Spinning Approach [Do u know ?]

[Veron1ka]

Anyone knows what approach software like TBS, SpinnerChief and others use to generate 600 or so optimized articles ?

 

[Tokarev]

Synonyms maybe?

 

[Scritty]

You have to tell them the keyword, they then manage the density and add associated words and phrases.
The real key is the associated words and phrases. These are VITAL for context and the SEs look for them.

Scritty

 

[Veron1ka]

I think my question wasn't clear enough
I was given the task to code an application that generates articles from a spintax input, I am looking for away NOT to get a random generation but rather a calculated one so I don't kill the potential of the spintax content

 

[bluehatface]

What language? Web or desktop? Calculated in what way? Keyword/LSI density?

 

[Veron1ka]

it is in C# but the language is not important as I need to know what is the methodology to be used to generate different articles using different paths each time an article is generated.
I guess it is a math formula that considers the Spin levels and the number of alternatives inside each level for a given spintax...

 

[deki33]

You need to calculate number of variations for each block (sentence if it is spanned on sentence level).
Put each block in array and do all variations.
Array1 {a,b,c}
Array2 {d,e,f}

So, you will get
a d
a e
a f

b d
b e
b f

etc....

 

[bluehatface]

You want to make sure each article is different from another rather than pick at random? You could run through each nest as an array to make sure you're picking one not used before. For a full article, if you store the previous path(s)/route(s) as an array of integers (or possibly an array of arrays), (choosing the path randomly), you can then use an if statement to pick an alternate route the next time(s).

 

[Veron1ka]

@deki33 : this won't help when the nestspin is 6 levels and very huge, also not interested in getting all the variations
@bluehatface : u are suggesting a comparision approach on the fly, this can be very resources consuming and may end up crashing the server or the application, I need to adopt a formula that extracts article without doing any comparison between the article so it is a formula that is going to be adopted by the application without caring about how different are the articles, the comparison can be done later on a different step

 

[bluehatface]

OK so just a simple spin then? http://www.blackhatworld.com/blackhat-seo/black-hat-seo/340392-your-own-article-spinner-php.html

 

[Veron1ka]

OK so just a simple spin then?

Not so simple as I try to avoid randomness, U know, if u extract 4000 articles from spintax, u can end up with 100 exactly similar articles and also u did not full utilize the potential of your spintax. it is a matter of probability.

I want to break down the probability factor and make it based on a calculation which should be there before processing the generation

 

[FutureProofSeo]

Not so simple as I try to avoid randomness, U know, if u extract 4000 articles from spintax, u can end up with 100 exactly similar articles and also u did not full utilize the potential of your spintax. it is a matter of probability.

I want to break down the probability factor and make it based on a calculation which should be there before processing the generation

How about you calculate all of the possibilities and then let it review final articles for certain densities, and keep the most optimised variations?

 

[bluehatface]

Not so simple as I try to avoid randomness, U know, if u extract 4000 articles from spintax, u can end up with 100 exactly similar articles and also u did not full utilize the potential of your spintax. it is a matter of probability.

I want to break down the probability factor and make it based on a calculation which should be there before processing the generation

You could iterate through, and save the number of possible variations of each nest and layer similar to deki33's suggestion into an array (of arrays), to create a navigable tree of values. You can then create a set of predefined navigation routes, to extract the spin based on non-reoccurring patterns. You can compare previously created routes to try not to re-use previously used synonyms. You will then have an array of 'routes' which you can plug the spintax through, giving unique, methodically spun content.

 

[Veron1ka]

@FutureProofSeo
Can u elaborate ? May be I got you wrong
I don't think the density is a good way to go as we can have for example 2 articles using exactly the same words which are organized in the articles in a different
manner, I wouldn't consider these articles as duplicate

 

[Veron1ka]

You could iterate through, and save the number of possible variations of each nest and layer similar to deki33's suggestion into an array (of arrays), to create a navigable tree of values. You can then create a set of predefined navigation routes, to extract the spin based on non-reoccurring patterns. You can compare previously created routes to try not to re-use previously used synonyms. You will then have an array of 'routes' which you can plug the spintax through, giving unique, methodically spun content.

With the complexity of levels deep we may have in the spintax, I prefer to avoid to do any calculation or comparison during this step.
The comparison can be done later when the articles are generated.