Errata

on the example 5-14 the __init__ method documentation mentions "filter_width (int): width of the convolutional kernels" but that is not a parameter to that method.

I wondered why the figure in Example 1-2 doesn't show 0's in the "flies" column, since the text emphasizes that words occurring in all documents should have an TF-IDF of 0. According to the sklearn documentation, its TF-IDF calculator uses nonstandard definitions of IDF.

On the Preface of the book the Supplemental material (code examples, exercises, etc.) link is unavailable. nlproc.info/PyTorchNLPBook/repo/ does not work anymore. Is there a different URL or could this one be updated?

In Chapter 1. Introduction under the section "TF Representation",

Figure 1-4. The collapsed one-hot representation generated by Example 1-1

contains only 7 columns. the field 'a' seems to be missing.. The length of the columns should be 8 i believe..

in "Example 4-8. Hyperparameters and program options for the MLP-based Yelp review classifier" a comma is missing after "hidden_dim=300"

- The output shows two y labels which are Sentence 1 and Sentence 2, however the code only includes Sentence 2.

- Also, vocab is not defined in the code which returns an error.

The code should be:

from sklearn.feature_extraction.text import CountVectorizer
import seaborn as sns
corpus = ['Time flies like an arrow.',
'Fruit flies like a banana.']
one_hot_vectorizer = CountVectorizer(binary=True)
one_hot = one_hot_vectorizer.fit_transform(corpus).toarray()
vocab = one_hot_vectorizer. get_feature_names()
sns.heatmap(one_hot, annot=True,
cbar=False, xticklabels=vocab,
yticklabels=['Sentence 1', 'Sentence 2'])

"By definition, the weight matrix of a Linear layer that accepts as input this one-­hot vector must have the same number of rows as the size of the one-­hot vector..."
one-­hot vector must have the same number of columns according to Fig.5.1