, 1836; Chomsky, 1965) (ex: words / sentences > Here we explore how to improve modern AI's capacity for **combinatorial generalization** by biasing learning towards structured representations and computations, and in particular, systems that operate on graphs. (papier recommandé par [Peter Bloem](tag:peter_bloem)) Adam Santoro 2018-06-13 Andrea Tacchetti 2018-06-13T13:34:03Z Francis Song Relational inductive biases, deep learning, and graph networks Charles Nash Daan Wierstra Jessica B. Hamrick 2018-06-02T10:23:37Z 2018-06-02 How To Create Natural Language Semantic Search For Arbitrary Objects With Deep Learning 2018-06-28T01:21:31Z 2018-07-12T09:31:10Z [1806.04470] Design Challenges and Misconceptions in Neural Sequence Labeling Shuailong Liang 2018-06-12T12:43:42Z Jie Yang Yue Zhang Design Challenges and Misconceptions in Neural Sequence Labeling Jie Yang 2018-06-28 design challenges of constructing effective and efficient neural sequence labeling systems We investigate the design challenges of constructing effective and efficient neural sequence labeling systems, by reproducing twelve neural sequence labeling models, which include most of the state-of-the-art structures, and conduct a systematic model comparison on three benchmarks (i.e. NER, Chunking, and POS tagging). Misconceptions and inconsistent conclusions in existing literature are examined and clarified under statistical experiments. In the comparison and analysis process, we reach several practical conclusions which can be useful to practitioners. 1806.04470 Despite the fast developmental pace of new sentence embedding methods, it is still challenging to find comprehensive evaluations of these different techniques. In the past years, we saw significant improvements in the field of sentence embeddings and especially towards the development of universal sentence encoders that could provide inductive transfer to a wide variety of downstream tasks. In this work, we perform a comprehensive evaluation of recent methods using a wide variety of downstream and linguistic feature probing tasks. We show that a simple approach using bag-of-words with a recently introduced language model for deep context-dependent word embeddings proved to yield better results in many tasks when compared to sentence encoders trained on entailment datasets. We also show, however, that we are still far away from a universal encoder that can perform consistently across several downstream tasks. Christian S. Perone [1806.06259] Evaluation of sentence embeddings in downstream and linguistic probing tasks 2018-06-16T16:07:49Z 2018-06-19T10:15:34Z 1806.06259 2018-06-19 a simple approach using bag-of-words with a recently introduced language model for deep context-dependent word embeddings proved to yield better results in many tasks when compared to sentence encoders trained on entailment datasets > We also show, however, that we are still far away from a universal encoder that can perform consistently across several downstream tasks. Evaluation of sentence embeddings in downstream and linguistic probing tasks Roberto Silveira Christian S. Perone 2018-06-16T16:07:49Z Thomas S. Paula 2018-06-08 2018-06-08T00:20:41Z Chatbots were the next big thing: what happened? – The Startup – Medium Au Sahara, voyager devient un crime 2018-06-03T15:11:20Z 2018-06-03 2018-06-24 2018-06-24T20:06:31Z Banksy peint les murs de Paris pour illustrer la crise des migrants Un Univers sans matière noire? | CNRS Le journal 2018-06-08T14:02:56Z 2018-06-08 D2KLab/entity2rec: entity2rec generates item recommendation from knowledge graphs 2018-06-04T00:10:02Z 2018-06-04 2018-06-09T09:26:53Z 2018-06-09 Reinforcement Learning from scratch – Insight Data