Text pairs give “right sequence of text” intuition to strong AI. With text pairs strong AI can feel which text is grammatically correct text and which is incorrect grammatically.
Text pair is pair of text concepts. Several text pairs define text order within a clause.
(Text concepts could be words or phrases).
Example:
Let’s consider a clause: “My name is Sid”.
The clause can be
easily separated into words: “My”, “name”, “is”, and “Sid”.
Text pairs would be:
“My” -> “name”
“name” -> “is”
“is” -> “Sid”
Phrases can be found
within the same text: “My name is”, “name is”, “is Sid”.
So, text pairs would be also:
“My” -> “name is”
“My name is” ->
“Sid”
“name is” -> “Sid”
Text pairs not only define the order within a pair of text concepts. Text pairs also define order within a whole clause. It’s possible because a set of text pairs creates a “chain” of pairs. The chain helps to order the whole clause.
Text pairs don’t serve for keeping one particular sentence in order. Instead text pairs are used to collect and keep statistical information about what is usual order of text concepts within any text.
While strong AI reads a text, text pair statistical information is stored into text pair table.
Example:
Let’s consider the
same clause again: “My name is Sid”.
We have words:
|
Word itself |
Word name |
Start position |
Next position |
|
My |
Word1 |
1 |
2 |
|
Name |
Word2 |
2 |
3 |
|
Is |
Word3 |
3 |
4 |
|
Sid |
Word4 |
4 |
5 |
We have phrases:
|
Phrase itself |
Phrase name |
Start position |
Next position |
|
“My name is” |
Phrase1 |
1 |
4 |
|
“name is” |
Phrase2 |
2 |
4 |
|
“is Sid” |
Phrase3 |
3 |
5 |
Now we can create “temporary text concept table”:
|
Text concept name |
Start position |
Next position |
|
Word1 |
1 |
2 |
|
Word2 |
2 |
3 |
|
Word3 |
3 |
4 |
|
Word4 |
4 |
5 |
|
Phrase1 |
1 |
4 |
|
Phrase2 |
2 |
4 |
|
Phrase3 |
3 |
5 |
This simple algorithm shows how to find text pairs within “temporary text concept table”: