I agree with Faldage's analysis of 'eoxlibnsdaienvlab.' Eoxlibnsdaienvlab is a random clustering of letters that may appear to be word-like, possibly even a word from another language, but unlikely so unless I was a very lucky monkey at the keyboard.
Symbols can represent meaningful utterances we readily understand, and it could be argued that letters, too, are symbols representing meaningful utterances. However, I think that symbols are representations of words and not actually words themselves. I would restrict words to being:
meaningful utterances
alphabetical representations of those meaningful utterances
...but only when examining the written language in a search for words. I can write the code: 3-1-19-20 to represent the word cast, but I would not argue that
3-1-19-20 is a word. I would say 3-1-19-20 is a symbolic representation of the word cast.
The entire situation dramatically changes when we examine other languages, such as sign languages that we've discussed here before.
So, in my own consideration of words and language without including sign languages, I would not include symbols, such as %, to be called and recognized as words in and of themselves, but representations of words, such as the word percent. I'm fairly sure that I'll be in the minority here. But so it goes...
