如何在 R 中转换文档术语矩阵？

Question

您好，我有一个文档术语矩阵，我使用tidy() 函数对其进行了转换，它运行良好。我想根据单词的频率绘制一个词云。所以我的转换表如下所示：

> head(Wcloud.Data)
# A tibble: 6 x 3
  document term       count
  <chr>    <chr>      <dbl>
1 1        accept         1
2 1        access         1
3 1        accomplish     1
4 1        account        4
5 1        accur          2
6 1        achiev         1

我有 33,647,383 个观察值，因此它是一个非常大的数据框。如果我使用max() 函数，我会得到一个非常高的数字（64116），但我的数据框中没有一个词的频率为 64116。此外，如果我用 wordcloud() 绘制闪亮的数据框，它会多次绘制相同的词。此外，如果我想对我的列进行排序 count 它不起作用 - sort(Wcloud.Data$count,decreasing = TRUE)。所以有些事情是不正确的，但我不知道，什么以及如何解决它。有人知道吗？

这是我的文档术语矩阵的摘要，在将其转换为数据框之前：

> observations.tf
<<DocumentTermMatrix (documents: 76717, terms: 4234)>>
Non-/sparse entries: 33647383/291172395
Sparsity           : 90%
Maximal term length: 15
Weighting          : term frequency (tf)

更新：我添加了我的数据框的图片

Answer 1

使用dplyr您可以如下：

library("tm")
library("SnowballC")
library("wordcloud")
library("RColorBrewer")

Wcloud.Data<- data.frame(Document= c(rep(1,6)), 
                         term = c("accept", "access","accomplish", "account", "accur", "achiev"),
                         count = c(1,1,1,4,2,1))

Data<-Wcloud.Data %>% 
  group_by(term) %>% 
  summarise(Frequency = sum(count))
set.seed(1234)
wordcloud(words = Data$term, freq = Data$Frequency, min.freq = 1,
          max.words=200, random.order=FALSE, rot.per=0.35, 
          colors=brewer.pal(8, "Dark2"))

在另一边，图书馆quanteda和tibble可以帮助您重写术语频率矩阵。我会把你的一个例子与它合作：

library(tibble)
library(quanteda)
Data <- data_frame(text = c("Chinese Beijing Chinese",
                              "Chinese Chinese Shanghai",
                              "this is china",
                              "china is here",
                              'hello china',
                              "Chinese Beijing Chinese",
                              "Chinese Chinese Shanghai",
                              "this is china",
                              "china is here",
                              'hello china',
                              "Kyoto Japan",
                              "Tokyo Japan Chinese",
                              "Kyoto Japan",
                              "Tokyo Japan Chinese",
                              "Kyoto Japan",
                              "Tokyo Japan Chinese",
                              "Kyoto Japan",
                              "Tokyo Japan Chinese",
                              'japan'))
DocTerm <- quanteda::dfm(Data$text)
DocTerm
# Document-feature matrix of: 19 documents, 11 features (78.5% sparse).
# 19 x 11 sparse Matrix of class "dfm"
# features
# docs     chinese beijing shanghai this is china here hello kyoto japan tokyo
# text1        2       1        0    0  0     0    0     0     0     0     0
# text2        2       0        1    0  0     0    0     0     0     0     0
# text3        0       0        0    1  1     1    0     0     0     0     0
# text4        0       0        0    0  1     1    1     0     0     0     0
# text5        0       0        0    0  0     1    0     1     0     0     0
# text6        2       1        0    0  0     0    0     0     0     0     0
# text7        2       0        1    0  0     0    0     0     0     0     0
# text8        0       0        0    1  1     1    0     0     0     0     0
# text9        0       0        0    0  1     1    1     0     0     0     0
# text10       0       0        0    0  0     1    0     1     0     0     0
# text11       0       0        0    0  0     0    0     0     1     1     0
# text12       1       0        0    0  0     0    0     0     0     1     1
# text13       0       0        0    0  0     0    0     0     1     1     0
# text14       1       0        0    0  0     0    0     0     0     1     1
# text15       0       0        0    0  0     0    0     0     1     1     0
# text16       1       0        0    0  0     0    0     0     0     1     1
# text17       0       0        0    0  0     0    0     0     1     1     0
# text18       1       0        0    0  0     0    0     0     0     1     1
# text19       0       0        0    0  0     0    0     0     0     1     0

Mat<-quanteda::convert(DocTerm,"data.frame")[,2:ncol(DocTerm)] # Converting to a Dataframe without taking into account the text variable
Result<- colSums(Mat) # This is what you are interested in
names(Result)<-colnames(Mat)
# > Result
# chinese  beijing shanghai     this       is    china     here    hello    kyoto    japan 
# 24        4        4        4        8       12        4        4        8       18