Wednesday, March 27, 2019


--- How complicated is language? Try speech synthesis ---

We take language for granted until we try to learn a new one. I don't think anyone would say that Spanish is more difficult than English, in fact, there are a lot of similarities. There's a lot more that you have to match up between nouns, verbs, modifiers and other word forms, and modifiers tend to follow the words they modify, unlike in English ("Little Red Riding Hood" instead of "Hood Riding Little Red") but the differences aren't huge.

I had little trouble in my 20s learning German and American Sign Language, but Spanish, I find unaccountably hard. But I'm having fun anyway.

Consider, though, the problem of computer language recognition and speech synthesis. They've come very far in the last 20 years. This century has seen some huge breakthroughs. In the 1990s, computer speech was still pretty much science fiction. What there was, was pretty clunky. But now, I routinely let my computer read books to me, a God-send given my dyslexia.

My computer still has a few problems though. Consider the word "read". Should it be pronounced "reed" or "red". It depends on it's context. Words exist in context. In fact, there are four things that have to be considered when you are parsing out words. It's amazing how easily a human does all that without breaking a sweat.

The meanings of any particular words are usually many. Just choose any word you thought you knew well and look it up in the unabridged Oxford Dictionary of the English Language. Then words exist in relation to the words around them. How you pronounce "read" has a lot to do whether you read a book yesterday or where you will read it today - tense determines the pronunciation. But "read" can be an adjective since a book can be a "read book" (or one that you have already read and, so, can mark off your reading list). Where words appear in a work also has meaning. One reason "It's was the best of time. It was the worst of times." is so well known is that it was the fist statement in Dickens "A Tale of Two Cities." Everything else being equal, first and last are most likely to make an impression. And other things can strongly influence how words are understood and pronounced. Intonation is one. Read "You are a fine one," out loud several times,placing emphasis on different words and see how the meaning changes.

I use a screen reader called NaturalReader, produced by NaturalSoft Ltd. It does a great job but, when I first started using it, I noticed a humorous snag. Frankly, I don't know if it was in the screen reader, the Windows dictionary or what but, when it came to the word "Nazi", it said "ne'er-do-well". "Appropriate" maybe, but not accurate. Today, my screenreader just boringly announces "Nazi". The older way was more entertaining.


No comments: