R- text挖掘中的水平合并数据帧

Question

如何在R中水平合并/合并3个数据帧？我有三个数据帧，一列中有一个单词，下一列中从文本中提取的单词计数，有点像这样：

  word.        count
1 hello.         6
2 test.          3
3 how.           8
4 are.           4
5 you.           1

我们来看一下dataframe 2：

  word.        count
1 hello.         6
2 test.          3
3 i.             3
4 am.            6
5 good.          2

我如何像这样合并它们：

  word.         df1.     df2.      total
1 hello.         6.       6.         12
2 test.          3        3           6
3 how.           8        0.          8 
4 are.           4        0           4
5 you.           1        0           1
6 i              0        3           3
7 am             0        6           6
8 good           0        2           2

因此，我有3个数据帧而不是2个数据帧

谢谢!

Answer 1

假设每一个都有相同的连接列：

out <- Reduce(function(a,b) merge(a, b, by = "word.", all = TRUE),
              list(df1, df2))
# normally this has `NA` in not-shared words, convert these NAs to 0
out[,-1] <- lapply(out[,-1], function(a) replace(a, is.na(a), 0))
out
#    word. count.x count.y
# 1    am.       0       6
# 2   are.       4       0
# 3  good.       0       2
# 4 hello.       6       6
# 5   how.       8       0
# 6     i.       0       3
# 7  test.       3       3
# 8   you.       1       0

不管你有多少帧，把它们都放在一个list中，这样就行了。(有关“帧列表”效率的讨论，请参阅https://stackoverflow.com/a/24376207/3358272。)

现在您需要做的就是更改列名。(有很多技术可以做到这一点。)

数据

df1 <- structure(list(word. = c("hello.", "test.", "how.", "are.", "you."), count = c(6L, 3L, 8L, 4L, 1L)), class = "data.frame", row.names = c("1", "2", "3", "4", "5"))
df2 <- structure(list(word. = c("hello.", "test.", "i.", "am.", "good."), count = c(6L, 3L, 3L, 6L, 2L)), class = "data.frame", row.names = c("1", "2", "3", "4", "5"))

Answer 2

大多数合并的R函数，或者使用更专业的术语，在表之间执行连接操作，被设计为一次在两个表上使用。在base R中，我们有用于这些连接操作的merge函数。由于您有两个以上的表要合并，即使使用此行为，您也可以使用Reduce函数轻松地跨表复制te join。定义如下：

这里是输入数据：

word <- c("hello.", "test.", "how.", "are.", "you.") 

df1 <- data.frame(
  word = word,
  count = 11:15
)

set.seed(1)
df2 <- data.frame(
  word = sample(word, size = 8, replace = T),
  value2 = rnorm(8)
)

set.seed(1)
df3 <- data.frame(
  word = word[c(3,4)],
  value3 = rnorm(2)
)

join的操作如下：：

list_dfs <- list(df1, df2, df3)

multi_inner <- Reduce(
  function(x, y, ...) merge(x, y, by = "word", all = TRUE), 
  list_dfs
)

结果如下：：

    word count       value2     value3
1   are.    14 -0.294720447  0.1836433
2 hello.    11 -0.928567035         NA
3 hello.    11 -0.005767173         NA
4   how.    13 -0.799009249 -0.6264538
5   how.    13 -0.289461574 -0.6264538
6  test.    12  2.404653389         NA
7  test.    12 -1.147657009         NA
8   you.    15  0.763593461         NA

现在，在这个示例中，我考虑到您希望保存这3个表之间的所有可能组合。如果只想识别所有表中出现的匹配项(或word列中的单词)，则需要将all参数设置为FALSE。如下所示：

multi_inner <- Reduce(
  function(x, y, ...) merge(x, y, by = "word", all = FALSE), 
  list_dfs
)

产生以下结果的：

  word count     value2     value3
1 are.    14 -0.2947204  0.1836433
2 how.    13 -0.7990092 -0.6264538
3 how.    13 -0.2894616 -0.6264538