NLTK Essentials: Build cool NLP and machine learning applications using NLTK and other Python libraries

Grammatical and code errors abound. Someone new to NLTK would be better served with the outdated O'Reilly Whale book.

This is a review for the PDF version of “NLTK Essentials” from packtpub.com. Other format (e.g. epub, kindle) may have minor difference in content layout when a page is referred.“NLTK Essentials” is a very concise (169 pages), incomplete overview of the Python NLTK module and other related technology. About half the content is not directly related to NLTK but to natural language processing (NLP) and data science in general. This book does not provide as many code snippets as other NLTK books (e.g. Python 3 Text Processing with NLTK 3 Cookbook), and many of the snippets still need debugging or require more instructions to run. The writing style is conversational and informal, and could use an editor for better clarity and organization.Chapter 1: Introduction to NLPIn the preface, the author intended the book “ideal for expert Python programmers who want to learn NLTK quickly”. There was really no need to review the basic data structures in python.Chapter 2:The table on pg20 was badly designed, mixing python syntax, module names, links to pypi and stackoverflow pages as table content.When first introducing NLTK to the audience, there was a need to introduce nltk.org/data.html and nltk.download() to run the example code. Otherwise, a “LookUpError” would be raised. Such error was never mentioned until near the end of the book (pg 176) for “improperly installing NLTK”. This was unhelpful, for this book only required a very small subset of the 500+Mb NLTK data.Chapter 3:This chapter introduced the Standford Parser (295Mb) and the Standford Tagger (290Mb). Only the download urls were provided, but more instructions were needed in order for the snippet to run -- it required JRE1.8 and some tweaking.Named Entity Recognition (NER) was first introduced here (pg40). Why? NER required chunking, which was not mentioned until Chapter 4 (pg56).Chapter 4:The code snippet on pg56 had an indentation error. An example on real data would be appreciated where readers could see the result of NER. “f=open(# absolute path for the file of text for which we want NER)” was not a real example.The code snippet on pg57 on Relation Extraction was an exact copy of the example at http://www.nltk.org/book/ch07.html (Chapter 7: 6. Relation Extraction) on a parsed news doc, NYT_19980315. This example was not instructive without any explanation as nltk.sem.extract_rels() was not a trivial function.Chapter 5: NLP ApplicationsThis was a puzzling chapter. There were 2 code examples of a “content summarizer” in the beginning, and the rest of NLP application examples had neither code example nor direct relevance to NLTK. The author could merge the bulk of this Chapter to Chapter 1.The 2 code examples had bugs: undefined variable on the first (pg61) and syntax error on the second (pg62). Neither example showed us the end result of the auto text summary.Chapter 6:This chapter was about the scikit-learn (sklearn) module, not the nltk module. Although both modules were capable of text classification, only the sklearn module was demonstrated. (Thus the book title was puzzling)Again the code examples needed debugging on pg77 (sms), pg78 (sms_list), pg81 (y_pred), pg91 (indentation) . Why skip the code on Logistic Regression (pg 85) and show us only the formula? Gensim was a great module on topic modeling in text, but trying to find a topic on a collections of unrelated SMS messages (as shown in the example) was puzzling.Chapter 7:The web crawler scrapy was interesting although the examples shown were too simple, and its installation needed some (undocumented) work. The code on pg108 needed debugging on AttributeError: 'HtmlResponse' object has no attribute 'URLs'Chapter 8 was unnecessary. The chapter surveyed the very basics of numpy, scipy, pandas and matplotlib modules. No direct application on NLTK (or NLP) was demonstrated.Chapter 9: It would great if the author could show us the D3.js code for the Geo visualization on pg144.Chapter 10:This chapter attempted to re-demonstrate previous NLTK examples in the context of Big Data, all in 12 pages. The author expected the readers to set up a Hadoop cluster, learn all hdfs commands, run MapReduce jobs in various ways, and familiarize themselves with the complex Hadoop ecosystem (Hive/Pig(Latin)/Mahout/Hbase... /Spark/...etc) before proceeding to the NLTK examples.All in all, I rated this book favorably (3/5 > 50%) because I have learned something new.

This book is so full of grammatical and logical errors that you never get to appreciate the content. The Punctuation seems to be random. Unreadable.

As someone passionate about the natural language and machine learning, I enjoyed this book from the first chapter. Despite being an essential book, it contains everything you need to give a first plunge into the ocean of natural language, from as simple as tokenization and parts of speech, to more complex tasks such as text classification and sentiment analysis in texts drawn from social networks. The code samples are concise and really easy to implement, so you don't have to expend a lot of time and effort testing and playing with them.Even if you are experienced in such topics, it's a fabulous compendia of web resources for code libraries and data samples that may help you to tune up your applications.As if this were not enough, it has three chapters that are short courses in their respective topic: web crawling and mining, machine learning and data analysis, and Big Data, all applied to natural language.From now on, it's going to be a reference book for my future natural language projects. I'll be waiting eagerly for the next book of the author in this series.

The Book is great for beginners who wants to learn natural language processing and text analytics. Concepts are explained very clearly and Python is being used to implement NLP models which made the book very easy to understand and will get you up and running with your NLP app development in no time.Python is not a prerequisite for this book as the author has described basics of Python programming which will help you to understand concepts and codes in book without any difficulties.

This is an excellent book to learn more about NLTK and a good start for NLP.