Unlock the full potential of AI with Building LLMs for Production—our 470+ page guide to mastering LLMs with practical projects and expert insights!


NLP News Cypher | 05.17.20
Latest   Machine Learning   Newsletter

NLP News Cypher | 05.17.20

Last Updated on July 27, 2023 by Editorial Team

Author(s): Ricky Costa

Originally published on Towards AI.

Photo by Sherly Tay on Unsplash


NLP News Cypher U+007C 05.17.20


You may have heard, Elon Musk isn’t very happy. And it didn’t help when the Head of AI at Facebook, Jerome Pesenti, decided to throw gas to the fire by calling Musk out on Twitter over his AI knowledge:

U+1F608 The Beef Thread U+1F608

Jerome Pesenti@an_open_mind

I believe a lot of people in the AI community would be ok saying it publicly. @elonmusk has no idea what he is talking about when he talks about AI. There is no such thing as AGI and we are nowhere near matching human intelligence. #noAGI


Elon Musk@elonmusk

Facebook sucks


Yann LeCun@ylecun

Tesla engineers and scientists be like “can we still use PyTorch though?”


Elon Musk@elonmusk

Fair point, PyTorch is great!


Zuck be like:


Meanwhile, history was made when NVIDIA CEO Mr. Huang (and his leather jacket) held the first ever keynote speech in a kitchen:

FYI, this past week we released another set of notebooks for the Super Duper NLP Repo! Thank you to all contributors: Aditya Malte, Kapil Chauhan, Veysel Kocaman & Sayak Paul. U+1F60E

The Super Duper NLP Repo

Colab notebooks for various tasks in NLP


P.S. – don’t click on this U+1F9D0 :




This Week:


Nested JSON

Visualizing AI Model Training


Text 2 Speech on CPUs

T5 Inspires BART to Question

Colab of the Week: T5 Tuning U+1F525U+1F525

CMUs ML Video Collection

Dataset of the Week: Street View Text (SVT)


You may have already used the Tacotron model found in the Super Duper NLP Repo for text 2 speech experimentation. Well now NVIDIA has released FlowTron and it comes with its own controllable style modulation. In fact, if you hear the keynote narration in the Huang video above, FlowTron is the model being used. If interested, check out their blog page showing various style demos alongside Tacotron 2.


Flowtron: an Autoregressive Flow-based Generative Network for Text-to-Speech Synthesis

LJSpeech Ground Truth Flowtron Tacotron 2 audio not supported audio not supported audio not supported With Flowtron we…




In our recent paper we propose Flowtron: an autoregressive flow-based generative network for text-to-speech synthesis…


Nested JSON

Those JSON records can get convoluted really quick especially if two objects share the same key like “name” for people and “name” for company name. Below is a quick guide to get through nested JSON data with a function for isolating the right key, giving all of us a new hope. U+1F601

Parsing Nested JSON Records in Python

JSON is the typical format used by web services for message passing that's also relatively human-readable. Despite…


Visualizing AI Model Training

The title says it all. This is a step by step guide (w/Colab) for infusing Weights and Biases visualizations and Hugging Face’s Transformers library. For this example, DistilBERT on CoLA dataset is used to observe the Mathew’s correlation coefficient metric:

A Step by Step Guide to Tracking Hugging Face Model Performance

This tutorial explains how to train a model (specifically, an NLP classifier) using the Weights & Biases and…



Searching over large amount of documents can often lead to multi-hop problems. Oftentimes, a question may require to search multiple areas of a knowledge base to answer a query accurately. In this work, the authors at CMU attempt to comb through documents (like a graph) without converting documents into a graph (leaving documents in original state) — which is easier to build than a knowledge graph and offering a major speed boost.

How does it perform?

On the MetaQA task:

the model outperforms the previous best by 6% and 9% on 2-hop and 3-hop questions, respectively, while being up to 10x faster. U+1F525U+1F525U+1F525U+1F525

On the HotpotQA:

method trades off some accuracy (F1 score 43 vs 61) to deliver a 100x improvement in terms of speed.


Differentiable Reasoning over Text

We all rely on search engines to navigate the massive amount of online information published every day. Modern search…


16-core CPU Demo:



Text 2 Speech on CPUs

New text-2-speech model from Facebook can generate one second of audio in 500 milliseconds on CPU. In addition, they’ve included style embeddings allowing the AI voice to mimic an assistant, soft, fast, projected, and formal style!

There’s demo speech in the link below. Bad news though as this seems to not be open-sourced. U+1F60C

A highly efficient, real-time text-to-speech system deployed on CPUs

Facebook AI has built and deployed a real-time neural text-to-speech system on CPU servers, delivering industry-leading…


T5 Inspires BART to Question

Open-domain QA, made famous by DrQA, usually involves a 2 stage model approach where you search over an external knowledge base (e.g. Wikipedia) and then use another model to retrieve data for a query. For closed-domain QA, like the SQuAD task, the downstream task involves feeding a general pre-trained model text and a question, and the model is tasked to find the answer span in the text. However, in this repo using the BART-large model, Sewon Min uses a model pre-trained on the knowledge itself and then fine-tuned to answer questions! This style, called open-domain closed-book, was inspired and described in the T5 paper below. Straight fire U+1F525U+1F525.

BART GitHub:


This is a BART version of sequence-to-sequence model for open-domain QA in a closed-book setup, based on PyTorch and…


T5 GitHub:


This repository contains the code for reproducing the experiments in How Much Knowledge Can You Pack Into the…


Paper based off the T5:


Colab of the Week: T5 Tuning U+1F525U+1F525

Learn to use T5 for review classification, emotion classification and commonsense inference!

Google Colaboratory

Edit description


CMUs ML Video Collection

From Graham Neubig, this great collection offers 24 lecture videos for your machine learning edification. You know the collection is good when attention is discussed in the 7th videoU+1F601. In these video clips we get everything from search trees, document level models to machine reading and NLG:

Dataset of the Week: Street View Text (SVT)

What is it?

Dataset contains street scene images with annotations used for scene text recognition task.


Where is it?

The Street View Text Dataset

Datasets -> Datasets List -> Current Page Kai Wang EBU3B, Room 4148 Department of Comp. Sci. and Engr. University of…


Every Sunday we do a weekly round-up of NLP news and code drops from researchers around the world.

If you enjoyed this article, help us out and share with friends!

For complete coverage, follow our Twitter: @Quantum_Stat


Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI

Feedback ↓