NATURAL LANGUAGE PROCESSING (NLP) WEEKLY NEWSLETTER
The NLP Cypher | 04.17.22
Dark Mode
An updated chart w/r/t the last newsletter. This is the DeepSparse Engine and how it performs against ONNX runtime (deets above). With each new DeepSparse version release; inference performance just keeps on improving while you drink your Mai Tai. 🍾
Sad Tales from the Dark Web
Dread our favorite dark site continues to deal with active DDoS attacks.
A PGP message…
OH … ever had anyone post a file you didn’t trust? Dangerzone comes in handy, but as usual, I wouldn’t know anything about this. 😉
Business is hard bruh
Elon got bored and decided to buy Twitter this past week. He claims to want to make Twitter’s algo open-source. The first flex:
Elon applied further pressure, by initially joining and then deciding not to join Twitter’s board, then offering 43Billi to get the whole pie … Twitter CEO explains 🤣👇
We can expect further trolls from Elon in the near future.
… the source code: Elon’s Twitter Buyout Doc
… Substack throwing up gang signs already 😭:
Dall-E 2 vs. Latent-Diffusion | Thoughts?
Colab of the Week 🏆Latent-Diffusion Notebook
Megatron-DeepSpeed Hacks from BigScience
What’s Megatron-DeepSpeed?
Quick start script for Megatron-DeepSpeed:
GraphGallery
A gallery for benchmarking Graph Neural Networks (GNNs) based on pure PyTorch backend.
HTTPie: Human-Friendly CLI HTTP Client
UFOs and OSINT | A Repo (Yes, this is real)
UFO research is so lit, it’s now expanding into GitHub. 😭
ok, ok, last thing on UFOs (promise)
❤️Thanks for the big up Francesco!
KeyBART & KBIR
Keyphrase Boundary Infilling with Replacement (KBIR) achieves SOTA performance for the task of keyphrase extraction.
KBART model achieves SOTA performance on the task of keyphrase generation.
Paper: https://arxiv.org/abs/2112.08547
YouTube: https://youtu.be/W31PKHl5gvo
NLP Models | A Timeline
Knowledge Graphs ICLR 2021
ICLR 2021 papers summary.
Knowledge Distillation with Haystack
Hey Deepset, if you’re reading this, let’s parlay homies!😍⭐
Question Answering from UKP et al
Question Answering (QA) platform to enable users to easily implement, manage and share their custom QA pipelines.
It’s got a UI too.
Papers to Read 📚
NLP Index👨💻
A collection of recently released repos that caught our 👁
End-to-End Speech Translation for Code Switched Speech
Code switching in the context of English/Spanish conversations for the task of speech translation (ST), generating and evaluating both transcript and translation.
FewIE: Few-Shot Named Entity Recognition
An encoder evaluation framework for comparing the performance of SOTA pre-trained representations on the task of low-resource NER.
InCoder: A Generative Model for Code Infilling and Synthesis
A unified generative model that can perform program synthesis (via left-to-right generation) as well as editing (via infilling). The model is the first generative model that is able to directly perform zero-shot code infilling.
BioBART: Pretraining and Evaluation of A Biomedical Generative Language Model
A generative language model BioBART that adapts BART to the biomedical domain.
