quanteda.dictionaries

This is a fork of the original quanteda.dictionaries maintained by kbenoit. Please go to https://github.com/kbenoit/quanteda.dictionaries for the original package.

An R package consisting of dictionaries for text analysis and associated utilities. Designed to be used with quanteda but can be used more generally with any text analytic package (e.g. tidytext, tm, etc.).

Experimental Regex support

This fork enables regular regex-based patterns in liwcalike by implementing a complementary function, codenamed liwcahead, which allows the following regex patterns to be used:

• ., e.g. .ome for {some, home, ...}
• *, e.g. fil.* for {fil, file, filing, filet, ...}
• ?, e.g. s.?lice for {splice, slice, ...}

It accomplishes this by using the optional arg regex in tokens_lookup:

tokens_lookup(toks, dictionary, valuetype="regex")

DISCLAIMER

Note that the entire regex pattern set is not guaranteed to work. For example, the brackets characters {(, )} are not functional, as liwcahead's dependencies remove the brackets.

Installing

# the devtools package needs to be installed for this to work
devtools::install_github("lullabysinger/quanteda.dictionaries") 

Usage

This is a drop-in replacement for the original liwcalike -- nothing has changed for liwcalike, which forks the original code.

However, the liwcahead function is the regex-enabled version. Arguments remain the same as the original kbenoit liwcalike, with one small difference.

If the LIWC-style dic file consists of wildcards, use the following:

output_lsd <- liwcahead(corpus, dictionary, regex = TRUE)

Demonstration - as adapted from the original kbenoit/quanteda.dictionaries

With the liwcalike() function from the quanteda.dictionaries package, you can easily analyze text corpora using exising or custom dictionaries. Here we show how to apply the Lexicoder Sentiment Dictionary (Young and Soroka 2012) to a corpus consting of 2000 movie reviews (from the quanteda.corpora package).

library("quanteda.dictionaries")

output_lsd <- liwcalike(quanteda.corpora::data_corpus_movies, 
                        dictionary = data_dictionary_NRC)

head(output_lsd)

##           docname Segment  WC       WPS Sixltr   Dic anger anticipation
## 1 neg_cv000_29416       1 847  78.88889  13.11 19.36  0.71         2.95
## 2 neg_cv001_19502       2 278 240.00000  10.79 24.46  3.24         3.24
## 3 neg_cv002_17424       3 559 162.66667  16.46 21.29  1.07         2.86
## 4 neg_cv003_12683       4 594 239.50000  16.67 22.73  1.35         3.03
## 5 neg_cv004_12641       5 872 366.50000  19.04 19.38  1.26         1.49
## 6 neg_cv005_29357       6 753 671.00000  18.33 27.49  3.32         1.59
##   disgust fear  joy negative positive sadness surprise trust AllPunc
## 1    0.83 1.42 2.36     2.24     4.01    1.42     1.30  2.13   18.06
## 2    2.16 1.80 1.44     4.32     2.88    1.44     1.08  2.88   18.35
## 3    0.72 2.68 1.43     3.40     4.11    1.79     0.54  2.68   14.67
## 4    1.18 1.85 1.52     4.21     5.05    1.85     1.18  1.52   22.90
## 5    0.69 2.06 1.49     3.90     3.33    1.72     1.38  2.06   17.66
## 6    2.39 4.25 0.80     5.98     3.45    1.99     1.86  1.86   11.02
##   Period Comma Colon SemiC QMark Exclam Dash Quote Apostro Parenth OtherP
## 1   4.01  5.19  0.35  0.00  0.71   0.35 1.18  3.07    1.89       0  14.76
## 2   5.04  6.47  0.00  0.00  0.00   0.00 0.00  5.40    4.68       0  16.91
## 3   3.94  5.55  0.00  0.00  0.54   0.00 0.54  2.68    2.68       0  12.70
## 4   3.37  4.38  0.00  0.00  0.51   0.00 4.71  7.58    4.21       0  15.82
## 5   4.24  6.19  0.69  0.23  0.11   0.00 1.95  2.18    1.72       0  13.65
## 6   4.65  3.98  0.13  0.00  0.00   0.00 0.53  0.93    0.40       0   9.69

Code of Conduct

Please note that this project is released with a Contributor Code of Conduct, which we adhere to, based on the original from kbenoit/quanteda.dictionaries.

By participating in this project you agree to abide by its terms.