Sign in to follow this  

English Speaking Computers

This topic is 3574 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I was just wondering, what sort of issues are there in making a computer actually understand English. I've been doing a lot of thinking on this. I already know a few of the issues but I would really want to know what kind of hurdles that need to be overcome before this becomes a reality.

Share this post


Link to post
Share on other sites
Apparently, there exist algorithms that have 99%+ recognition rate. Downside is that they need to run on clusters, and take a week or so to process the not exactly large amounts of data.

Quote:
but I would really want to know what kind of hurdles that need to be overcome before this becomes a reality


Join a research team the works in the area. To produce anything useful, the domain isn't accessible to common user.

Or better yet, license one of existing APIs.

Share this post


Link to post
Share on other sites
English and most other human languages are contextual and open to interpretation.

Consider the following two sentences:
Quote:

I wanted to go to the shop but it started raining.

I wanted to go to the shop but it was closed.


How does a program know what "it" means? Obviously to you and I, "it" refers to the weather in the first sentence and the shop in the second sentence. We can infer this because shops don't rain and weather doesn't close. So in order to correctly interpret either of those sentences a machine needs to understand what a shop is and if it can rain or close.

It gets even worse. Consider
Quote:
She drove me to it.


Depending on a completely external context to this sentence, it could mean "she transported me to something (it) in a car" or "her actions prompted me to something".

Share this post


Link to post
Share on other sites
Another fun example that, I think, comes from "What computers still can't do":

Scene: girl standing in front of a pet store window with puppies on display

Girl: "Mommy! I want it!!!"

To what does 'it' refer:
a) the puppy
b) the window
c) the box holding the puppies
d) the air at the tip of her fingers
e) the shop in it's entirety
f) the mound of puppy poop
g) the puppy food
h) etc

There is so much "common sense" reasoning bundled up in that simple example it'd take ages to develop algorithms that could reliably predict the answer to other similar scenarios.

I'd be interested to learn about how/if the algorithms to which Antheus refers solve this problem. Do they require vast databases of "common sense" rules; do they require that all pronouns and whatnot have direct references contained within the parsed language block, etc.

-me

Share this post


Link to post
Share on other sites
IMHO, it is extremely unlikely that we will ever be able to sit down and write a formal set of rules and definitions (i.e. a traditional program) that will understand human speech. The best results will likely come from evolutionary algorithms and machine learning. I'll happily admit I'm largely ignorant of these fields, but it feels like a more realistic approach to me.

Share this post


Link to post
Share on other sites

This topic is 3574 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this