Re: Why is NLP such a hard topic? - Broken rules.

Rod Davison wrote :

> Question:  Why is NLP a hard topic?
> Answer: Because natural language is a a rules based system
> where users don't follow the rules.
>

Yes, I think that we more-or-less use rules to understand English,
but we can still understand sentences that are full of errors.


> Listen to conversations, speakers do
> not speak in sentences or well formed linguistic units:

Thats true.

Just to illustrate the truth of this,
I'll show you some of the sentences in that last email
which are hard for a computer to understand.

These sentences all have typing errors or grammatical errors.
They are still quite easy for a human to understand.

Here are the difficult sentences now...

> The answer in simple in statement but incredibly complex in its
> implications.

The first word "in" was meant to be "is".

> For example, many years ago, I lead a project that developed a machine
> translation program form English to French.

"Form"?  That was meant to be "from".

> The question was asked how
> accurate was the translation -- 80%, 90%.

What is the grammatical structure of that sentence?

How are the words "80%, 90%" relevant to the rest of the sentence?


> I suggested that the way to
> answer this was to have two human translators translate the text first
> then one they agreed on a benchmark translation, rate our programs
> output.

The words "one they agreed" should be "when they agreed".

> No two translators ever agreed on a right translation for the
> benchmark.  Several times we they even came to blows.
>

The words "we they" was intended to be "they".

Also, does it mean "they came to blows" literally ?
I.e. does it mean they really hit each other?
It takes a lot of real world lknowledge to answer this question.
In fact, I'm not even sure of the answer myself.

> > I have long been wandering about that question,
> > and did not really find any good answers:

"Wandering"?  That was meant to be "wondering".


> > There are many (profitalbe) uses for natural language processing ...

"Profitalbe" ?
I can understand you, but how could a computer program
understand this?

Also, the brackets around "profitable" should not be there.
These brackets give that sentence a unique grammatical
structure, that can't be found in a normal grammar tree.

--
Martin Sondergaard,
London.