Keyword question

Come here for help or to post comments on Sphider
Post Reply
nhaas
Posts: 8
Joined: Tue Jul 12, 2022 12:31 am
Contact:

Keyword question

Post by nhaas »

I have a question looking through the Code at Keywords. I have a basic understand of the Keyword feature. But the weight is perplexing me. Is the weighting score basically generated from the fact that the Keyword is in the Title or Body? Or are there other factors, that generate the weight?

Thanks
User avatar
captquirk
Site Admin
Posts: 299
Joined: Sun Apr 09, 2017 8:49 pm
Location: Arizona, USA
Contact:

Re: Keyword question

Post by captquirk »

Key factors in weighting is location (title, domain, path, meta tags, and body) and number of occurrences.

The keyword weighting remains just as it was when Ando Saabas left it. I have not had the courage to screw with it, although I can't say I always feel it is ideal. For example, one particular case of my own, the first of 33 results is a relative 100% (which I accept, but the SECOND result is 15%!!! The final 26 results all hover around 5%. The skewing seems to be a bit off. But then again, IF I were to look at each occurrence in detail it might make sense. (??)
nhaas
Posts: 8
Joined: Tue Jul 12, 2022 12:31 am
Contact:

Re: Keyword question

Post by nhaas »

wow, seems intense. Now that I look at Keywords more, I am assuming that the saving to the Keywords table(s) is just to speed up the query of the searches and for the shear number of keywords. I have not even looked into the search side of the house.
User avatar
captquirk
Site Admin
Posts: 299
Joined: Sun Apr 09, 2017 8:49 pm
Location: Arizona, USA
Contact:

Re: Keyword question

Post by captquirk »

Compared to indexing, searching is a piece of cake.
Any particular keyword can occur on any particular page. And if multiple sites are indexed, keywords can appear in multiple sites. Yes, there are many keywords. Now imagine the number of link (page) to keyword relationships. That is why there are 16 such relationship tables.

When you do a keyword search, the keyword is found in the keyword table. It has an indexing number. Then all 16 link-keyword tables are searched for the index number. The result includes the page and a weight for each page. These results and the highest weight is given a relative weighting of 100%. A weight of 0 (which will not actually occur in the results) is 0%. Everything else has a RELATIVE percentage. On the one hand, I think the relative weighting system could be improved, but on the other hand the way it is now makes darn good sense. I have yet to have a brainstorm telling me how to improve it! So I leave it alone.

Phrase searches are much simpler. The search just looks at the full text of a page in the links table and returns matches. I have never delved into the ranking system for phrase searches.
nhaas
Posts: 8
Joined: Tue Jul 12, 2022 12:31 am
Contact:

Re: Keyword question

Post by nhaas »

The only way I could see improving it is having a Keyword date. So if a keyword searched has date that is closer to todays day it would get a 10% boost in weight, or something like that. Thanks for the explanation. that helps me a lot.
Post Reply