Log in

No account? Create an account

Peter Rogers's Blog
Artist-in-Residence at Chez Firth

Monday (4/3/17) 12:27am - ... wherein Peter posts a Weekly Media Update.

Books:  Natural Language Processing with Python, The Science of Good Cooking
Movies:  <none>
TV:  <none>

Natural Language Processing with Python: Analyzing Text with the Natural Language Toolkit by Steven Bird, Ewan Klein, and Edward Loper
This is an O'Reilly programming book that covers nltk, Python's "natural language toolkit" for handling linguistics.

This book kind of falls between a few stools. It's not quite a book about linguistics, but it needs to cover some linguistics topics. And it's not quite a book about Python, but it needs to get readers up to speed with reading programs written in Python. And it's not quite a book about computational linguistics, but it's trying to cover one specific computational-linguistics toolkit in a way that newbies will understand, so it needs to provide context.

All that said, it's still a readable and interesting book. It alternates between Python lessons and linguistics information, and slowly covers all the various tools that nltk provides. And some of the information, I found very useful: for example, the toolkit includes an exhaustive English pronunciation dictionary, which I could use to determine how many English words don't have rhymes[1] If you want to do some very basic, simple things in computational linguistics, nltk can give you the tools you need, and this book can explain how to use it.

But it's clearly hobbled by having to cover so many topics and having to explain everything to a newbie target audience. You know that there's much more in-depth work being done with the nltk. And for that matter, you know that, since this book was published, machine learning and Big Data have provided fascinating alternate approaches to processing and generating natural language. So: you're not going to, say, rewrite google search via information in this book. You won't even be up for a very very dumb version of that.

And then there's the end of the book.

The book is attributed to multiple authors, and I'm assuming the closing chapters are where one of those gentlemen took over from the others. Because, as soon as they start covering first-order logic and lambda calculus, the book takes a hard right turn into baffling incomprehensibility. It's that rare perfect storm: an *introduction* to a topic that is incomprehensible to anybody who doesn't already *know* the topic. Code snippets with single-letter variable names fly by. The prose gets as terse and as turgid as a high-level math textbook. Lots of sentences begin with "Obviously," implying a misguided confidence from the authors in how well they're explaining the material. It's a mess.

And the funniest part is, this is such complex material that they can't even provide useful examples. The best they can do are little toy problems, and only vague hand-wave about how you might use these high-level nltk features in real life. It's the birdiest-eye view of these advanced topics. Even if I had understood these chapters, I would have learned nothing practical.

So: it has some nice introductory material about this specific toolkit -- and I genuinely did find a fun use for what I learned -- so if you have a weird linguistic question to explore, this will tell you what you need to know. But if you don't, I suspect this book will waste your time.

The Science of Good Cooking by Cooks, Illustrated
This is the Cooks, Illustrated cookbook based around 50 basic principles of how to cook.

It's nice to find the exact book you're looking for.

After watching a half-dozen or so seasons of Good Eats, I found myself wanting a very particular book. I thought it was great that Alton Brown covered lots of intimidating cooking concepts in terms of chemistry -- I could understand that. "Mix the sugar and butter in a blender so that the sugar granules punch tiny holes in the butter and aerates it" makes sense to me. "Mix the sugar and butter in a blender because you have to cream them" is frustrating and mysterious.

But Good Eats is, by design, scattershot. You cover one ingredient, or one dish, or one topic, per episode. Some concepts come up repeatedly, but subsequent mentions never go into more depth -- instead, they just refer back to the earlier episode, and rattle off a quick summary to get viewers up to speed. Wouldn't it be nice, I thought, to have a book that covered these basic cooking concepts, explained them in detail in terms of science, and worked their way up from simple, basic concepts to more complex ones.

The Science of Good Cooking is basically that book.

It has fifty chapters. Each chapter gives a basic principle of cooking -- the second chapter, for example, is "High Heat Develops Flavor". Then it explains, chemistry-wise, what it's getting at. In this case, it summarizes what we know about the Maillard Reaction, where surface-level proteins and starches react to high temperatures to create lots of new (and delicious) molecules. It lists various dos and do-nots to get that reaction working like you want it to. And then finally, it lists several recipes that put this knowledge to work: first you get the recipe, and then you get a "Why it works" section to show how what you just learned applies to this particular dish.

And so it goes for fifty chapters, through meat and vegetable preparations, through baking and roasting and grilling, through desserts and entrées and sides. It includes regular little digressions where the Cooks, Illustrated folks run an experiment to test some piece of kitchen folk wisdom. For example, some folks say you should add sugar to whipped cream at the start; some say you should add it at the end; so... let's try doing both a bunch of times and see what happens.

Honestly, my only complaint about the book is that I read it wrong. I was stupidly completist about it and insisted on reading every chapter, and every recipe in every chapter, in order. If you're going to read the whole book -- instead of just sampling individual chapters you think are neat -- at least skip the recipes. Then, later on, if you want to really explore one of those fifty topics, you can read, or even try cooking, some of the listed recipes.

Honestly, you can get decent recipes anywhere. It's the knowledge of why and how these recipes work that feels welcome and unique.

For next week: lots of media-writing to catch up on, including Crazy Ex-Girlfriend, Logan, Paper Girls, Phonogram, and Stuff Matters.  At the moment, I'm watching OJ: Made in America, catching up on podcasts, and finally reading On the Origin of Species. (Weird to get caught up in arguments about evolution and realize I'd never read that.) While exercising, I'm on season 11 of Good Eats and season 3 of Last Week Tonight.

[1] The answer: 29% of all words. Crazy, right?

Tags: ,
Mood: [mood icon] contemplative · Music: none
Previous Entry Share Next Entry