Python实战：打造自然语言处理应用

需积分: 32 42 浏览量更新于2024-07-17 1 收藏 9.9MB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

"《自然语言处理实战》是一本由Manning出版社发行的Apress系列书籍，它专注于使用Python这一强大的编程语言来实际操作和解决自然语言处理(NLP)中的各种问题。该书特别强调实践和应用，提供了丰富的编码练习，涵盖了文本分类、语音识别（部分）、主题建模、文本摘要、文本生成、实体提取和情感分析等多个关键领域。书中首先介绍了文本数据清洗和预处理的重要性，通过实例展示如何对文本进行规范化、高级预处理，包括词干提取、词性标注(pos tagging)等步骤，这些都是NLP基础中的基石。接着，作者深入探讨了语义和句法分析的实际应用，展示了如何利用诸如正则表达式、命名实体识别(NER)等技术提取信息，并将文本转化为结构化数据，便于后续分析。在机器学习与深度学习的章节中，读者将接触到循环神经网络(RNN)、卷积神经网络(CNN)、受限玻尔兹曼机(RBM)和生成对抗网络(GAN)等先进模型，它们在文本理解和生成、情感分析以及对话系统中的应用被详细讲解。例如，书中提到的聊天机器人中采用了循环管道（recirculating pipeline）设计，结合数据库管理和用户行为分析，确保对话流程的顺畅和个性化。整个过程分为四个主要阶段：解析、分析、执行和通用化/分类。解析阶段涉及词汇分割、词法分析和结构化数据转换；分析阶段则检查语法、情感、人性化和风格，使用深度学习模型如CNN和RNN进行这些任务；执行阶段执行决策，如根据分析结果更新对话计划并选择响应；最后，通过对大量数据的训练，模型会逐渐学习和泛化，从而提高整体性能。通过阅读本书，读者不仅能掌握实际的NLP技巧，还能建立一个可用于自身项目的解决方案工具箱，显著提升开发效率。此外，该书还配有翔实的案例研究和实用的代码示例，确保读者能够迅速上手并在实践中熟练掌握自然语言处理技术。无论是对于初学者还是经验丰富的开发者来说，这都是一本极具价值的参考资料。"

资源详情

资源推荐

FOREWORDxiv

news preferences are monitored so that our interests, opinions, and beliefs are

indulged. What is at the heart of all these powerful technologies?

The answer is natural language processing. In this book you will learn both the the-

ory and practical skills needed to go beyond merely understanding the inner workings

of these systems, and start creating your own algorithms or models. Fundamental

computer science concepts are seamlessly translated into a solid foundation for the

approaches and practices that follow. Taking the reader on a clear and well-narrated

tour through the core methodologies of natural language processing, the authors

begin with tried and true methods, such as TF-IDF, before taking a shallow but deep

(yes, I made a pun) dive into deep neural networks for NLP.

Language is the foundation upon which we build our shared sense of humanity.

We communicate not just facts, but emotions; through language we acquire knowl-

edge outside of our realm of experience, and build understanding through sharing

those experiences. You have the opportunity to develop a solid understanding, not

just of the mechanics of NLP, but the opportunities to generate impactful systems that

may one day understand humankind through our language. The technology of NLP

has great potential for misuse, but also great potential for good. Through sharing

their knowledge, via this book, the authors hope to tip us towards a brighter future.

R. ARWEN GRIFFIOEN

SENIOR DATA SCIENTIST - RESEARCH

ZENDESK

PREFACExvi

gained confidence, and learned more and more from my mentors and mentees, it

seemed like I might be able to build something new and magical myself.

One of the tricks I learned was to iterate through a collection of documents and

count how often words like “War” and “Hunger” are followed by words like “Games”

or “III.” If you do that for a large collection of texts, you can get pretty good at guess-

ing the right word in a “chain” of words, a phrase, or sentence. This classical approach

to language processing was intuitive to me.

Professors and bosses called this a Markov chain, but to me it was just a table of

probabilities. It was just a list of the counts of each word, based on the preceding

word. Professors would call this a conditional distribution, probabilities of words con-

ditioned on the preceding word. The spelling corrector that Peter Norvig built for

Google showed how this approach scales well and takes very little Python code.

All

you need is a lot of natural language text. I couldn’t help but get excited as I thought

about the possibilities for doing such a thing on massive free collections of text like

Wikipedia or the Gutenberg Project.

Then I heard about latent semantic analysis (LSA). It seemed to be just a fancy way

of describing some linear algebra operations I’d learned in college. If you keep track

of all the words that occur together, you can use linear algebra to group those words

into “topics.” LSA could compress the meaning of an entire sentence or even a long

document into a single vector. And, when used in a search engine, LSA seemed to

have an uncanny ability to return documents that were exactly what I was looking for.

Good search engines would do this even when I couldn’t think of the words that

might be in those documents!

Then

gensim

released a Python implementation of Word2vec word vectors, mak-

ing it possible to do semantic math with individual words. And it turned out that this

fancy neural network math was equivalent to the old LSA technique if you just split up

the documents into smaller chunks. This was an eye-opener. It gave me hope that I

might be able to contribute to the field. I’d been thinking about hierarchical semantic

vectors for years—how books are made of chapters of paragraphs of sentences of

phrases of words of characters. Tomas Mikolov, the Word2vec inventor, had the insight

that the dominant semantics of text could be found in the connection between two

layers of the hierarchy, between words and 10-word phrases. For decades, NLP

researchers had been thinking of words as having components, like niceness and emo-

tional intensity. And these sentiment scores, components, could be added and sub-

tracted to combine the meanings of multiple words. But Mikolov had figured out how

to create these vectors without hand-crafting them, or even defining what the compo-

nents should be. This made NLP fun!

See the web page titled “How to Write a Spelling Corrector” by Peter Norvig (http://www.norvig.com/spell-

correct.html).

If you appreciate the importance of having freely accessible books of natural language, you may want to keep

abreast of the international effort to extend copyrights far beyond their original

“use by” date: gutenberg.org

(http://www.gutenberg.org) and gutenbergnews.org (http://www.gutenbergnews.org/20150208/copyright-

term-extensions-are-looming:).

PREFACE xvii

About that time, Thunder introduced me to his mentee, Cole. And later others

introduced me to Hannes. So the three of us began to “divide and conquer” the field

of NLP. I was intrigued by the possibility of building an intelligent-sounding chatbot.

Cole and Hannes were inspired by the powerful black boxes of neural nets. Before

long they were opening up the black box, looking inside and describing what they

found to me. Cole even used it to build chatbots, to help me out in my NLP journey.

Each time we dug into some amazing new NLP approach it seemed like something

I could understand and use. And there seemed to be a Python implementation for

each new technique almost as soon as it came out. The data and pretrained models we

needed were often included with these Python packages. “There’s a package for that”

became a common refrain on Sunday afternoons at Floyd’s Coffee Shop where

Hannes, Cole, and I would brainstorm with friends or play Go and the “middle button

game.” So we made rapid progress and started giving talks and lectures to Hack Ore-

gon classes and teams.

In 2015 and 2016 things got more serious. As Microsoft’s Tay and other bots began

to run amok, it became clear that natural language bots were influencing society. In

2016 I was busy testing a bot that vacuumed up tweets in an attempt to forecast elec-

tions. At the same time, news stories were beginning to surface about the effect of

Twitter bots on the US presidential election. In 2015 I had learned of a system used to

predict economic trends and trigger large financial transactions based only on the

“judgment” of algorithms about natural language text.

These economy-influencing

and society-shifting algorithms had created an amplifier feedback loop. “Survival of

the fittest” for these algorithms appeared to favor the algorithms that generated the

most profits. And those profits often came at the expense of the structural founda-

tions of democracy. Machines were influencing humans, and we humans were train-

ing them to use natural language to increase their influence. Obviously these

machines were under the control of thinking and introspective humans, but when you

realize that those humans are being influenced by the bots, the mind begins to bog-

gle. Could those bots result in a runaway chain reaction of escalating feedback? Per-

haps the initial conditions of those bots could have a big effect on whether that chain

reaction was favorable or unfavorable to human values and concerns.

Then Brian Sawyer at Manning Publishing came calling. I knew immediately what

I wanted to write about and who I wanted to help me. The pace of development in

NLP algorithms and aggregation of natural language data continued to accelerate as

Cole, Hannes, and I raced to keep up.

The firehose of unstructured natural language data about politics and economics

helped NLP become a critical tool in any campaign or finance manager’s toolbox. It’s

unnerving to realize that some of the articles whose sentiment is driving those predic-

tions are being written by other bots. These bots are often unaware of each other. The

bots are literally talking to each other and attempting to manipulate each other, while

See the web page titled “Why Banjo Is the Most Important Social Media Company You’ve Never Heard Of”

(https://www.inc.com/magazine/201504/will-bourne/banjo-the-gods-eye-view.html).

PREFACExviii

the health of humans and society as a whole seems to be an afterthought. We’re just

along for the ride.

One example of this cycle of bots talking to bots is illustrated by the rise of fintech

startup Banjo in 2015.

By monitoring Twitter, Banjo’s NLP could predict newsworthy

events 30 minutes to an hour before the first Reuters or CNN reporter filed a story.

Many of the tweets it was using to detect those events would have almost certainly been

favorited and retweeted by several other bots with the intent of catching the “eye” of

Banjo’s NLP bot. And the tweets being favorited by bots and monitored by Banjo

weren’t just curated, promoted, or metered out according to machine learning algo-

rithms driven by analytics. Many of these tweets were written entirely by NLP engines.

More and more entertainment, advertisement, and financial reporting content

generation can happen without requiring a human to lift a finger. NLP bots compose

entire movie scripts.

Video games and virtual worlds contain bots that converse with

us, sometimes talking about bots and AI themselves. This “play within a play” will get

ever more “meta” as movies about video games and then bots in the real world write

reviews to help us decide which movies to watch. Authorship attribution will become

harder and harder as natural language processing can dissect natural language style

and generate text in that style.

NLP influences society in other less straightforward ways. NLP enables efficient

information retrieval (search), and being a good filter or promoter of some pages

affects the information we consume. Search was the first commercially successful

application of NLP. Search powered faster and faster development of NLP algorithms,

which then improved search technology itself. We help you contribute to this virtuous

cycle of increasing collective brain power by showing you some of the natural lan-

guage indexing and prediction techniques behind web search. We show you how to

index this book so that you can free your brain to do higher-level thinking, allowing

machines to take care of memorizing the terminology, facts, and Python snippets

here. Perhaps then you can influence your own culture for yourself and your friends

with your own natural language search tools.

The development of NLP systems has built to a crescendo of information flow and

computation through and among human brains. We can now type only a few charac-

ters into a search bar, and often retrieve the exact piece of information we need to

complete whatever task we’re working on, like writing the software for a textbook on

NLP. The top few autocomplete options are often so uncannily appropriate that we

feel like we have a human assisting us with our search. Of course we authors used vari-

ous search engines throughout the writing of this textbook. In some cases these

Banjo, https://www.inc.com/magazine/201504/will-bourne/banjo-the-gods-eye-view.html

The 2014 financial report by Twitter revealed that >8% of tweets were composed by bots, and in 2015 DARPA

held a competition (https://arxiv.org/ftp/arxiv/papers/1601/1601.05140.pdf) to try to detect them and

reduce their influence on society in the US.

Five Thirty Eight, http://fivethirtyeight.com/features/some-like-it-bot/

NLP has been used successfully to help quantify the style of 16th century authors like Shakespeare

(https://pdfs.semanticscholar.org/3973/ff27eb173412ce532c8684b950f4cd9b0dc8.pdf).

剩余544页未读，继续阅读

iebukes

粉丝: 9
资源: 132

Python实战：打造自然语言处理应用

【Manning经典教材】Natural Language Processing in Action

Natural Language Processing in Action (ePUB)

Handbook of Natural Language Processing Second Edition Chapman & Hall

natural language processing in action 下载

请给我关于这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing的各小结标题

这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing包含哪些具体小节

请给我这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing中Removing Duplicates的原始内容

请给我关于这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing中3.4的原始内容

基于机器学习的情感分析研究文献

请给我关于这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing的标题有哪些

handbook of natural language processing pdf

亲给我这篇文献的信息Cleaning GeoNames Data: A Case Study for Natural Language Processing

transformers for natural language processing pdf

请帮我提取关于这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing的The Case Study部分的内容

natural language processing and computational linguistics

请给我这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing中3.1的原始内容

请帮我提取关于这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing的Description of the case study部分的原始内容

请给我这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing中3.3Normalizing Data的原始内容

这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing有哪些小节

请给我这篇文献Cleaning GeoNames Data: A Case Study for Natural Language Processing中的各级标题信息

最新资源