]> 2018-09-20 2018-09-20T13:21:47Z Machine Learning - Cheatsheet (Teaching - CS 229) Extracting word senses from embeddings. [About this paper](/doc/?uri=https%3A%2F%2Farxiv.org%2Fabs%2F1601.03764) Linear algebraic structure of word meanings – Off the convex path 2018-09-20T23:47:32Z 2018-09-20 Can Global Semantic Context Improve Neural Language Models? - Apple (2018) 2018-09-27T21:37:54Z 2018-09-27 Ubang: The Nigerian village where men and women speak different languages - BBC News 2018-09-03 2018-09-03T11:45:46Z Off the convex path 2018-09-09 2018-09-09T15:38:14Z 2018-09-11 2018-09-11T00:58:07Z [Learning Note] StarSpace For Multi-label Text Classification AI Can Recognize Images, But Text Has Been Tricky—Until Now | WIRED 2018-09-08 2018-09-08T00:19:53Z Named Entity Recognition and Classification with Scikit-Learn 2018-09-16T10:15:39Z 2018-09-16 The Grand Budapest Hotel 2018-09-16T22:39:24Z 2018-09-16 datas-frame – Tabular Data in Scikit-Learn and Dask-ML 2018-09-17 2018-09-17T18:06:59Z 2018-09-09 GitHub - marcotcr/lime: Lime: Explaining the predictions of any machine learning classifier 2018-09-09T15:25:49Z 2018-09-04T03:15:56Z Open Domain Question Answering Using Early Fusion of Knowledge Bases and Text 2018-09-06 Bhuwan Dhingra Ruslan Salakhutdinov QA over the combination of a KB and entity-linked text, which is appropriate when an incomplete KB is available with a large text corpus. > In practice, some questions are best answered using text, while others are best answered using KBs. A natural question, then, is how to effectively combine both types of information. Surprisingly little prior work has looked at this problem. Manzil Zaheer 1809.00782 Kathryn Mazaitis Open Domain Question Answering (QA) is evolving from complex pipelined systems to end-to-end deep neural networks. Specialized neural models have been developed for extracting answers from either text alone or Knowledge Bases (KBs) alone. In this paper we look at a more practical setting, namely QA over the combination of a KB and entity-linked text, which is appropriate when an incomplete KB is available with a large text corpus. Building on recent advances in graph representation learning we propose a novel model, GRAFT-Net, for extracting answers from a question-specific subgraph containing text and KB entities and relations. We construct a suite of benchmark tasks for this problem, varying the difficulty of questions, the amount of training data, and KB completeness. We show that GRAFT-Net is competitive with the state-of-the-art when tested using either KBs or text alone, and vastly outperforms existing methods in the combined setting. Source code is available at https://github.com/OceanskySun/GraftNet . 2018-09-06T01:38:28Z 2018-09-04T03:15:56Z Haitian Sun Haitian Sun [1809.00782] Open Domain Question Answering Using Early Fusion of Knowledge Bases and Text William W. Cohen Creating a Machine Learning Web API with Flask - Wintellect 2018-09-25T09:51:21Z 2018-09-25 Le Dickinsonia, le plus ancien animal sur Terre, était ovale et plat 2018-09-21T08:19:59Z 2018-09-21 2018-09-15 2018-09-15T16:24:21Z Découverte d’un rare cimetière d’urnes funéraires en Amazonie 2018-09-28T22:21:15Z > **What's wrong with our unsupervised training objectives ? They are in pixel space rather than in abstract space** > Many more entropy bits in acoustics details than linguistic content. Related to [this paper](/doc/?uri=https%3A%2F%2Farxiv.org%2Fabs%2F1709.08568) From Deep Learning of Disentangled Representations to Higher-level Cognition - YouTube 2018-09-28 2018-09-05T22:58:06Z 2018-09-05 The Brazil Museum Fire: What Was Lost - The New York Times Can Mark Zuckerberg Fix Facebook Before It Breaks Democracy? | The New Yorker 2018-09-17T13:00:18Z 2018-09-17 Dix ans après Lehman Brothers : en attendant la prochaine crise 2018-09-15T16:41:51Z 2018-09-15 Boliang Zhang Zhiying Jiang 2018-09-06T02:56:58Z Qingyun Wang Heng Ji 2018-09-07 Xiaoman Pan Describing a Knowledge Base 2018-09-30T04:36:18Z We aim to automatically generate natural language descriptions about an input structured knowledge base (KB). We build our generation framework based on a pointer network which can copy facts from the input KB, and add two attention mechanisms: (i) slot-aware attention to capture the association between a slot type and its corresponding slot value; and (ii) a new \emph{table position self-attention} to capture the inter-dependencies among related slots. For evaluation, besides standard metrics including BLEU, METEOR, and ROUGE, we propose a KB reconstruction based metric by extracting a KB from the generation output and comparing it with the input KB. We also create a new data set which includes 106,216 pairs of structured KBs and their corresponding natural language descriptions for two distinct entity types. Experiments show that our approach significantly outperforms state-of-the-art methods. The reconstructed KB achieves 68.8% - 72.6% F-score. [1809.01797] Describing a Knowledge Base 2018-09-07T12:57:23Z 1809.01797 Qingyun Wang Lifu Huang Kevin Knight 2018-09-21 2018-09-21T13:40:35Z Embeddings@Twitter 2018-09-18 2018-09-18T18:15:49Z A La Carte Embedding: Cheap but Effective Induction of Semantic Feature Vectors (2018) 2018-09-18T15:05:58Z 2018-09-18 Key topics extraction and contextual sentiment of users’ reviews Strategies to scale computationally: bigger data — scikit-learn documentation using out-of-core learning 2018-09-15 2018-09-15T18:42:54Z 2018-09-15T18:22:29Z Visualising Top Features in Linear SVM with Scikit Learn and Matplotlib 2018-09-15 > The weights obtained from svm.coef_ represent the vector coordinates which are orthogonal to the hyperplane and their direction indicates the predicted class. The absolute size of the coefficients in relation to each other can then be used to determine feature importance for the data separation task 2018-09-09 Depends on the definition - it's about machine learning, data science and more 2018-09-09T15:32:10Z Graph embedding Day - Lyon 2018-09-10T22:56:23Z 2018-09-10 2018-09-19 2018-09-19T16:59:33Z Towards Natural Language Semantic Code Search | GitHub Engineering 2018-09-18 A La Carte embeddings > Distributional methods for capturing meaning, such as word embeddings, often require observing many examples of words in context. But most humans can infer a reasonable meaning from very few or even a single occurrence... > > “Porgies live in shallow temperate marine waters” > > Inducing word embedding from their contexts: a surprising linear relationship Simple and efficient semantic embeddings for rare words, n-grams, and language features – Off the convex path 2018-09-18T18:07:01Z Distill — Latest articles about machine learning 2018-09-09T15:41:18Z 2018-09-09 2018-09-18 2018-09-18T20:30:27Z La plus vieille biodiversité de communauté bactérienne, datée de 2,1 milliards d’années et son implication dans la conservation du biota francevillien 2016-08-09T17:54:52Z 1602.04938 "Why Should I Trust You?": Explaining the Predictions of Any Classifier technique that explains the predictions of any classifier by learning an interpretable model locally around the prediction [1602.04938] "Why Should I Trust You?": Explaining the Predictions of Any Classifier Carlos Guestrin 2018-09-09 Sameer Singh Despite widespread adoption, machine learning models remain mostly black boxes. Understanding the reasons behind predictions is, however, quite important in assessing trust, which is fundamental if one plans to take action based on a prediction, or when choosing whether to deploy a new model. Such understanding also provides insights into the model, which can be used to transform an untrustworthy model or prediction into a trustworthy one. In this work, we propose LIME, a novel explanation technique that explains the predictions of any classifier in an interpretable and faithful manner, by learning an interpretable model locally around the prediction. We also propose a method to explain models by presenting representative individual predictions and their explanations in a non-redundant way, framing the task as a submodular optimization problem. We demonstrate the flexibility of these methods by explaining different models for text (e.g. random forests) and image classification (e.g. neural networks). We show the utility of explanations via novel experiments, both simulated and with human subjects, on various scenarios that require trust: deciding if one should trust a prediction, choosing between models, improving an untrustworthy classifier, and identifying why a classifier should not be trusted. Marco Tulio Ribeiro 2016-02-16T08:20:14Z 2018-09-09T15:22:41Z Marco Tulio Ribeiro 2018-09-17 java - How to debug stream().map(...) with lambda expressions? - Stack Overflow 2018-09-17T12:35:46Z 2018-09-27 2018-09-27T11:29:18Z Paris NLP Season 3 Meetup #1 | Meetup slides présentées au [Paris NLP meetup](/doc/?uri=https%3A%2F%2Fwww.meetup.com%2Ffr-FR%2FParis-NLP%2Fevents%2Fxzstdqyxmbjc%2F) Unsupervised Machine Translation. G. Lample (slides) 2018-09-29 2018-09-29T10:29:24Z