tm 包中的 stemDocment 不适用于过去时词

Question

我有一个文件“check_text.txt”，其中包含“said say says make made”。我想对其进行词干提取以获得“say say say make make”。我尝试在tm 包中使用stemDocument，如下所示，但只得到“说说制造”。有没有办法对过去时词进行词干提取？在现实世界的自然语言处理中是否有必要这样做？谢谢！

filename = 'check_text.txt'
con <- file(filename, "rb")
text_data <- readLines(con,skipNul = TRUE)
close(con)
text_VS <- VectorSource(text_data)
text_corpus <- VCorpus(text_VS)
text_corpus <- tm_map(text_corpus, stemDocument, language = "english")
as.data.frame(text_corpus)$text

编辑：我也在SnowballC包中尝试了wordStem

> library(SnowballC)
> wordStem(c("said", "say", "says", "make", "made"))
[1] "said" "sai"  "sai"  "make" "made"

Answer 1

如果一个包中有一个不规则英语动词的数据集，这个任务就很容易了。我只是不知道任何包含此类数据的包，所以我选择通过抓取创建自己的数据库。我不确定这个网站是否涵盖所有不规则的单词。如有必要，您想搜索更好的网站以创建自己的数据库。一旦你有了你的数据库，你就可以开始你的任务了。

首先，我使用stemDocument() 并使用-s 清理当前表单。然后，我在words（即past）中收集了过去形式，过去形式的不定式（即inf1），在temp中确定了过去形式的顺序。我进一步确定了temp中过去表格的位置。我终于用不定式形式替换了 sat 形式。我对过去分词重复了同样的过程。

library(tm)
library(rvest)
library(dplyr)
library(splitstackshape)


### Create a database
x <- read_html("http://www.englishpage.com/irregularverbs/irregularverbs.html")

x %>%
html_table(header = TRUE) %>%
bind_rows %>%
rename(Past = `Simple Past`, PP = `Past Participle`) %>%
filter(!Infinitive %in% LETTERS) %>%
cSplit(splitCols = c("Past", "PP"),
       sep = " / ", direction = "long") %>%
filter(complete.cases(.)) %>%
mutate_each(funs(gsub(pattern = "\\s\\(.*\\)$|\\s\\[\\?\\]",
                      replacement = "",
                      x = .))) -> mydic

### Work on the task

words <- c("said", "drawn", "say", "says", "make", "made", "done")

### says to say
temp <- stemDocument(words)

### past forms become present form
### Collect past forms
past <- mydic$Past[which(mydic$Past %in% temp)]

### Collect infinitive forms of past forms
inf1 <- mydic$Infinitive[which(mydic$Past %in% temp)]

### Identify the order of past forms in temp
ind <- match(temp, past)
ind <- ind[is.na(ind) == FALSE]

### Where are the past forms in temp?
position <- which(temp %in% past)

temp[position] <- inf1[ind]

### Check
temp
#[1] "say"   "drawn" "say"   "say"   "make"  "make"  "done" 


### PP forms to infinitive forms (same as past forms)

pp <- mydic$PP[which(mydic$PP %in% temp)]
inf2 <- mydic$Infinitive[which(mydic$PP %in% temp)]
ind <- match(temp, pp)
ind <- ind[is.na(ind) == FALSE]
position <- which(temp %in% pp)
temp[position] <- inf2[ind]

### Check
temp
#[1] "say"  "draw" "say"  "say"  "make" "make" "do"