Home
Name Modified Size InfoDownloads / Week
README.txt 2012-11-30 1.5 kB
Totals: 1 Item   1.5 kB 0
The project is under construction!

How does this work, and what does it do?

My earlier project consisted of comparing the strings (sentences) based on their semantic value:
Calculate Semantic Similarity - https://sourceforge.net/projects/semantics/

It has come to my attention that such a comparison is 1-dimensional comparison since output is one single value showing a semantic distance between the sentences.
Now, the idea is to take a range of words according to their distances from the main word, and to split the range into N units. Each unit represents a dimension, so that first unit is N=1, second N=2 (since the words that are being matched to a main word are sorted) and so on...

Then, for each N, the group of words are collected and combined into N number of sentences.
Each sentence obtained by combining is then compared to an original sentence by the vector-space analysis.
The result is therefore the N-dimensional array of doubles, or an N-dimensional representation of a sentence.

What does this mean?
It means that we can represent a sentence as a point. Furthermore, we can also assign the N-dimensional vectors to a point to show how the sentence relates to a context. This itself is a mathematical representation of the "Meaning" of the sentences, which can be used to calculate any Artificial Intelligence tasks, such as clustering, classifying, and so on. 

For the further study and research:
To find a valid way of going backwards, that is, from a N-dimensional mathematical representation, to a sentence!


Source: README.txt, updated 2012-11-30