Dear Nadia,
Unitex is appropriate for tagging corpus with lexical resources not for string processing. The preprocessing with "Replace" is not appropriate as well to the task you describe. A clean solution is to use find/replace with regular expression (Regex), for instance by using Unix utilities such as grep/sed.
You describe the inflexion of job names for Spanish, which is not a simple. It will be more simple to use a scripting language such as Python or PERL, using the embedded regex library. This will give you more readability, flexibility and maintainability than the Unix utilities (grep/sed) .
I don't know if you wanna go in such direction, take a look to my Python script (attachment) to have an idea of what you should expect by using such Scripting language (see attachment) and in term of developer time-consuming. the script is clear and simple ( for a Python programmer, you can read). Such script can preprocess many files or whatever!
if you have further questions ....
Cheers,
Alex
================================
Here the output of my Python script for the 4 examples you gave,
of course, it works for similar Jobs like "Doutor/a; Engenheiro/a electrico" :
"<<<" read as input line string
">>>" read as output line string
wc : word count
------OUTPUT -------------------------------
<<< line, wc: vendedor/a 1
>>> vendedor
>>> vendedor a
------------------
<<< line, wc: instalador/a electrico 2
>>> instalador electrico
>>> instalador a electrica
------------------
<<< line, wc: encargado/a de administracion 3
>>> encargado de administracion
>>> encargad a de administracion
------------------
<<< line, wc: peluqueros/as 1
>>> peluqueros
>>> peluquer as