Deep Learning Notations

Deep Learning Notations#

We largely follow the Machine Learning: The Basics book in terms of notations.

NLP Terminologies#

Consider a toy corpus of five text units:

1. This is Foo, how can I help you
2. It might rain today
3. I love football
4. Crazy, Stupid & Love
5. I shot the sheriff

The terminology below follows common usage (see this discussion):

  • A document is the basic unit of text. What counts as a document is task-dependent — it can be an entire article, a passage, or a single sentence. In the toy corpus above, each line is one document, so “This is Foo, how can I help you” is a document and “It might rain today” is another.

  • A corpus (or collection) is a set of documents. The toy corpus above is composed of five documents.

  • The vocabulary is the set of all distinct words contained in the corpus. Here it is [&, can, crazy, foo, football, help, how, i, is, it, love, might, rain, sheriff, shot, stupid, the, this, today, you].

Table 4 Basic NLP Terminologies#

Notation

Description

\(V\)

The number of distinct vocabulary words in the corpus.

Corpus/Collection

A collection of documents.

These terms are used throughout the vector semantics and embeddings chapter.