Python 解决文本是一项十分常见的性能，本文整顿了多种文本提取及NLP相干的案例，还是十分用心的
文章很长，高下要忍一下，如果忍不了，那就珍藏吧，总会用到的
萝卜哥也贴心的做成了PDF，在文末获取！
[TOC]
提取 PDF 内容
# pip install PyPDF2 装置 PyPDF2import PyPDF2from PyPDF2 import PdfFileReader # Creating a pdf file object.pdf = open("test.pdf", "rb") # Creating pdf reader object.pdf_reader = PyPDF2.PdfFileReader(pdf) # Checking total number of pages in a pdf file.print("Total number of Pages:", pdf_reader.numPages) # Creating a page object.page = pdf_reader.getPage(200) # Extract data from a specific page number.print(page.extractText()) # Closing the object.pdf.close()提取 Word 内容
# pip install python-docx 装置 python-docximport docx def main(): try: doc = docx.Document('test.docx') # Creating word reader object. data = "" fullText = [] for para in doc.paragraphs: fullText.append(para.text) data = '\n'.join(fullText) print(data) except IOError: print('There was an error opening the file!') return if __name__ == '__main__': main()提取 Web 网页内容
# pip install bs4 装置 bs4from urllib.request import Request, urlopenfrom bs4 import BeautifulSoup req = Request('http://www.cmegroup.com/trading/products/#sortField=oi&sortAsc=false&venues=3&page=1&cleared=1&group=1', headers={'User-Agent': 'Mozilla/5.0'}) webpage = urlopen(req).read() # Parsingsoup = BeautifulSoup(webpage, 'html.parser') # Formating the parsed html filestrhtm = soup.prettify() # Print first 500 linesprint(strhtm[:500]) # Extract meta tag valueprint(soup.title.string)print(soup.find('meta', attrs={'property':'og:description'})) # Extract anchor tag valuefor x in soup.find_all('a'): print(x.string) # Extract Paragraph tag value for x in soup.find_all('p'): print(x.text)读取 Json 数据
import requestsimport jsonr = requests.get("https://support.oneskyapp.com/hc/en-us/article_attachments/202761727/example_2.json")res = r.json()# Extract specific node content.print(res['quiz']['sport'])# Dump data as stringdata = json.dumps(res)print(data)读取 CSV 数据
import csvwith open('test.csv','r') as csv_file: reader =csv.reader(csv_file) next(reader) # Skip first row for row in reader: print(row)删除字符串中的标点符号
import reimport string data = "Stuning even for the non-gamer: This sound track was beautiful!\It paints the senery in your mind so well I would recomend\it even to people who hate vid. game music! I have played the game Chrono \Cross but out of all of the games I have ever played it has the best music! \It backs away from crude keyboarding and takes a fresher step with grate\guitars and soulful orchestras.\It would impress anyone who cares to listen!" # Methood 1 : Regex# Remove the special charaters from the read string.no_specials_string = re.sub('[!#?,.:";]', '', data)print(no_specials_string) # Methood 2 : translate()# Rake translator objecttranslator = str.maketrans('', '', string.punctuation)data = data.translate(translator)print(data)应用 NLTK 删除停用词
from nltk.corpus import stopwords data = ['Stuning even for the non-gamer: This sound track was beautiful!\It paints the senery in your mind so well I would recomend\it even to people who hate vid. game music! I have played the game Chrono \Cross but out of all of the games I have ever played it has the best music! \It backs away from crude keyboarding and takes a fresher step with grate\guitars and soulful orchestras.\It would impress anyone who cares to listen!'] # Remove stop wordsstopwords = set(stopwords.words('english')) output = []for sentence in data: temp_list = [] for word in sentence.split(): if word.lower() not in stopwords: temp_list.append(word) output.append(' '.join(temp_list)) print(output)应用 TextBlob 更正拼写
from textblob import TextBlobdata = "Natural language is a cantral part of our day to day life, and it's so antresting to work on any problem related to langages."output = TextBlob(data).correct()print(output)应用 NLTK 和 TextBlob 的词标记化
import nltkfrom textblob import TextBlobdata = "Natural language is a central part of our day to day life, and it's so interesting to work on any problem related to languages."nltk_output = nltk.word_tokenize(data)textblob_output = TextBlob(data).wordsprint(nltk_output)print(textblob_output)Output:
['Natural', 'language', 'is', 'a', 'central', 'part', 'of', 'our', 'day', 'to', 'day', 'life', ',', 'and', 'it', "'s", 'so', 'interesting', 'to', 'work', 'on', 'any', 'problem', 'related', 'to', 'languages', '.']['Natural', 'language', 'is', 'a', 'central', 'part', 'of', 'our', 'day', 'to', 'day', 'life', 'and', 'it', "'s", 'so', 'interesting', 'to', 'work', 'on', 'any', 'problem', 'related', 'to', 'languages']应用 NLTK 提取句子单词或短语的词干列表
from nltk.stem import PorterStemmer st = PorterStemmer()text = ['Where did he learn to dance like that?', 'His eyes were dancing with humor.', 'She shook her head and danced away', 'Alex was an excellent dancer.'] output = []for sentence in text: output.append(" ".join([st.stem(i) for i in sentence.split()])) for item in output: print(item) print("-" * 50)print(st.stem('jumping'), st.stem('jumps'), st.stem('jumped'))Output:
where did he learn to danc like that?hi eye were danc with humor.she shook her head and danc awayalex wa an excel dancer.--------------------------------------------------jump jump jump应用 NLTK 进行句子或短语词形还原
from nltk.stem import WordNetLemmatizerwnl = WordNetLemmatizer()text = ['She gripped the armrest as he passed two cars at a time.', 'Her car was in full view.', 'A number of cars carried out of state license plates.']output = []for sentence in text: output.append(" ".join([wnl.lemmatize(i) for i in sentence.split()]))for item in output: print(item)print("*" * 10)print(wnl.lemmatize('jumps', 'n'))print(wnl.lemmatize('jumping', 'v'))print(wnl.lemmatize('jumped', 'v'))print("*" * 10)print(wnl.lemmatize('saddest', 'a'))print(wnl.lemmatize('happiest', 'a'))print(wnl.lemmatize('easiest', 'a'))Output:
She gripped the armrest a he passed two car at a time.Her car wa in full view.A number of car carried out of state license plates.**********jumpjumpjump**********sadhappyeasy应用 NLTK 从文本文件中查找每个单词的频率
import nltkfrom nltk.corpus import webtextfrom nltk.probability import FreqDist nltk.download('webtext')wt_words = webtext.words('testing.txt')data_analysis = nltk.FreqDist(wt_words) # Let's take the specific words only if their frequency is greater than 3.filter_words = dict([(m, n) for m, n in data_analysis.items() if len(m) > 3]) for key in sorted(filter_words): print("%s: %s" % (key, filter_words[key])) data_analysis = nltk.FreqDist(filter_words) data_analysis.plot(25, cumulative=False)Output:
[nltk_data] Downloading package webtext to[nltk_data] C:\Users\amit\AppData\Roaming\nltk_data...[nltk_data] Unzipping corpora\webtext.zip.1989: 1Accessing: 1Analysis: 1Anyone: 1Chapter: 1Coding: 1Data: 1...从语料库中创立词云
import nltkfrom nltk.corpus import webtextfrom nltk.probability import FreqDistfrom wordcloud import WordCloudimport matplotlib.pyplot as plt nltk.download('webtext')wt_words = webtext.words('testing.txt') # Sample datadata_analysis = nltk.FreqDist(wt_words) filter_words = dict([(m, n) for m, n in data_analysis.items() if len(m) > 3]) wcloud = WordCloud().generate_from_frequencies(filter_words) # Plotting the wordcloudplt.imshow(wcloud, interpolation="bilinear") plt.axis("off")(-0.5, 399.5, 199.5, -0.5)plt.show()NLTK 词法分布图
import nltkfrom nltk.corpus import webtextfrom nltk.probability import FreqDistfrom wordcloud import WordCloudimport matplotlib.pyplot as plt words = ['data', 'science', 'dataset'] nltk.download('webtext')wt_words = webtext.words('testing.txt') # Sample data points = [(x, y) for x in range(len(wt_words)) for y in range(len(words)) if wt_words[x] == words[y]] if points: x, y = zip(*points)else: x = y = () plt.plot(x, y, "rx", scalex=.1)plt.yticks(range(len(words)), words, color="b")plt.ylim(-1, len(words))plt.title("Lexical Dispersion Plot")plt.xlabel("Word Offset")plt.show()应用 countvectorizer 将文本转换为数字
import pandas as pdfrom sklearn.feature_extraction.text import CountVectorizer # Sample data for analysisdata1 = "Java is a language for programming that develops a software for several platforms. A compiled code or bytecode on Java application can run on most of the operating systems including Linux, Mac operating system, and Linux. Most of the syntax of Java is derived from the C++ and C languages."data2 = "Python supports multiple programming paradigms and comes up with a large standard library, paradigms included are object-oriented, imperative, functional and procedural."data3 = "Go is typed statically compiled language. It was created by Robert Griesemer, Ken Thompson, and Rob Pike in 2009. This language offers garbage collection, concurrency of CSP-style, memory safety, and structural typing." df1 = pd.DataFrame({'Java': [data1], 'Python': [data2], 'Go': [data2]}) # Initializevectorizer = CountVectorizer()doc_vec = vectorizer.fit_transform(df1.iloc[0]) # Create dataFramedf2 = pd.DataFrame(doc_vec.toarray().transpose(), index=vectorizer.get_feature_names()) # Change column headersdf2.columns = df1.columnsprint(df2)Output:
Go Java Pythonand 2 2 2application 0 1 0are 1 0 1bytecode 0 1 0can 0 1 0code 0 1 0comes 1 0 1compiled 0 1 0derived 0 1 0develops 0 1 0for 0 2 0from 0 1 0functional 1 0 1imperative 1 0 1...应用 TF-IDF 创立文档术语矩阵
import pandas as pdfrom sklearn.feature_extraction.text import TfidfVectorizer# Sample data for analysisdata1 = "Java is a language for programming that develops a software for several platforms. A compiled code or bytecode on Java application can run on most of the operating systems including Linux, Mac operating system, and Linux. Most of the syntax of Java is derived from the C++ and C languages."data2 = "Python supports multiple programming paradigms and comes up with a large standard library, paradigms included are object-oriented, imperative, functional and procedural."data3 = "Go is typed statically compiled language. It was created by Robert Griesemer, Ken Thompson, and Rob Pike in 2009. This language offers garbage collection, concurrency of CSP-style, memory safety, and structural typing."df1 = pd.DataFrame({'Java': [data1], 'Python': [data2], 'Go': [data2]})# Initializevectorizer = TfidfVectorizer()doc_vec = vectorizer.fit_transform(df1.iloc[0])# Create dataFramedf2 = pd.DataFrame(doc_vec.toarray().transpose(), index=vectorizer.get_feature_names())# Change column headersdf2.columns = df1.columnsprint(df2)Output:
Go Java Pythonand 0.323751 0.137553 0.323751application 0.000000 0.116449 0.000000are 0.208444 0.000000 0.208444bytecode 0.000000 0.116449 0.000000can 0.000000 0.116449 0.000000code 0.000000 0.116449 0.000000comes 0.208444 0.000000 0.208444compiled 0.000000 0.116449 0.000000derived 0.000000 0.116449 0.000000develops 0.000000 0.116449 0.000000for 0.000000 0.232898 0.000000...为给定句子生成 N-gram
NLTK
import nltkfrom nltk.util import ngrams# Function to generate n-grams from sentences.def extract_ngrams(data, num): n_grams = ngrams(nltk.word_tokenize(data), num) return [ ' '.join(grams) for grams in n_grams]data = 'A class is a blueprint for the object.'print("1-gram: ", extract_ngrams(data, 1))print("2-gram: ", extract_ngrams(data, 2))print("3-gram: ", extract_ngrams(data, 3))print("4-gram: ", extract_ngrams(data, 4))TextBlob
from textblob import TextBlob # Function to generate n-grams from sentences.def extract_ngrams(data, num): n_grams = TextBlob(data).ngrams(num) return [ ' '.join(grams) for grams in n_grams] data = 'A class is a blueprint for the object.' print("1-gram: ", extract_ngrams(data, 1))print("2-gram: ", extract_ngrams(data, 2))print("3-gram: ", extract_ngrams(data, 3))print("4-gram: ", extract_ngrams(data, 4))Output:
1-gram: ['A', 'class', 'is', 'a', 'blueprint', 'for', 'the', 'object']2-gram: ['A class', 'class is', 'is a', 'a blueprint', 'blueprint for', 'for the', 'the object']3-gram: ['A class is', 'class is a', 'is a blueprint', 'a blueprint for', 'blueprint for the', 'for the object']4-gram: ['A class is a', 'class is a blueprint', 'is a blueprint for', 'a blueprint for the', 'blueprint for the object']应用带有二元组的 sklearn CountVectorize 词汇标准
import pandas as pdfrom sklearn.feature_extraction.text import CountVectorizer # Sample data for analysisdata1 = "Machine language is a low-level programming language. It is easily understood by computers but difficult to read by people. This is why people use higher level programming languages. Programs written in high-level languages are also either compiled and/or interpreted into machine language so that computers can execute them."data2 = "Assembly language is a representation of machine language. In other words, each assembly language instruction translates to a machine language instruction. Though assembly language statements are readable, the statements are still low-level. A disadvantage of assembly language is that it is not portable, because each platform comes with a particular Assembly Language" df1 = pd.DataFrame({'Machine': [data1], 'Assembly': [data2]}) # Initializevectorizer = CountVectorizer(ngram_range=(2, 2))doc_vec = vectorizer.fit_transform(df1.iloc[0]) # Create dataFramedf2 = pd.DataFrame(doc_vec.toarray().transpose(), index=vectorizer.get_feature_names()) # Change column headersdf2.columns = df1.columnsprint(df2)Output:
Assembly Machinealso either 0 1and or 0 1are also 0 1are readable 1 0are still 1 0assembly language 5 0because each 1 0but difficult 0 1by computers 0 1by people 0 1can execute 0 1...应用 TextBlob 提取名词短语
from textblob import TextBlob#Extract nounblob = TextBlob("Canada is a country in the northern part of North America.")for nouns in blob.noun_phrases: print(nouns)Output:
canadanorthern partamerica如何计算词-词共现矩阵
import numpy as npimport nltkfrom nltk import bigramsimport itertoolsimport pandas as pd def generate_co_occurrence_matrix(corpus): vocab = set(corpus) vocab = list(vocab) vocab_index = {word: i for i, word in enumerate(vocab)} # Create bigrams from all words in corpus bi_grams = list(bigrams(corpus)) # Frequency distribution of bigrams ((word1, word2), num_occurrences) bigram_freq = nltk.FreqDist(bi_grams).most_common(len(bi_grams)) # Initialise co-occurrence matrix # co_occurrence_matrix[current][previous] co_occurrence_matrix = np.zeros((len(vocab), len(vocab))) # Loop through the bigrams taking the current and previous word, # and the number of occurrences of the bigram. for bigram in bigram_freq: current = bigram[0][1] previous = bigram[0][0] count = bigram[1] pos_current = vocab_index[current] pos_previous = vocab_index[previous] co_occurrence_matrix[pos_current][pos_previous] = count co_occurrence_matrix = np.matrix(co_occurrence_matrix) # return the matrix and the index return co_occurrence_matrix, vocab_index text_data = [['Where', 'Python', 'is', 'used'], ['What', 'is', 'Python' 'used', 'in'], ['Why', 'Python', 'is', 'best'], ['What', 'companies', 'use', 'Python']] # Create one list using many listsdata = list(itertools.chain.from_iterable(text_data))matrix, vocab_index = generate_co_occurrence_matrix(data) data_matrix = pd.DataFrame(matrix, index=vocab_index, columns=vocab_index)print(data_matrix)Output:
best use What Where ... in is Python usedbest 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 1.0use 0.0 0.0 0.0 0.0 ... 0.0 1.0 0.0 0.0What 1.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0Where 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0Pythonused 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 1.0Why 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 1.0companies 0.0 1.0 0.0 1.0 ... 1.0 0.0 0.0 0.0in 0.0 0.0 0.0 0.0 ... 0.0 0.0 1.0 0.0is 0.0 0.0 1.0 0.0 ... 0.0 0.0 0.0 0.0Python 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0used 0.0 0.0 1.0 0.0 ... 0.0 0.0 0.0 0.0 [11 rows x 11 columns]应用 TextBlob 进行情感剖析
from textblob import TextBlobdef sentiment(polarity): if blob.sentiment.polarity < 0: print("Negative") elif blob.sentiment.polarity > 0: print("Positive") else: print("Neutral")blob = TextBlob("The movie was excellent!")print(blob.sentiment)sentiment(blob.sentiment.polarity)blob = TextBlob("The movie was not bad.")print(blob.sentiment)sentiment(blob.sentiment.polarity)blob = TextBlob("The movie was ridiculous.")print(blob.sentiment)sentiment(blob.sentiment.polarity)Output:
Sentiment(polarity=1.0, subjectivity=1.0)PositiveSentiment(polarity=0.3499999999999999, subjectivity=0.6666666666666666)PositiveSentiment(polarity=-0.3333333333333333, subjectivity=1.0)Negative应用 Goslate 进行语言翻译
import goslatetext = "Comment vas-tu?"gs = goslate.Goslate()translatedText = gs.translate(text, 'en')print(translatedText)translatedText = gs.translate(text, 'zh')print(translatedText)translatedText = gs.translate(text, 'de')print(translatedText)应用 TextBlob 进行语言检测和翻译
from textblob import TextBlob blob = TextBlob("Comment vas-tu?") print(blob.detect_language()) print(blob.translate(to='es'))print(blob.translate(to='en'))print(blob.translate(to='zh'))Output:
fr¿Como estas tu?How are you?你好吗？应用 TextBlob 获取定义和同义词
from textblob import TextBlobfrom textblob import Word text_word = Word('safe') print(text_word.definitions) synonyms = set()for synset in text_word.synsets: for lemma in synset.lemmas(): synonyms.add(lemma.name()) print(synonyms)Output:
['strongbox where valuables can be safely kept', 'a ventilated or refrigerated cupboard for securing provisions from pests', 'contraceptive device consisting of a sheath of thin rubber or latex that is worn over the penis during intercourse', 'free from danger or the risk of harm', '(of an undertaking) secure from risk', 'having reached a base without being put out', 'financially sound']{'secure', 'rubber', 'good', 'safety', 'safe', 'dependable', 'condom', 'prophylactic'}应用 TextBlob 获取反义词列表
from textblob import TextBlobfrom textblob import Wordtext_word = Word('safe')antonyms = set()for synset in text_word.synsets: for lemma in synset.lemmas(): if lemma.antonyms(): antonyms.add(lemma.antonyms()[0].name()) print(antonyms)Output:
{'dangerous', 'out'}本文由mdnice多平台公布