quanteda kwic regex operation

问题

Further edit to original question.
Question originated by expectation that regexes would work identically or nearly to "grep" or to some programming language. This below is what I expected and the fact that it did not happen generated my question (using cygwin):

echo "regex unusual operation will deport into a different" > out.txt
grep "will * dep" out.txt
"regex unusual operation will deport into a different"

Originary question
Trying to follow https://github.com/kbenoit/ITAUR/blob/master/README.md
to learn Quanteda after seeing that everybody that uses this package finds it very good.
In demo.R, line 22 I find the line:

kwic(immigCorpus, "deport", window = 3)

Its output is -

[BNP, 157]        The BNP will | deport | all foreigners convicted  
[BNP, 1946]                . 2. | Deport | all illegal immigrants    
[BNP, 1952] immigrants We shall | deport | all illegal immigrants  
[BNP, 2585]  Criminals We shall | deport | all criminal entrants

To try/learn the basics I execute

kwic(immigCorpus, "will *depo", window = 3, valuetype = "regex")

expecting to get

[BNP, 157]        The BNP will | deport | all foreigners convicted

but I get:

kwic object with 0 rows

Similar attempts like

kwic(immigCorpus, ".*will *depo.*", window = 3, valuetype = "regex")

Get the same result:

kwic object with 0 rows

Why is that? Tokenization? if so how should I write the regex?

PS Thanks for this great package

回答1:

You are trying to match a phrase with your pattern. By default, the pattern argument is treated as a space separated list of keywords, and the search is performed against this list. So, you may get your expected result using

> kwic(immigCorpus, phrase("will deport"), window = 3)
[BNP, 156:157] - The BNP | will deport | all foreigners convicted

A valuetype = "regex" makes sense if you are using a regex. E.g. to get both shall and will deport use

> kwic(immigCorpus, phrase("(will|shall) deport"), window = 3, valuetype = "regex")

   [BNP, 156:157]             - The BNP | will deport  | all foreigners convicted
 [BNP, 1951:1952] illegal immigrants We | shall deport | all illegal immigrants  
 [BNP, 2584:2585]  Foreign Criminals We | shall deport | all criminal entrants

See this kwic documentation.

回答2:

The examples from the ITAUR repository are based on an older syntax. What you need is the phrase() wrapper - see ?phrase. You should also probably brush up on the regular expression syntax you are trying to achieve with the *, since it may not be what you want, and since a regular expression cannot start with a "*". (This might help.) The default "glob" valuetype will probably achieve what you want.

library("quanteda")
## Package version: 1.1.4
## Parallel computing: 2 of 8 threads used.
## See https://quanteda.io for tutorials and examples.

kwic(data_char_ukimmig2010, phrase("will deport"))

## [BNP, 156:157] nation.- The BNP | will deport | all foreigners convicted of crimes

kwic(data_char_ukimmig2010, phrase("will .*deport.*"), valuetype = "regex")

## [BNP, 156:157] nation.- The BNP | will deport | all foreigners convicted of crimes

来源：https://stackoverflow.com/questions/49478723/quanteda-kwic-regex-operation

标签

regex

nlp

quanteda