Linguistic Computing with UNIX Tools - Natural Language Processing and Text Mining - page 223

program lowerCaseStrings of section 12.2.1 for a non-trivial example using this
technique.
We observe that one can also store sed commands without the two framing
single quotes in a file (say) sedCommands and use sed -f sedCommands instead of or
in any sh program as above.
Applications (tagging programs for grammatical analysis): The above example
indicates how to examine a text for the overall structure of grammatical patterns
occurring in that text. Thereby, lists of verbs, nouns and other grammatical entities
can be automatically transformed into sed programs that perform identification (see,
e.g. , the listing of the UNIX command eliminateList in Section 12.3.4). Similarly,
a dedicated list of words can be automatically formatted into a (search) program in
another programming language.
Applications (synonyms/translations through substitution): A custom-de- signed
sed program similar to the above can be used to replace every word in a file by a
bracketed list of synonyms. This can be used as a feedback device to encourage stu-
dents to use a diversified vocabulary. In addition, note that sed can handle Japanese
kanji characters. Thus, a custom-designed sed program similar to the above where
every English word is replaced by a bracketed family of romaji, hiragana/katakana
and kanji characters can be used to assist translation of text documents.

Application (cleaning files of control sequences in preprocessing): The replace-
ment in a substitution can be empty. This can, e.g. , be used to “clean” a tex / latex
[33] file of control sequences ( e.g. , \subsection ).
The Format of an Addressed sed Command
The format of an addressed sed command is Address Command Address can
be omitted. In that case, Command is then applied to every pattern space. (On first
reading, one may think of the pattern space as being the current line of input). If
an Address is given, then Command is applied to the pattern space only if the pattern
space matches Address . Address can be a pattern (regular expression) enclosed within
slashes / as described in Appendix A.1, a line number not enclosed within slashes
or $ standing for the last line. In addition, a range of addresses in the format
StartActionAddress , EndActionAddress Command can be used. In that case, Command
is applied to every pattern space after StartActionAddress has been found until and
including the pattern space where EndActionAddress has been found.
Example: The following program replaces TX with Texas in all lines that contain
the string USA .

#!/bin/sh
sed '/USA/s/TX/Texas/g' $1

sed commands are terminated by either an immediately following newline character,
a semicolon, or the end of the program.
One may wish to process the single quote ' using sed or awk . In a program
similar to the one listed above, the single quote ' needs to be encoded as: '\'' .
This means for sh : (1) terminate the string listing the sed program temporarily at
the first ' , (2) concatenate the latter with the literal ( \ ) character ' and (3) continue
the string listing the sed program by concatenating with the string following the
third ' .Ifa sed or awk program is stored in a file, then the the single quote ' is