sentiment is a command-line utility that
evaluates the emotional sentiment of natural language text.
$ echo "With a comfortable fit, great sound, and awesome noise canceling, these are the best AirPods ever." | sentiment
1.0
$ echo "No wireless, less space than a Nomad. Lame." | sentiment
-0.8
For more information about natural language processing, check out Chapter 7 of the Flight School Guide to Swift Strings.
- macOS 10.15+
Install sentiment with Homebrew using the following command:
$ brew install flight-school/formulae/sentiment
Text can be read from either standard input or file arguments, and named entities are written to standard output on separate lines.
$ echo "It was a good day." | sentiment
0.4
$ sentiment
Great!
1.0
$ head -n 1 anna_karenina.txt
Happy families are all alike; every unhappy family is unhappy in its own way.
$ sentiment anna_karenina.txt
-0.400151
Note:
sentimentcan take a long time for large documents.
$ time sentiment anna-karenina.txt
-0.400151
33.44 real 61.84 user 4.62 sys
sentiment can be used with ls, find and other Unix commands
to analyze the sentiment of multiple files at once.
$ find . -iname '*.txt' -exec \
sh -c 'printf "%s\t%s\n" {} $(sentiment {})' \;
./positive.txt 1.0
./somewhat-negative.txt -0.5
./neutral.txt 0.0
If the command is able to determine an emotional sentiment for the provided text,
it writes a number between -1.0 (negative) and 1.0 (positive)
to standard output (0.0 indicates neutral or unknown sentiment).
The resulting number has an unlocalized format
consisting of an optional leading minus sign - (U+002D HYPHEN-MINUS)
followed by a decimal point . (U+002E FULL STOP)
and between one and six decimal digits 0 – 9 (U+0030 – U+0039):
^-?[01]\.\d{1,6}$sentiment uses
NLTagger
with the
sentimentScore
tag scheme.
The overall sentiment of a text input is calculated from
the average sentiment of each paragraph.
MIT
Mattt (@mattt)