We're accepting team registrations for the #NLP4Gov #hackathon between 2pm and 4pm in the foyer. #nlproc @info_vlaanderen Marijke Verhavert

We wish everyone a great 2019! We're starting the year with a new website. Take a look at nlp.town and let us know what you think! #NaturalLanguageProcessing #ArtificialIntelligence #TextMining #DeepLearning

Transfer Learning works great for Natural Language Processing, but sometimes its large models can be hard to handle. At NLP Town we used model distillation to train spaCy text classifiers that rival BERT for sentiment analysis nlp.town/blog/distillin… #deeplearning #NLProc

Slowly but surely #transferlearning models are being released for languages other than English. In a recent project we found that the German #BERT model from the MDZ Digital Library team outperforms multilingual BERT on a German #sentimentanalysis task. #nlproc #deeplearning #ai

We're still excited by the great turnout and interesting talks at yesterday's 12th Belgium NLP Meetup. We hope to see you all again for the next meetup in early 2020! #nlproc #ai #deeplearning #ArtificialIntelligence #textmining #metup

Evaluation and communication are central to the development of successful AI software. At NLP Town, we've recently started using Streamlit to experiment with our machine learning models and share the results with our clients. #nlproc #ArtificialIntelligence #machinelearning

Our #Python #NLProc Toolbox for 2019 has some shiny newcomers (Hugging Face's Transformers, Streamlit, Sebastián Ramírez's FastAPI), rising stars (Rasa and Zalando Technology's Flair) and of course the usual suspects (spaCy). Let us know what you think! #machinelearning #deeplearning

We're loving the inference widget in v2 of the Hugging Face model hub, which lets you try out all shared #nlproc models, including our 6-language #SentimentAnalysis model. Also great to see our model has been downloaded almost 64,000 times in the past 30 days! #deeplearning #ai

Hey Hugging Face, it looks like the model card for our finetuned sentiment model is down. It worked perfectly a few weeks ago, including the inference widget. Anything we can do to fix this? clem 🤗 Thomas Wolf

Data #annotation and quality is critical in #ArtificialIntelligence and #NLProc. That's why we developed #Taggle, a state-of-the-art labeling platform for all your text annotation projects. Take a look at nlp.town/taggle/ to find out more.

Today we were experimenting with Google Translate glossaries. We were expecting some nifty #deeplearning solution, but instead it just appears to replace words blindly, with agreement and other grammatical errors as a result. Can't Google AI do better? #neuralempty #nlproc

We thought that #deeplearning in #nlproc had reached peak hype with #BERT, but that was nothing compared to OpenAI's #GPT3. Just about every client thinks their problem is solved now. #AI

😍 We're simply in awe at spaCy's v3.0 release. So many new features will have an immediate impact on our projects: transformer pipelines, the new workflow system, etc. Kudos to Matthew Honnibal, Ines Montani and Sofie Van Landeghem! explosion.ai/blog/spacy-v3-… #nlproc #deeplearning #ai

Today we're taking spaCy v3 for a spin in a project on grammar correction. ✍️ Loving the config-based training and the integration with . 😍 Also, these #transformers look pretty promising. 🧐 Are they a thing? 😏 Matthew Honnibal Ines Montani Sofie Van Landeghem #NLProc

We've given our data #labeling tool a revamp, as well as a new name: Tagalog. Get in touch if you're struggling with an #annotation project and need help. #machinelearning #datascience #textmining #nlproc

Many #ai and #machinelearning projects fail because of low-quality labeled #data. For challenging #annotation tasks, it's crucial to have several annotators label the documents and review the conflicts. Our labeling tool #Tagalog is there to support you in those projects. #nlproc

#NLProc data #annotation is a team effort. The latest version of our labeling tool #Tagalog allows project managers to organize labeling campaigns and track the progress of annotators and reviewers in real time. tagalog.ai #ai #machinelearning #artificialintelligence

We evaluated spaCy's entity recognition on 14 days of English #financial news. The non-neural models all struggle with the identification of organizations, but look how the transformer pipeline shines! 🤯 #nlproc #deeplearning Labeling with tagalog.ai

What #nlproc models work best for the identification of organizations in English #financial news? Stanford NLP Group Stanza's BiLSTM-CRF comfortably beats spaCy's CNN-based models, but can't compete with spaCy's transformer-based pipeline. #deeplearning #AI #machinelearning

Can you do text classification without any labeled training examples? We put the zero-shot #classification pipelines in Hugging Face Transformers and Flair to the test, and wrote up our findings: nlp.town/blog/zero-shot… #nlp #nlproc #deeplearning #artificialintelligence #ai