1. Advertising
    y u no do it?

    Advertising (learn more)

    Advertise virtually anything here, with CPM banner ads, CPM email ads and CPC contextual links. You can target relevant areas of the site and show ads based on geographical location of the user if you wish.

    Starts at just $1 per CPM or $0.10 per CPC.

New way to evaluate keywords and html

Discussion in 'Keywords' started by jocs, Jul 28, 2004.

0
  1. #1
    I wan to expose to this comunity my idea and when we all think its the best way then make a new tool:

    When counting keyword density and prominence we have many ways to evaluate the keywords but I think we can have a tool that joins all ways and gives a concrete total view of each keyword.


    note: when i say word I mean words that are not stopwords.

    We'll view it by this Example:
    <html><body> tal tal tal tal tal tal tal tal tal tal <b> yam tam pam </b></body></html>

    keyword "yam": (values are not real)
    it is 0 non stop words far from <b> tag - 100% prominence (it is the first word)​
    it is 10 non stop words far from <body> - tag 2% prominence​
    it is 10 non stop words far from <html> - tag 2%p rominence​

    Ok now we have a tag value table: (values are not real)

    <HTML> 100
    <BODY> 80
    <B> 30

    note: if there are 5 <b> tags in the html page, <b> tag value is divided by 5, the meaning of this is that if there is only one <b> tag in the html it will be more important than if there are 5.


    then we'll calculate the total Rank of keyword yam doing:
    VALUE += (% of prominence)*(tag value) for each tag that affects the keyword.

    Value of "yam"=100%*30 (<B> tag value) + 2%*80 (<body> tag value) + 2%*100(<html> tag value)
    Value of "yam"= 30+1.6+2=33.6

    ...
    ...
    ...


    when we have all words values calculated we apply this rules:

    if a keyword is just before another it will gain 1% of other value
    if a keyword is just After another it will gain 0.8% of other value
    if a keyword is the second word before another it will gain 0.5% of other value
    if a keyword is the second word after another it will gain 0.4% of other value

    meaning if a word is next to another, if one is important the other also will be a little important.

    here is my idea. What do you think of?

    Hope you understan dit :confused: , at least the idea, I would appreciate all comments, thanx.
     
    jocs, Jul 28, 2004 IP
    hdpt00 likes this.
  2. rocky

    rocky Guest

    Messages:
    23
    Likes Received:
    1
    Best Answers:
    0
    Trophy Points:
    0
    #2
    Very interesting. How about building a search engine?
     
    rocky, Jul 28, 2004 IP
  3. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #3
    Maybe we can do it together!!! anyone else ???

    Did you at least understand my Idea ?
     
    jocs, Jul 28, 2004 IP
  4. steve5jack

    steve5jack Peon

    Messages:
    28
    Likes Received:
    0
    Best Answers:
    0
    Trophy Points:
    0
    #4
    Proximity to <body> tag is interesting. That makes sense.

    Why count just <b>? Why not <em> and the header <h1> tags as well?
     
    steve5jack, Jul 28, 2004 IP
  5. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #5
    This was only a example, it will do it with all html tags.
     
    jocs, Jul 28, 2004 IP
  6. hdpt00

    hdpt00 Peon

    Messages:
    427
    Likes Received:
    15
    Best Answers:
    0
    Trophy Points:
    0
    #6
    Sounds like a good idea, obviously, like google's page rank, you would have to apply some sort of iterative count to calculate the bottom part (ex: 1% gain from word before, 0.8% from word in front). So finding the right number of iterations would be key.
     
    hdpt00, Jul 28, 2004 IP
  7. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #7
    Maybe I'm wrong, but I was thinking to do only one iteration, precalculating how much value will gain each keyword and when all calculated add the initial value and the gain.
     
    jocs, Jul 28, 2004 IP
  8. mopacfan

    mopacfan Peon

    Messages:
    3,273
    Likes Received:
    164
    Best Answers:
    0
    Trophy Points:
    0
    #8
    That's definitely over my head...

    "The plane, the plane"
     
    mopacfan, Jul 28, 2004 IP
  9. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #9
    Maybe I can explain it in another way, just tell me which part you didn't understand.


    To everyone who read this: Please POST AND TELL ME WHAT YOU THINK on my idea, its really important to me.
     
    jocs, Jul 28, 2004 IP
  10. hdpt00

    hdpt00 Peon

    Messages:
    427
    Likes Received:
    15
    Best Answers:
    0
    Trophy Points:
    0
    #10
    I guess one iteration will work, but then it won't take into account this scenario lets say:
    keyword1 keyword2 keyword3

    using that string... keyword1 will not benefit from the additional percentage keyword2 gains via keyword3, hence for more accuracy a few iterations might be beneficial. Of course that is up to you though.
     
    hdpt00, Jul 28, 2004 IP
  11. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #11
    Maybe the best way to count how many times i've to iterate will be:
    (I apologise about always using rule1, rule2, rule3...)
    - Stop iterating if you passed the limit of N1 iterations
    - Stop iterating if you find a word with less than N2 value
    - Stop iterating if you are in the limit of a limit_Tag (maybe not all tags will be limit_tags, example <body> will be a limit_tag ...we'll need a list of limit_tags)


    any idea?
     
    jocs, Jul 28, 2004 IP
  12. hdpt00

    hdpt00 Peon

    Messages:
    427
    Likes Received:
    15
    Best Answers:
    0
    Trophy Points:
    0
    #12
    Maybe a better idea would just be run lets say 10-20 iterations, each subsequent iteration the value of keywords near each other will go up, you can then use that scaled value everytime or set up rules to get some type of weighted average. But I think it should be a finite number of iterations each time that us global for every site.

    Do you plan on running this as a cgi script or sometihng or a robot. I mean the number of iterations and complexity will depend on if you are running this via a cgi script that someone inputs a domain or if you are using a spider and calculating the values and storing them in a db...
     
    hdpt00, Jul 28, 2004 IP
  13. jocs

    jocs Peon

    Messages:
    103
    Likes Received:
    6
    Best Answers:
    0
    Trophy Points:
    0
    #13
    I'm planing to do it, but I still think i need to madurate this.
    I'm sure it would be a nice idea, but i'm not sure if i will have the time to do it :confused:
     
    jocs, Jul 29, 2004 IP
  14. Johnny Linker

    Johnny Linker Peon

    Messages:
    2
    Likes Received:
    0
    Best Answers:
    0
    Trophy Points:
    0
    #14
    I don't know enough about the topic, yet, to give you constructive feedback but I APPLAUD YOUR EFFORTS!

    Johnny
    http://www.webbikinistore.com
     
    Johnny Linker, Dec 27, 2004 IP
  15. nohaber

    nohaber Well-Known Member

    Messages:
    276
    Likes Received:
    18
    Best Answers:
    0
    Trophy Points:
    138
    #15
    I don't understand the purpose of making such a tool (using this kind of algorithm). SEs don't use your idea, so what's the usefulness of building a tool working on an imaginary algorithm? It's just a waste of time.

    I don't care how your tool weights keywords. I care how search engines do it.
     
    nohaber, Jan 1, 2005 IP