Eureka Man » AI

Natural Language Search Engine Optimization

Eureka Man — Sun, 28 Oct 2007 01:51:43 +0000

This is a tip for search engine optimizers of the future, a future in which Powerset, or a similar system, grabs a major piece of the search market. Powerset has language understanding capabilities well below those of a human. In particular it deals much better with sentences constructed in what might be referred to as some form of “Standard English”. This is not a peculiarity of Powerset. If you ever find a NL system that understands humans as well as they understand each other then you’ve found genuine AGI and you’ll be doing more interesting things with it than making search engines. I’m sure the guys at Powerset are thinking about ways to parse sentences that don’t quite match their grammar. Nevertheless, it will always remain true that natural language search engines do not read the web as well as humans do.

Given this tip, you can guess what search engine optimizers will do. If you want to increase your chances of being indexed right by Powerset, you’d better work out what grammar it likes best and conform to it. At last, prescriptive linguistics will find its true purpose. Imagine the written comment of a future English teacher, amid the red marks, “Incorrect grammar. Will not be well indexed”.

Is there an uncanny valley for natural language search?

Eureka Man — Mon, 09 Apr 2007 03:11:31 +0000

There is a hypothesis pertaining to robotics and computer generated, human-like characters called the Uncanny Valley. It states that as robots become more and more human-like, the comfort of humans interacting with them takes a sharp dip before it rises again towards fully human interaction.

In Fernando Pereira’s latest take on Powerset, he eludes to a similar nonmonotonic function for natural language search. Conversing efficiently in natural language requires that you can predict, to some degree of accuracy, what will be understood by your conversation partner. Systems that get uncannily close to human understanding yet fail in cases that are only predictable to the system designers may be frustrating to use. How will the tradeoff between increasing utility and increasing inscrutability pan out? Where is Powerset going to be on this hypothetical graph?

As Fernando says, only testing on the final system will tell.

Artificial Artificial Intelligence at ChaCha

Eureka Man — Fri, 29 Sep 2006 07:48:37 +0000

ChaCha is a new search engine that will connect you to a personal search assistant (a human!) to help you with your query. Matthias decided to test it out with a challenging query:

Status:

Looking for a guide …

Status:

Connected to guide: SomeGuide

SomeGuide:

Hi there. I will be helping with your search.

SomeGuide:

Hey, there. How are you today?

SomeGuide:

So you’re looking for the ‘answer to everything’?

You:

i am good

You:

and you?

You:

yes

SomeGuide:

I’m great. Thanks for asking

You:

SomeGuide, I am really curious what might be a good answer for everything

SomeGuide:

You mean in conversation?

SomeGuide:

Or ‘everything’ in an all-encompassing philosophical scope.

You:

yes, philosophical

You:

It sounds like a hard task, but I am looking for that to put into a slideshow for my next presentation

You:

as an opening quote

You:

it will be a technical talk about research opportunities

You:

for the future

SomeGuide:

Hmm, that’s quite a question…

…….

Go read the rest at Matthias’ excuse for a permalink. It has a happy ending!