在我的经验中,这是一个数据准备和争论的问题,而不是一个解析问题,因为在本例中,tabulizer的解析算法除了在方法之间进行更改之外,没有提供太多的余地。从我所看到的,当我试图提取你的表,它不仅是第7页的表是错误的分析。每个页面的解析方式都不同,但所有的数据似乎都被保留了下来。我可以看到第一个表有13列,第二个17列,第三个12列,第四个10列,最后三个11列。我建议做的是单独解析每个页面,并根据每个页面上所需的输出执行数据清理,然后将它们绑定在一起。这是一个漫长的过程,对于解析的每个表都非常具体,因此我只提供一个示例脚本:
library(dplyr)
library(tidyr)
library(tabulizer)
# I create a dummy list to iterate through all the pages and push a data.frame in
result <- list()
for (i in 15:21){
out <- as.data.frame(extract_tables("mydocument.pdf", page = i, method = 'stream'), stringsAsFactors = FALSE)
result[[i]] <- out
}
# Remove excess list items -
# there is probably a better way to do this from within the for loop
result <- result[-(1:14)]
## ------- DATA CLEANING OPERATIONS examples:
# Remove top 3x lines from the first page of table1 not part of data
result[[1]] <- result[[1]][-(1:3),]
# Perform data cleaning operations such as split/ merge columns according to your liking
# for instance if you want to split column X1 into 4 (as in your original post), you can do that by splitting by whitespace
result[[1]] <- separate(result[[1]], 1, into = c('X1.1','X1.2','X1.3', 'X1.4'),sep = ' ', remove = TRUE)
## ---- After data cleaning operations:
# Bind all dataframes (they should have equal number of columns by now into one and make sure the colnames match as well)
df <-bind_rows(result)
# Write your output csv file
write.csv(df, 'yourfilename.csv')
另外,您可能想看看tabulizer的不同解析方法(我在这里将其设置为'stream',因为根据我的经验,这通常会产生最好的结果,但是也许'lattice'对某些表更有效)。
