Some words to CSDN readers: be grateful for the company of these ten years, live up to the meeting and disappear for a short time

1. Data crawling

The specific code is not introduced here to protect the original CSDN, but the corresponding core knowledge points will be given. Readers are advised to grasp text knowledge in combination with their own direction.

Core expansion pack:

• import requests
• from lxml import etree
• import csv

Core process:

• Solve the header problem
• Solve page turning problem
• Review element analysis DOM tree structure
• The positioning node is analyzed by Xpath
• Earn titles, URL s, times, reads, and comments, respectively
• Detail page capture

Crawler output results, it is recommended to learn pile driving output (multi-purpose print).

The sorted results are shown in the figure below, and the contents are output to CSV storage.

2. Metrological statistics and visual analysis

(1) Analyze the author's writing habits by hour First, let's analyze the blog writing habits of the author "Eastmount". At the same time, we use Matplotlib and pyecarts to draw graphics. We find that the graphics drawn by ecarts are better. As can be seen from the figure, the author has been writing blogs late at night and in the afternoon for a long time.

The source code is as follows:

# encoding:utf-8 """ By: Easmount CSDN 2021-04-19 """ import re import time import csv import pandas as pd import numpy as np #------------------------------------------------------------------------------ #Step 1: read data dd = [] #date tt = [] #time with open("data.csv", "r", encoding="utf8") as csvfile: csv_reader = csv.reader(csvfile) k = 0 for row in csv_reader: if k==0: #Skip title k = k + 1 continue #Get data 2021-04-08 21:52:21 value_date = row[4] value_time = row[5] hour = value_time.split(":")[0] hour = int(hour) dd.append(row[4]) tt.append(hour) #print(row[4],row[5]) #print(hour) k = k + 1 print(len(tt),len(dd)) print(dd) print(tt) #------------------------------------------------------------------------------ #The second step is to count the number of different hours from collections import Counter cnt = Counter(tt) print(cnt.items()) #dict_items #Dictionary key sorting list_time = [] list_tnum = [] for i in sorted(cnt): print(i,cnt[i]) list_time.append(i) list_tnum.append(cnt[i]) #------------------------------------------------------------------------------ #Step 3 draw a histogram import matplotlib.pyplot as plt N = 24 ind = np.arange(N) width=0.35 plt.bar(ind, list_tnum, width, color='r', label='hour') plt.xticks(ind+width/2, list_time, rotation=40) plt.title("The Eastmount's blog is distributed by the hour") plt.xlabel('hour') plt.ylabel('numbers') plt.savefig('Eastmount-01.png',dpi=400) plt.show() #------------------------------------------------------------------------------ #Step 4: PyEcharts draw a histogram from pyecharts import options as opts from pyecharts.charts import Bar bar=( Bar() .add_xaxis(list_time) .add_yaxis("quantity", list_tnum, color="blue") .set_global_opts(title_opts=opts.TitleOpts( title="Eastmount Blogs are distributed by hour", subtitle="hour")) ) bar.render('01-Eastmount Blogs are distributed by hour.html')

(2) Blog statistics by month The author writes a blog by month, as shown in the figure below. In 2015, he wrote a lot of LeetCode code when looking for a job. Later, he shared more security during his blog reading.

The source code is as follows:

# encoding:utf-8 """ By: Easmount CSDN 2021-04-19 """ import re import time import csv import pandas as pd import numpy as np #------------------------------------------------------------------------------ #Step 1: read data dd = [] #date tt = [] #time with open("data.csv", "r", encoding="utf8") as csvfile: csv_reader = csv.reader(csvfile) k = 0 for row in csv_reader: if k==0: #Skip title k = k + 1 continue #Get data 2021-04-08 21:52:21 value_date = row[4] value_time = row[5] hour = value_time.split(":")[0] #Get hours hour = int(hour) month = value_date[:7] #Get month dd.append(month) tt.append(hour) #print(row[4],row[5]) #print(hour,month) print(month) k = k + 1 #break print(len(tt),len(dd)) print(dd) print(tt) #------------------------------------------------------------------------------ #Step 2 count the number of different dates from collections import Counter cnt = Counter(dd) print(cnt.items()) #dict_items #Dictionary key sorting list_date = [] list_dnum = [] for i in sorted(cnt): print(i,cnt[i]) list_date.append(i) list_dnum.append(cnt[i]) #------------------------------------------------------------------------------ #Step 3: PyEcharts draw a histogram from pyecharts import options as opts from pyecharts.charts import Bar from pyecharts.charts import Line from pyecharts.commons.utils import JsCode line = ( Line() .add_xaxis(list_date) .add_yaxis('quantity', list_dnum, is_smooth=True, markline_opts=opts.MarkLineOpts(data=[opts.MarkLineItem(type_="average")]), markpoint_opts=opts.MarkPointOpts(data=[opts.MarkPointItem(type_="max"), opts.MarkPointItem(type_="min")])) # Hide numbers set area .set_series_opts( areastyle_opts=opts.AreaStyleOpts(opacity=0.5), label_opts=opts.LabelOpts(is_show=False)) # Sets the rotation angle of the x-axis label .set_global_opts(xaxis_opts=opts.AxisOpts(axislabel_opts=opts.LabelOpts(rotate=-30)), yaxis_opts=opts.AxisOpts(name='quantity', min_=3), title_opts=opts.TitleOpts(title='Eastmount Blog distribution by date')) ) line.render('02-Eastmount Blog distribution by date.html')

(3) Blog statistics by week Statistics by week are as follows. Call the date.weekday() function to output the corresponding week. The author updates a little less at the weekend.

The core code is as follows:

# encoding:utf-8 """ By: Easmount CSDN 2021-04-19 """ import re import time import csv import pandas as pd import numpy as np import datetime #Define week function def get_week_day(date): week_day_dict = { 0 : 'Monday', 1 : 'Tuesday', 2 : 'Wednesday', 3 : 'Thursday', 4 : 'Friday', 5 : 'Saturday', 6 : 'Sunday' } day = date.weekday() return week_day_dict[day] #------------------------------------------------------------------------------ #Step 1: read data dd = [] #date tt = [] #time ww = [] #week with open("data.csv", "r", encoding="utf8") as csvfile: csv_reader = csv.reader(csvfile) k = 0 for row in csv_reader: if k==0: #Skip title k = k + 1 continue #Get data 2021-04-08 21:52:21 value_date = row[4] value_time = row[5] hour = value_time.split(":")[0] #Get hours hour = int(hour) month = value_date[:7] #Get month dd.append(month) tt.append(hour) #Get week date = datetime.datetime.strptime(value_date, '%Y-%m-%d').date() week = get_week_day(date) ww.append(week) #print(date,week) k = k + 1 print(len(tt),len(dd),len(ww)) print(dd) print(tt) print(ww) #------------------------------------------------------------------------------ #Step 2 count the number of different dates from collections import Counter cnt = Counter(ww) print(cnt.items()) #dict_items #Dictionary key sorting list_date = ['Monday','Tuesday','Wednesday','Thursday','Friday','Saturday','Sunday'] list_dnum = [0,0,0,0,0,0,0] for key,value in cnt.items(): k = 0 while k<len(list_date): if key==list_date[k]: list_dnum[k] = value break k = k + 1 print(list_date,list_dnum) #------------------------------------------------------------------------------ #Step 3: PyEcharts draw a histogram from pyecharts import options as opts from pyecharts.charts import Bar from pyecharts.charts import Line from pyecharts.commons.utils import JsCode bar=( Bar() .add_xaxis(list_date) .add_yaxis("quantity", list_dnum, color='pink') .set_global_opts(title_opts=opts.TitleOpts( title="Eastmount Blog distribution by week", subtitle="week")) ) bar.render('03-Eastmount Blog distribution by week.html')

3. Core word statistics and word cloud analysis

Word cloud analysis is very suitable for beginners. Here, the author also briefly shares the process of core topic word statistics and word cloud analysis.

(1) Statistical core keywords and word frequency The output results are shown in the figure below:

The code is as follows:

# coding=utf-8 """ By: Easmount CSDN 2021-04-19 """ import jieba import re import time import csv from collections import Counter #------------------------------------Chinese word segmentation---------------------------------- cut_words = "" all_words = "" stopwords = ["[", "]", ")", "(", ")", "(", "[", "]", ".", ",", "-", "—", ":", ": ", "<", ">", "of", "and", "of", "and", """, """, "？", "?"] #Import custom dictionary #jieba.load_userdict("dict.txt") f = open('06-data-fenci.txt', 'w') with open("data.csv", "r", encoding="utf8") as csvfile: csv_reader = csv.reader(csvfile) k = 0 for row in csv_reader: if k==0: #Skip title k = k + 1 continue #get data title = row[1] title = title.strip('\n') #print(title) #participle cut_words = "" seg_list = jieba.cut(title,cut_all=False) for seg in seg_list: if seg not in stopwords: cut_words += seg + " " #cut_words = (" ".join(seg_list)) f.write(cut_words+"\n") all_words += cut_words k = k + 1 f.close() #Output results all_words = all_words.split() print(all_words) #------------------------------------Word frequency statistics---------------------------------- c = Counter() for x in all_words: if len(x)>1 and x != '\r\n': c[x] += 1 #Output the top 10 words with the highest word frequency print('\n Word frequency statistics:') for (k,v) in c.most_common(10): print("%s:%d"%(k,v)) #Store data name ="06-data-word.csv" fw = open(name, 'w', encoding='utf-8') i = 1 for (k,v) in c.most_common(len(c)): fw.write(str(i)+','+str(k)+','+str(v)+'\n') i = i + 1 else: print("Over write file!") fw.close()

(2) PyEcharts word cloud visualization The output results are shown in the figure below. The words with higher word frequency are displayed larger and brighter.

The code is as follows:

# coding=utf-8 """ By: Easmount CSDN 2021-04-19 """ import jieba import re import time import csv from collections import Counter #------------------------------------Chinese word segmentation---------------------------------- cut_words = "" all_words = "" stopwords = ["[", "]", ")", "(", ")", "(", "[", "]", "01", "02", "03", "04", "05", "06", "07", "08", "09", "what"] f = open('06-data-fenci.txt', 'w') with open("data.csv", "r", encoding="utf8") as csvfile: csv_reader = csv.reader(csvfile) k = 0 for row in csv_reader: if k==0: #Skip title k = k + 1 continue #get data title = row[1] title = title.strip('\n') #print(title) #participle cut_words = "" seg_list = jieba.cut(title,cut_all=False) for seg in seg_list: if seg not in stopwords: cut_words += seg + " " #cut_words = (" ".join(seg_list)) f.write(cut_words+"\n") all_words += cut_words k = k + 1 f.close() #Output results all_words = all_words.split() print(all_words) #------------------------------------Word frequency statistics---------------------------------- c = Counter() for x in all_words: if len(x)>1 and x != '\r\n': c[x] += 1 #Output the top 10 words with the highest word frequency print('\n Word frequency statistics:') for (k,v) in c.most_common(10): print("%s:%d"%(k,v)) #Store data name ="06-data-word.csv" fw = open(name, 'w', encoding='utf-8') i = 1 for (k,v) in c.most_common(len(c)): fw.write(str(i)+','+str(k)+','+str(v)+'\n') i = i + 1 else: print("Over write file!") fw.close() #------------------------------------Word cloud analysis---------------------------------- from pyecharts import options as opts from pyecharts.charts import WordCloud from pyecharts.globals import SymbolType # Generate data word = [('A',10), ('B',9), ('C',8)] list + Tuple words = [] for (k,v) in c.most_common(200): # print(k, v) words.append((k,v)) # Rendering def wordcloud_base() -> WordCloud: c = ( WordCloud() .add("", words, word_size_range=[20, 40], shape='diamond') #shape=SymbolType.ROUND_RECT .set_global_opts(title_opts=opts.TitleOpts(title='Eastmount Ten year blog word cloud map')) ) return c # Generate graph wordcloud_base().render('05-Eastmount Ten year blog word cloud map.html')

4.LDA theme mining

LDA model is a very classic algorithm in text mining or topic mining. Readers can read the author's previous articles and introduce the model in detail. Here, we use it to mine the author's blog topics. The number of topics set is 4. It is usually necessary to calculate the confusion comparison.

At the same time, calculate the subject words corresponding to each subject, as shown below. Note that readers are recommended to optimize their own text accordingly, which will get subject words that are more in line with the real situation, and there will be mutual blending between topics. For example, there will be Python related penetration articles in the security series blog.

The complete code is as follows:

#coding: utf-8 import pandas as pd from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer #---------------------Step 1: read data (word segmentation)---------------------- corpus = [] # Read expected a line is expected to be a document for line in open('06-data-fenci.txt', 'r').readlines(): corpus.append(line.strip()) #-----------------------Step 2: calculate TF-IDF value----------------------- # Set number of features n_features = 2000 tf_vectorizer = TfidfVectorizer(strip_accents = 'unicode', max_features=n_features, stop_words=['of','or','etc.','yes','have','of','And','sure','still','here', 'One','and','also','cover','Do you','to','in','most','however','everybody', 'once','How many days?','200','also','A look','300','50','Ha ha ha ha', '"','"','. ','，','？',',','；','Yes?','originally','find', 'and','in','of','the','We','always','really','18','once', 'Yes','Some','already','no','such','one by one','one day','this','such', 'one kind','be located','one of','sky','No,','quite a lot','somewhat','what','Five', 'especially'], max_df = 0.99, min_df = 0.002) #Remove words that are too likely to appear in the document tf = tf_vectorizer.fit_transform(corpus) print(tf.shape) print(tf) #-------------------------Step 3 LDA analysis------------------------ from sklearn.decomposition import LatentDirichletAllocation # Set number of topics n_topics = 4 lda = LatentDirichletAllocation(n_components=n_topics, max_iter=100, learning_method='online', learning_offset=50, random_state=0) lda.fit(tf) # Show number of topics model.topic_word_ print(lda.components_) # Several topics are rows, and how many keywords are columns print(lda.components_.shape) # Computational confusion print(u'Confusion:') print(lda.perplexity(tf,sub_sampling = False)) # Topic keyword distribution def print_top_words(model, tf_feature_names, n_top_words): for topic_idx,topic in enumerate(model.components_): # lda.component is equivalent to model.topic_word_ print('Topic #%d:' % topic_idx) print(' '.join([tf_feature_names[i] for i in topic.argsort()[:-n_top_words-1:-1]])) print("") # After defining the function, temporarily output the first 20 keywords for each topic n_top_words = 20 tf_feature_names = tf_vectorizer.get_feature_names() # Call function print_top_words(lda, tf_feature_names, n_top_words) #------------------------Step 4 visual analysis------------------------- import pyLDAvis import pyLDAvis.sklearn #pyLDAvis.enable_notebook() data = pyLDAvis.sklearn.prepare(lda,tf,tf_vectorizer) print(data) #display graphics pyLDAvis.show(data) pyLDAvis.save_json(data,' 06-fileobj.html')

5. Hierarchical clustering topic tree

The tree view drawn by hierarchical clustering is also a common technology in the field of text mining. It will display the topics related to each field in the form of a tree. The output results here are shown in the following figure:

Note that here, the author can set filtering to display the number of subject words displayed in the tree view, and conduct relevant comparative experiments to find the best results.

# -*- coding: utf-8 -*- import os import codecs from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer from sklearn.manifold import TSNE from sklearn.cluster import KMeans import matplotlib.pyplot as plt import numpy as np import pandas as pd import jieba from sklearn import metrics from sklearn.metrics import silhouette_score from array import array from numpy import * from pylab import mpl from sklearn.metrics.pairwise import cosine_similarity import matplotlib.pyplot as plt import matplotlib as mpl from scipy.cluster.hierarchy import ward, dendrogram #---------------------------------------Loading corpus------------------------------------- text = open('06-data-fenci.txt').read() print(text) list1=text.split("\n") print(list1) print(list1[0]) print(list1[1]) mytext_list=list1 #Control the number of displays count_vec = CountVectorizer(min_df=20, max_df=1000) #Maximum ignored xx1 = count_vec.fit_transform(list1).toarray() word=count_vec.get_feature_names() print("word feature length: {}".format(len(word))) print(word) print(xx1) print(type(xx1)) print(xx1.shape) print(xx1[0]) #---------------------------------------Hierarchical clustering------------------------------------- titles = word #dist = cosine_similarity(xx1) mpl.rcParams['font.sans-serif'] = ['SimHei'] df = pd.DataFrame(xx1) print(df.corr()) print(df.corr('spearman')) print(df.corr('kendall')) dist = df.corr() print (dist) print(type(dist)) print(dist.shape) #define the linkage_matrix using ward clustering pre-computed distances linkage_matrix = ward(dist) fig, ax = plt.subplots(figsize=(8, 12)) # set size ax = dendrogram(linkage_matrix, orientation="right", p=20, labels=titles, leaf_font_size=12 ) #leaf_rotation=90., leaf_font_size=12. #show plot with tight layout plt.tight_layout() #save figure as ward_clusters plt.savefig('07-KH.png', dpi=200) plt.show()

6. Social network analysis

Social network analysis is often used in citation analysis. Some in the field of liberal arts become literature knowledge map (different from the knowledge map or ontology proposed by Google). It is also a common technical means in the field of literature mining. Here we draw the social network relationship map as shown below, mainly using Gephi software, and Neo4j or D3 are also recommended. It can be seen that the author's ten-year sharing mainly focuses on four pieces of content, which are interrelated and complementary.

• network security
• Python
• backward analysis
• Basic knowledge or programming technology

Recommended author's article:

• [Pyhon big data analysis] v. news topic capture and Gephi construction of topic knowledge map

In the first step, we need to calculate the pairwise co-occurrence matrix. Too much data may overflow the boundary. The output result is shown in the figure below. At this time, we hope to filter the stop words or delete the abnormal relationship.

# -*- coding: utf-8 -*- """ @author: eastmount CSDN 2020-04-20 """ import pandas as pd import numpy as np import codecs import networkx as nx import matplotlib.pyplot as plt import csv from scipy.sparse import coo_matrix #---------------------------Step 1: read data------------------------------- word = [] #Record keywords f = open("06-data-fenci.txt", encoding='gbk') line = f.readline() while line: #print line line = line.replace("\n", "") #Filter line breaks line = line.strip('\n') for n in line.split(' '): #print n if n not in word: word.append(n) line = f.readline() f.close() print(len(word)) #Total number of keywords 2913 #--------------------------The second step is to calculate the co-occurrence matrix---------------------------- a = np.zeros([2,3]) print(a) #Co-occurrence matrix #word_vector = np.zeros([len(word),len(word)], dtype='float16') #MemoryError: the matrix is too large. Coo is used to report memory errors_ The matrix function solves this problem print(len(word)) #Type < type 'numpy. Ndarray' > word_vector = coo_matrix((len(word),len(word)), dtype=np.int8).toarray() print(word_vector.shape) f = open("06-data-fenci.txt", encoding='gbk') line = f.readline() while line: line = line.replace("\n", "") #Filter line breaks line = line.strip('\n') #Filter line breaks nums = line.split(' ') #Cycle through the keyword location and set word_vector count i = 0 j = 0 while i<len(nums): #ABCD co-occurrence AB AC AD BC BD CD plus 1 j = i + 1 w1 = nums[i] #First word while j<len(nums): w2 = nums[j] #Second word #Find the subscript corresponding to the word from the word array k = 0 n1 = 0 while k<len(word): if w1==word[k]: n1 = k break k = k +1 #Find the second keyword location k = 0 n2 = 0 while k<len(word): if w2==word[k]: n2 = k break k = k +1 #Key: the assignment of word frequency matrix only calculates the upper triangle if n1<=n2: word_vector[n1][n2] = word_vector[n1][n2] + 1 else: word_vector[n2][n1] = word_vector[n2][n1] + 1 #print(n1, n2, w1, w2) j = j + 1 i = i + 1 #Read new content line = f.readline() #print("next line") f.close() print("over computer") #--------------------------Step 3: write CSV file-------------------------- c = open("word-word-weight.csv","w", encoding='utf-8', newline='') #Resolve blank lines #c.write(codecs.BOM_UTF8) #Prevent garbled code writer = csv.writer(c) #Write object writer.writerow(['Word1', 'Word2', 'Weight']) i = 0 while i<len(word): w1 = word[i] j = 0 while j<len(word): w2 = word[j] #Judge whether two words co-exist, and write the file whose co-occurrence word frequency is not 0 if word_vector[i][j]>0: #write file templist = [] templist.append(w1) templist.append(w2) templist.append(str(int(word_vector[i][j]))) #print templist writer.writerow(templist) j = j + 1 i = i + 1 c.close()

In the second step, we need to build CSV files of entities (nodes) and relationships (edges). As shown in the figure below:

• entity-clean.csv
• rela-clean.csv

The third step is to create a new project, select "data" and enter the spreadsheet. Import the node table and select the entity table.

Step 4: import data and set it as "edge table". Note that CSV table data must be set as Source, Target and Weight. This must be consistent with Gephi format, otherwise an error will be prompted when importing data.

Step 5: after successful import, click "Overview" to display as follows, and then adjust parameters.

Step 6: set modularity. Click "run" in the statistics on the right to set modularity. At the same time, set the average path length. Click "run" in the statistics on the right to set the edge overview.

Step 7: reset the node attributes. The node size value is set to "degree". The minimum value is 20 and the maximum value is 120. The node color value is set to "Modularity Class", indicating modularity.

Step 8: select "Fruchterman Reingold" in the layout. Adjust area, gravity and speed.

Step 9, click preview. Set the Song typeface, display the label, and adjust the transparency to 20, as shown in the figure below.

Step 10: map optimization and adjustment. At the same time, you can filter the weight or set the light color of the color module. For example, get a more detailed relationship map.

7. Blog emotion analysis

SnowNLP experiment is mainly used for emotion analysis. It is also recommended to use the emotion Dictionary of Dalian University of technology for optimization. The articles analyzed by the author are recommended here. The output results are shown in the figure below:

However, if we calculate the overall emotional score of daily or monthly news, we will reach the emotional analysis diagram of time series, so as to better predict the emotional trend. It is also widely used in the field of text mining or library and information.

# -*- coding: utf-8 -*- from snownlp import SnowNLP import codecs import os #Get emotional scores source = open("06-data-fenci.txt", "r", encoding='gbk') fw = open("09-result.txt", "w", encoding="gbk") line = source.readlines() sentimentslist = [] for i in line: s = SnowNLP(i) #print(s.sentiments) sentimentslist.append(s.sentiments) #Interval conversion to [- 0.5, 0.5] result = [] i = 0 while i<len(sentimentslist): result.append(sentimentslist[i]-0.5) fw.write(str(sentimentslist[i]-0.5)+"\n") print(sentimentslist[i]-0.5, line[i].strip("\n")) i = i + 1 fw.close() #Visual drawing import matplotlib.pyplot as plt import numpy as np plt.plot(np.arange(0, 598, 1), result, 'k-') plt.xlabel('Number') plt.ylabel('Sentiment') plt.title('Analysis of Sentiments') plt.show()

8. Analysis of blog theme evolution

The last is the subject of laboratory research. It is recommended that you read the core papers of NTU. In fact, theme evolution is usually divided into:

• Theme freshmen
• Theme extinction
• Theme fusion
• Theme loneliness

There are various calculation methods for topic fusion. You can find the most suitable method for your paper, such as word frequency, weight, O coefficient, relevance analysis and so on. It is recommended that you use Echarts to draw. The author's Atlas is shown in the figure below:

Note that the code given by the author here is another case. But the principle is the same, for reference only. The calculation process of the real situation is more complex, and the calculation evolution coefficient is usually decimal.

• https://echarts.apache.org/examples/ zh/index.html#chart-type-sankey

option = { series: { type: 'sankey', layout:'none', focusNodeAdjacency: 'allEdges', data: [ { name: 'T1-competitive intelligence ' },{ name: 'T1-enterprise' },{ name: 'T1-business management' }, { name: 'T1-Information research' },{ name: 'T1-competitor' },{ name: 'T1-Intelligence work' },{ name: 'T1-market economy' },{ name: 'T2-competitive intelligence ' },{ name: 'T2-enterprise' },{ name: 'T2-business management' },{ name: 'T2-Competitive intelligence system' },{ name: 'T2-competitor' },{ name: 'T2-knowledge management ' },{ name: 'T2-Anti competitive intelligence' },{ name: 'T3-competitive intelligence ' },{ name: 'T3-enterprise' },{ name: 'T3-Competitive intelligence system' },{ name: 'T3-business management' },{ name: 'T3-University Library' },{ name: 'T3-Anti competitive intelligence' },{ name: 'T3-knowledge management ' },{ name: 'T4-competitive intelligence ' },{ name: 'T4-enterprise' },{ name: 'T4-big data' },{ name: 'T4-Industrial Competitive Intelligence' },{ name: 'T4-Competitive intelligence system' },{ name: 'T4-University Library' },{ name: 'T4-business management' } ], links: [{ source: 'T1-competitive intelligence ', target: 'T2-competitive intelligence ', value: 10 }, { source: 'T1-enterprise', target: 'T2-enterprise', value: 7 }, { source: 'T1-business management', target: 'T2-business management', value: 6 },{ source: 'T1-Information research', target: 'T2-competitive intelligence ', value: 5 },{ source: 'T1-competitor', target: 'T2-competitor', value: 5 },{ source: 'T1-Intelligence work', target: 'T2-competitive intelligence ', value: 3 },{ source: 'T1-market economy', target: 'T2-competitive intelligence ', value: 3 },{ source: 'T1-competitive intelligence ', target: 'T2-Competitive intelligence system', value: 5 },{ source: 'T1-competitive intelligence ', target: 'T2-Competitive intelligence system', value: 3 },{ source: 'T1-competitive intelligence ', target: 'T2-knowledge management ', value: 3 },{ source: 'T1-competitive intelligence ', target: 'T2-Anti competitive intelligence', value: 3 }, { source: 'T2-competitive intelligence ', target: 'T3-competitive intelligence ', value: 10 },{ source: 'T2-enterprise', target: 'T3-enterprise', value: 7 },{ source: 'T3-Competitive intelligence system', target: 'T4-competitive intelligence ', value: 3 },{ source: 'T2-business management', target: 'T3-business management', value: 6 },{ source: 'T2-Competitive intelligence system', target: 'T3-Competitive intelligence system', value: 5 },{ source: 'T2-competitor', target: 'T3-competitive intelligence ', value: 5 },{ source: 'T2-knowledge management ', target: 'T3-knowledge management ', value: 3 },{ source: 'T2-Anti competitive intelligence', target: 'T3-Anti competitive intelligence', value: 3 },{ source: 'T2-competitive intelligence ', target: 'T3-University Library', value: 4 }, { source: 'T3-competitive intelligence ', target: 'T4-competitive intelligence ', value: 10 },{ source: 'T3-enterprise', target: 'T4-enterprise', value: 7 },{ source: 'T3-competitive intelligence ', target: 'T4-big data', value: 5 },{ source: 'T3-competitive intelligence ', target: 'T4-Industrial Competitive Intelligence', value: 5 },{ source: 'T3-Competitive intelligence system', target: 'T4-Competitive intelligence system', value: 6 },{ source: 'T3-business management', target: 'T4-business management', value: 4 }, { source: 'T3-University Library', target: 'T4-University Library', value: 4 },{ source: 'T3-Anti competitive intelligence', target: 'T4-competitive intelligence ', value: 3 },{ source: 'T3-knowledge management ', target: 'T4-competitive intelligence ', value: 2 } ] } };

The operation screenshot is as follows:

9. Expand knowledge

Readers can also carry out various text mining, such as:

• Named entity recognition
• Construction of knowledge map
• Intelligent question and answer processing
• Prediction of public opinion events
• ...

3, Summary

Finally, I sign my blog to end this article, "ignorance · optimism · low-key · humility · life". I always tell myself that I need to be optimistic to seek knowledge. The background of low-key is humility, and humility comes from the insight into life. We not only have work, study and programming, but also learn to enjoy life. Why should we go in such a hurry and do some happy things in life, It's good to write a few systematic articles and bring a loved one! Thank CSDN for your persistence and sharing. This is more than ten years.

Thank all readers for their ten years of company, and disappear for a better meeting. In the next three years, I am willing to accept my true self without inferiority or pride; I wish to work hard, live seriously, love what I love, no regrets, goodbye in the Jianghu.

Add new comment

0 comments