Forum Moderators: open
I would love to see your comments about this.
Does anyone here feel that google is using Semantics , if so for how long it's could have been in existance?
Eagerly Looking forward to hear from you all.
Aravind
Ps: I have noticed that words like best and free tend to shift SERPS a little bit in my industry.
>>Ps: I have noticed that words like best and free tend to shift SERPS a little bit in my industry
are you saying you observe a better ranking for say "widgets", when adding the text "free" or "best"?
Personally, I would doubt that. They tend to be words generally discarded, or even a little poisonous?
from Googleguy: [webmasterworld.com...]
About natural language, just because we don't talk much about something doesn't mean that we don't work on it.
You mean like the example in the Google phrase correction - More than just a spellchecker [webmasterworld.com] thread where it was mentioned that when searching for
dante "diving comedy"it gives you Did you mean:
dante "divine comedy"Or do you mean that a search for
york newmight give you pages about new products in an English town called 'York' whereas a search for
new yorkgives you pages about an American city called 'New York'?
But there is hope: the've recently bought a small search engine that
parses pages for ads so who knows what else they are going to use it for?
I keep my fingers crossed!
Agreed, could mean a yes or no.
From the dictionary- semantics - "the study of language meaning"
From the thread vitaplease referenced- Googleguy "The spellchecker is context-sensitive, so it's much more than just a set of suggested corrections"
So it's context sensitive, google must be using "semantics" :)
Using it to spell check a word is one thing, but to order their whole index in some context sensitive matter is another thing....I guess we'd have to pinpoint why and where they would need to apply context to a word. The last place I would have thought is a spellchecker!
Spelling usually deals with the word form, not the meaning.....I guess semantics is just an everyday thing at google ;)
Say Suppose I am searching for online Widget..
Then the site with
title:
"Accomplised Online Widget Store"
meta description :"largest online Wideget store of the world."
Then First Alt tag with link:
" Accomplised Fact! World's Largest Online Widget Store"
has more weightage then simply
title:
arsfksadg's online widget store
Friends! I have been confused with two of my clients who tend say this is the case..certainly superlatives before keywords seem to have some effect.
I am personally responsible for the decision to go ahead but not very sure of this.
Certainly I need clarification ,Please do post me where were ever semantics are posted at webmasterworld.
Next Question abt Semantics: Appiled semantics is bought by google, context based results are good if they deliver what they tend to promise,if this is the case then google must be using semantics in some form in serps, in wat form they must be using?
Eagerly looking forward to your posts.
Aravind
Can you please elaborate on semantics being everyday thing?
IMO if they have context between words in the spell checker, they have context for the SERP's they return. I just think that the place where you would least expect the "context" of a word to be needed would be in a spellchecker, most spell checkers would see what characters youve typed and any words that "look" similar, not "mean" something similar.
To use an example, searching for sheets [google.com], in which part of "Bedroom > Bedding" does it mention sheets? :) Certainly nothing to do with the spelling of the word, though the cat does seem somewhat related to the query in meaning.
Reading in alot of places, Google keeps their word dictionaries in memory for query time, could just be that words also have pointers to other words, giving it it's semantic "feel" :)
The way their dictionary works is probably way different to the way they collect pages, nonetheless, like in the sheets example, you have to wonder what they can do "across the board" with a spider that could understand the differenec between a bed sheet and a sheet of metal.
//
0.02 anyway, there seems to be alot of work going on out there to automate even the meaning of words, this search for example [google.com]
Have look at this [ai.mit.edu...] START
This is mit's useful Natural Language Processing Search Tool.
Unfortunately it doesn't return results on sheets
.Are we looking at google doing this at miniscule level.?
Yeah as you say it must have it's own query dictionary.There is some clue here.
I would like to hear more on this query based dictionary?
Vitaplease: IS Accomplished too a poisonous word?
Hitprof: the've recently bought a small search engine that
parses pages for ads so who knows what else they are going to use it for?
Can you elaborate on this?
you might find this interesting concering semantics:
[javelina.cet.middlebury.edu...]
and more specifically:
Discard articles, prepositions, and conjunctions
Discard common verbs (know, see, do, be)
Discard pronouns
Discard common adjectives (big, late, high)
Discard frilly words (therefore, thus, however, albeit, etc.)
Discard any words that appear in every document
not that Google should necessarily use or apply any of this...
It's unclear whether it will be used for the serps or even what it does at the moment.
[catalogs.google.com...]
Just started to develope a tool on the same.Will get back as soon as I am through with understanding of what ever I am reading.
I have made one observation :
When you type hotels on line google doesn't give spell check, but when you type on line hotels it gives did you mean online hotels.
Y does this happen?
Aravind
Hotels is a 'valid' search phrase all by itself so it doesn't require any correction.
Sorry.
But again Hotels On line is not quite often searched either..this doesn't generate 'did you mean' " Hotles Online"
Only when I type : "On line Hotels" does this spell check occur not when I type "hotels on line."
But when it comes to "widget on line" it gieves did you mean " widget online" and again when "on line widget" is there no spell is given.
Is it purely becoz of more searches for that type?
Aravind
I also wonder how well the semantic technologies transfer to other languages.
One paper I found very interesting is this one on Automatic extraction of synonyms: [inma.ucl.ac.be...]
One may wonder whether the results obtained are specific to the Webster’s dictionary or whether the same methods could work on other dictionaries, in English or in other languages. Although the latter is most likely since our techniques were not designed for the particular graph we worked on, there will undoubtedly be differences with other languages. For example, in French, postpositions do not exist and thus verbs have not as many different meanings as in English. Besides, it is much rarer in French to have the same word for the noun and for the verb than in English. Furthermore, linguistics teach us that the way words are defined vary from one language to another. This seems to be an interesting research direction.
[labs.google.com...] might be interesting as well.
re: synonym exctraction PDF: it's great to see the various ways people are using graph theory.
Another challenge associated with semantic analysis is having to rely heavily on accurate text parsing. Some asian languages, e.g. Japanese, do not have spaces between the characters in a statement so you need a smart parser.
As far as semantics at Google, this thread seems to cover minimally three separate topics:
1. context-sensitive spell correction
2. query disambiguation using some level of semantics (sometimes called 'utterance interpretation')
3. web page content analysis (strong NLP)
This thread looks to me to lead to somewhere near semantics.
[webmasterworld.com...]
LSI,Graph theory are being used by Monika R. Research Directory Google 2003) ...
With Regards
Aravind
