NLP   31704

« earlier    

roamresearch/modern-tensorflow.ipynb at master · roamanalytics/roamresearch
Contribute to roamanalytics/roamresearch development by creating an account on GitHub.
nlp  jupyter  tensorflow  best-practice  deep-learning 
2 hours ago by nharbour
302 Found
How To Create a ChatBot With tf-seq2seq For Free! - Added June 07, 2018 at 01:15PM
chatbot  machine-learning  nlp  read2of  tutorial 
yesterday by xenocid
302 Found
Building a Question-Answering System — Part 1 - Added June 07, 2018 at 01:14PM
data-science  nlp  read2of 
yesterday by xenocid
NLP's ImageNet moment has arrived
Word2vec and related methods are shallow approaches that trade expressivity for efficiency. Using word embeddings is like initializing a computer vision model with pretrained representations that only encode edges: they will be helpful for many tasks, but they fail to capture higher-level information that might be even more useful. A model initialized with word embeddings needs to learn from scratch not only to disambiguate words, but also to derive meaning from a sequence of words. This is the core aspect of language understanding, and it requires modeling complex language phenomena such as compositionality, polysemy, anaphora, long-term dependencies, agreement, negation, and many more. It should thus come as no surprise that NLP models initialized with these shallow representations still require a huge number of examples to achieve good performance.

In NLP, models are typically a lot shallower than their CV counterparts. Analysis of features has thus mostly focused on the first embedding layer, and little work has investigated the properties of higher layers for transfer learning. Let us consider the datasets that are large enough, fulfilling desideratum #1. Given the current state of NLP, there are several contenders.

Language modeling (LM) aims to predict the next word given its previous word. Existing benchmark datasets consist of up to 1B words, but as the task is unsupervised, any number of words can be used for training. See below for examples from the popular WikiText-2 dataset consisting of Wikipedia articles.

In light of this step change, it is very likely that in a year’s time NLP practitioners will download pretrained language models rather than pretrained word embeddings for use in their own models, similarly to how pre-trained ImageNet models are the starting point for most CV projects nowadays.
nlp  deeplearning 
yesterday by mike

« earlier    

related tags

ai  algorithms  analysis  annotation  api  apis  artificial-intelligence  arxiv  bbc  benchmark  benchmarks  bert  best-practice  bigdata  book  brazil  chatbot  china  chinese  chrome  classes  clojure  clustering  compling  comprehension  constraint-satisfaction  conversation  cv  data-science  data  datascience  dataset  datasets  dates  decanlp  deep-learning  deep_learning  deeplearning  directory  distributed-computing  docker  document-ranking  driverlessai  duckling  email  embeddings  encoder-decoder  entityextraction  ethics  evaluation  extension  facebook  funny  gan  generative-adversarial-networks  generator  gensim  glue  haskell  history  insight  interesting  interviews  ipynb  job  jupyter  language  languages  later  library  linguistics  machine-learning  machine  ml  natural-language  neural-machine-translation  neural-networks  neuralnets  nlp  nmt  opennlp  optimization  paper  papers  parsing  pdf  people  programming  psychology  python-cython  python  pytorch  qna  question-answer  r  read2of  readability  reference  research  resource  resources  salesforce  scikit-learn  scraping  search  sentiment  seq2seq  software  spacy  stemming  study-group  summarization  teaching  tensorflow  text-classifier  text  textual_analysis  tf-hub  tips  todo  tool  tools  transferlearning  translation  treebank  tutorial  tutorials  unicamp  universal-dependencies  video  webservices  webtools  wiki  wikipedia  word-embedding  word2vec  wordembedding  writing 

Copy this bookmark: