| 9:10 pm on Aug 5, 2003 (gmt 0)|
> In a nutshell how does the average "internaut" gain access to "the secret Google"?
WebmasterWorld is the answer. ;)
| 9:12 pm on Aug 5, 2003 (gmt 0)|
I suppose Google can also implement the "users who search for this also searched for that" into the algo and many more things.
As great as the ~ is, it will never replace the OR.
Simply because many words have multiple totally unrelated meanings to which a sea of related words from totally different themes would match. you will always need the more specifics to get what you really are looking for.
On the note of international compatibility, no ~ on Spanish keyboards either. Am also using the Alt 126 approach around that problem. But then it is a Spanish International keyboard layout, the latin american keyboards might have the ~. but if they don't either, you might be excluding quite a bunch of people with your special character.
| 9:14 pm on Aug 5, 2003 (gmt 0)|
Good question PBG. even worse, lot of people dont know google. they still use MSN search, since default, to search for anything when their handful URLs is spelled wrong.
Even many of those who use google, it would be interesting to know how many people actually surf for anything except basic search. i know lotsa of people, who search a lot and still who dont know "and" is default in google and type blue+widget.
Sad Sad Sad :(
|Please Be Gentle|
| 9:19 pm on Aug 5, 2003 (gmt 0)|
> In a nutshell how does the average "internaut" gain access to "the secret Google"?
WebmasterWorld is the answer. ;)
Indubitably, Giacomo. Admittedly, it is only since joining Webmasterworld that mine eyes have seen the glory of the "minus" and the "OR", but I don't believe that joining here should be a mandatory prerequisite to using Google properly (although who knows?)
| 9:25 pm on Aug 5, 2003 (gmt 0)|
The best way to spread the word about the ~ operator and improve the user experience would be, IMO, to suggest its use when appropriate, just like the spelling suggestions.
Something like: Have you tried: html ~help [google.com]
|Please Be Gentle|
| 9:39 pm on Aug 5, 2003 (gmt 0)|
The best way to spread the word about the ~ operator improving the user experience would be, IMO, to suggest its use when appropriate, just like the spell suggestions.
Have you tried: HTML ~help
Good point Giacomo, I like the suggestion. Just as a matter of interest my one question would be how would Google determine which part of the query was permutable? I mean is Google advanced enough to say
Have you tried: HTML ~help which would be a more obvious choice than Have you tried: ~HTML help (or even Have you tried: ~HTML ~help)?
With the second scenario there would be less synonyms for HTML whereas help would be an ideal candidate for permutations and combinations. How would Google determine this candidacy. Would it a) look at related queries posed by other users to determine which is the more flexible term, b)check to see the number of synonyms for html, the number of synonyms for help and therefore calculate the probability that help is the more fluid of the two terms, or c)none of the above.
With spelling suggestions it has a store of data from which to derive suggestions eg Elvis Presley for Elvis Parsely etc. but how would it determine the weight of each term on the fly? Again, I do like the suggestion of:
Have you tried: HTML ~help
google could elaborate on it to
Have you tried: HTML ~help as this will yield synonyms or related terms for help such as tips, guide...
and thus indicate to the users the accepted related terms should the user feel the need to later add in "OR primer" for instance.
Regards, kind as ever
| 9:47 pm on Aug 5, 2003 (gmt 0)|
I wonder how many people (percentages) actually use those special operators.
btw: never thought of using ¦ .
| 9:57 pm on Aug 5, 2003 (gmt 0)|
PBG, your question is not so easy to answer I'm afraid. :)
b) might be a good quick'n'dirty technique to start with; however, I guess Applied Semantics' engineers may have far better approaches. ;)
I think the biggest problem with deciding which keyword Google should suggest to change is context. It's very hard (nearly impossible, IMO) to try to guess, let alone "understand", the meaning of a search query, which is usually a very short sentence (compared to natural language questions).
Many two-word phrases have literally dozens of possible meanings, often belonging to totally different contexts. Since computers are so good at math, but not nearly as good at imagination/creativity etc., it's much easier for a search engine to do some type of syntactic analysis and return a set of possible spelling errors (heck, you can even do that with regular expressions!), than perform semantic analysis, which requires at least some form of contextual information.
Where could Google get/extract this contextual information from? Well, previous query sets, for example. Personally, I would be willing to let Google know a little more about my search history in exchange for a more "intelligent" guess at what I'm looking for.
|Please Be Gentle|
| 10:00 pm on Aug 5, 2003 (gmt 0)|
Percentages,in June I asked Googleguy "With respect to refining searches, does Google publish statistics on the frequency of use of search modifiers such as "minus" or "OR". ( On the google papers site I found “Improving Category Specific Web Search by Learning Query Modifications” ")...." and his response was as follows:
"I don't know if we've ever discussed the frequency of things like minus and double-quotes. I do know that it's a moving target as users get more familiar with searching.. ".
|Please Be Gentle|
| 10:17 pm on Aug 5, 2003 (gmt 0)|
I should also add I was somewhat surprised at this response, if I were Google I would be very concerned about this regardless of whether it was a moving target. I read the aforementioned pdf (“Improving Category Specific Web Search by Learning Query Modifications”) and was dismayed to find that it did not refer to human beings learning query modifications, rather it focussed on training a "support vector machine to classify pages by membership in the desired category". The title was promising but I do have a fear that human beings could somehow get left behind in all this! I am sure that helping the user is something of great import to Google, and often the helping hand we seek is at the end of our own arm. So I believe teaching the user to teach him/herself should be deemed as important as teaching classifiers, support vector machines etc. how to modify queries (Give a man a fish, feed him for a day, teach a man to fish, feed him for a lifetime.)
Regards, ever gentler
| 10:23 pm on Aug 5, 2003 (gmt 0)|
On a side note, I think Google's tilde operator is similar, in its functioning, to Teoma's "Refine" and AllTheWeb's "Clusters" features.
Try searching for "html help" on Teoma or ATW, for instance: you will get "html tutorials", "html tags", "html code help" and so on as suggestions (look right on Teoma, near the bottom on ATW).
Except for the (not very useful) "Similar Pages" feature (A.K.A. GoogleScout), I'm sorry to note that Google doesn't have anything like that to date.
ATW also has a "similar queries" function that I think should be pretty easy to implement (just select other queries having similar results sets, and there you go) and very useful to the user.
But I think I'm going a bit off-topic here, so I will just add that I hope Google will get more and more "intelligent" in the next few months/years, as we move towards a Semantic Web.
| 11:04 pm on Aug 5, 2003 (gmt 0)|
Hey, good thread--thanks all for the comments. It's always tricky to find the right balance of power features that users can try vs. trying to do things automatically. Certainly things like tips can help to educate users, but the flip side is that in an ideal world, users wouldn't need much training or special syntax to find what they want. Rest assured that we'll keep looking for ways to make searches more powerful (and easier, too)!
| 11:21 pm on Aug 5, 2003 (gmt 0)|
This discussion really intrigues me (partly because I have a semiotics background), so I hope nobody will mind if I add a few more considerations.
Google really lacks a "Refine/broaden/fuzzify your query" feature IMO (BTW GG, all concepts are "fuzzy" by definition), where a user could click on a list of related queries. I hope to see it soon in the Google Labs technology playground. Let's call it Google Serendipity: sounds great, doesn't it? ;)
More seriously, I agree in part with PBG where he/she says it's easier to teach humans to think than machines, although I believe that search engines do need to move further towards users. Even Google, no doubt the most user-friendly SE today, needs to get more and more easy to use, yet without overwhelming searchers with too many options/alternatives (info overload, anyone?).
Google must learn how to "intelligently" guess when users need suggestions about their queries. Since contextual (semantic) information is totally lacking from search queries as we know them, Google should rely on the "pragmatic" side of search activity to extract useful information about a user's scopes and intentions.
For example, let's consider the following (typical) scenarios:
a) user enters a query, clicks on first page of results;
b) user enters a query, clicks on first page of results, presses "Back" button, changes query;
c) user enters a query, pages through N pages of results, eventually changes query.
I think that both cases b) and c) have a very clear meaning: the user is not satisfied with the results for his/her first query (or, the user is an SEO doing rank checking, but those should be pretty easy to spot :) ). The reason might be poor results quality, or a poor search technique. In both these cases I think Google should provide the user with suggestions and/or alternatives (such as a clickable set of related queries) derived from recent search history and behaviour, in order to let him/her refine the query and get better results.
I don't know whether Google make use of this kind of information about user search behaviour on a large scale to improve results quality in real time; it is probably used in usability tests only. I think it would be nice if Google took a closer look at its users' behaviour, and tried to be a little more proactive at suggesting ways to improve their search experience.
<added>You're welcome GG, I'm always glad to give my $0.02 about Google's search quality. Plus it's a great way to kill time while waiting for the next update. ;) </added>
|Please Be Gentle|
| 11:46 pm on Aug 5, 2003 (gmt 0)|
To answer some of your questions, a. I am definitely a "she" (either that or I glean some sort of masochistic pleasure from joining long queues outside of ladies' restrooms), interestingly I assumed my gender was immediately apparent from the way I wrote. I digress. You said:
Google really lacks a "Refine/broaden/fuzzify your query" feature IMO (BTW GG, all concepts are "fuzzy" by definition), where a user could click on a list of related queries.
What I have found ( and Googleguy may comment on this) is that if you go to the adwords sandbox they can provide you with suggestions.
Google Guy said:
in an ideal world, users wouldn't need much training or special syntax to find what they want. Rest assured that we'll keep looking for ways to make searches more powerful (and easier, too)!
I agree that in an ideal world users wouldn't need much training, but in the real world some of us do.. We are simply endeavouring to highlight the gulf between the ideal and the real user and how Google addresses their respective needs. I have no doubt that Google will strive to make searches more powerful (it cannot afford not to) but the point is if they do not make these features painfully obvious to some of us, we will never find them or get the opportunity to use them. With all due respect, not all of us have degrees in Maths, Science or IT and maybe we don't think the same way that Google professionals do but I think Google should embrace the rich diversity of users and their search needs and explore different solutions to these challenges.(One size will not fit all). As an example the advanced search page for instance may not address the needs of many users. Personally the first time I saw it I found it a little offputting. Even the term "advanced search" implies advanced abilities, so perhaps Google could offer alternatives to this such as "customized search" or even "refined search" which would be less daunting. That is just a minor point, but what worries me most are the options th user is faced with.
with all of the words
with the exact phrase
with at least one of the words
I know it may sound stupid but an ordinary person may look at that and say "well, if i am using the exact phrase, surely by implication I am using all of the words" or perhaps "well if I am using all of the words I am using at least one of the words". They are not told the raison d'etre of these distinctions eg. that with the option "with at least one of the words", Googles wants either wordA or wordB ( wordA OR wordB in Google terms). It does not explicitly state this and thus the user probably won't learn how to use the OR from this. The user enters the query and probably looks straight at the results without checking the query generated. I do not find the phrasing here concise or unambiguous. My main gripe is with the term: "with all of the words". A user may put in a phrase "the rain in Spain" and would assume from the term "with all of the words" that all of the words are included. This as we know is not true. "The" and "in" are not included by default here as they are stopwords (a list of stopwords would also help, yes once again I digress). I know it is a minor point and a very trivial example, but the user won't check the generated query and will assume all the words are included when they are not. They won't get what they expect and won't learn anything in the process
What I am saying is that I appreciate any new features ( and frankly a -~ equivalent ie an operator to omit a word and its cognates/synonyms would be very appreciated!) but personally I would concentrate on basics and let people know what was already available.
I will leave you in peace now, sorry to bother you but I am curious (in every sense of the word!)
With peaceful regards
| 1:11 am on Aug 6, 2003 (gmt 0)|
Sorry I didn't guess your gender from your writing style/nickname, PBG. :)
I do use the keyword sandbox; however, that tool was made with AdWords advertisers in mind, not searchers. I would like Google to offer a similar tool, more easy to find and use, and at the same time more powerful and effective, specifically made for users' needs.
About the advanced search page:
|the first time I saw it I found it a little offputting. Even the term "advanced search" implies advanced abilities |
A couple of years ago Jakob Nielsen noted that the average searcher has poor search skills [useit.com]. They use 2-word searches and do not know how to use simple boolean operators like AND or OR. Therefore, Nielsen suggested they should not be offered advanced search options by default. If I'm not mistaken, he also explicitly suggested (in his best-seller Designing Web Usability) that the advanced search page be labeled with a scary name in order to deter non-power users from clicking on the link. ;)
Users are also not very prone to reformulate their queries when they are not satisfied with their search results, says Nielsen. Therefore,
That's exactly what Google did. And it has worked great so far.
So, how can Google introduce useful search refinement tools to the great mass of web searchers?
Google has adopted a "pull" strategy so far: let's make a "cool little fuzzy concept operator" and give it a geeky name like tilde (~); spread the word on Slashdot and WebmasterWorld, and if the Google geeks like it, their non-techie friends will follow. ;)
I think it's time for Google to switch to a "(soft) push" strategy, introducing and publicizing advanced search features one at a time, by giving discreet suggestions whenever they are appropriate.
In part, they already do that: if I search for "html AND help", I get the following suggestion:
|The "AND" operator is unnecessary -- we include all search terms by default. [details [google.com]] |
I just hope the same will happen soon with the tilde op and whatever useful search widget will come next. GG, I understand that finding the right balance between complexity and usefulness is a tricky task, especially when you're dealing with a worldwide audience of users. But hey, someone has to open a road sooner or later. :)
Finding a way to painlessly improve their users' search skills might reveal itself as Google's greatest success of all times: paraphrasing PBG, "Teach a man to search, and he will use Google forever". ;)
| 4:40 am on Aug 6, 2003 (gmt 0)|
I have found this new feature useful in a way not yet mentioned. We've talked before about the differences in spelling betwen countries and I know that I personally find this a problem while doing searchs (http://www.webmasterworld.com/forum3/4239-12-15.htm).
So when I read this thread I was hoping the tilde would help my specific problem. And voila, a nice SERP returning fairly specific scientific phrases in all forms of spelling, including the ones I don't know :D It's not perfect and not all terms I tried were in there (particularly the more obscure ones), but it works for most. It also returns different forms of non technical terms too, giving a wider range of results than without the tilde.
I'm not sure how much this will help your average web surfer. The type of spelling used on a site can be a good way to brand that site as being from a certain part of the world. But at the same time a lot of people I've talked to aren't really aware of extent of the differences, particularly between British and American English. So hopefully using the tilde will give them results they would not have otherwise seen.
It's likely that some of my pages will now be found by people from a wider ranger of countries (assuming a wide adoption of this operator of course). Whether this is a good thing probably depends on the intentions of each webmaster, but it suits me personally.
| 4:56 am on Aug 6, 2003 (gmt 0)|
They are certainly interested in how we are using it as they appear to be tracking the clicks from those search queries. Perhaps they are looking at whether we click on results from the original keyword set or results with the synonyms and suggestions.
| 7:53 am on Aug 6, 2003 (gmt 0)|
in places i'm finding it's really useful - for example:
~renal ~disorders --> kidney diseases
however, the biggest flaw is not being able to contract multiword phrases
- none of these map to chondrogenesis - the major keyword for this topic
and there are some rather dodgy synonyms:
| 8:25 am on Aug 6, 2003 (gmt 0)|
Thanks PBC. In my logs no more then about 10-15% of all searches show any sign special operators, "" included. I don't have exact numbers, but what I see most is "" and + . Most searches with special operators come from the 'Advanced' page.
Added: Hey, I actually used the ~ operator, it works great to search webmasterworld (except that I don't understand why reference is highlighted for ~commands).
| 8:51 am on Aug 6, 2003 (gmt 0)|
>~spam --> email, mail
Would you rather see ~spam --> email, seo? :)
| 9:14 am on Aug 6, 2003 (gmt 0)|
Last year around this time only 1% used Google advanced search in France:
and 16% use one word in searching.
I'd say Google should offer Altavista Prisma type search query expansion for users using only one word.
(force-fed education to be more specific)
Dec. 2002: Jeff Dean from Google mentioned some of Google's future plans, which include
|Conceptual understanding (perhaps that Google will try to guess synonyms) |
(and only 3% of searches are advanced search - 10% misspelling!)
also mentioned in this thread: [webmasterworld.com...]
| 6:23 pm on Aug 6, 2003 (gmt 0)|
> Would you rather see ~spam --> email, seo? :)
Well, ~seo matches search engine optimization and ~optimization matches performance (though not the other way round).
A very nice feature.
| 7:04 pm on Aug 6, 2003 (gmt 0)|
does this mean someday, one might be able to bid on ~widget in Adwords?
| 8:13 pm on Aug 6, 2003 (gmt 0)|
So ~ just means (sym1 OR sym2 OR sym3) search
So instead of doing a:
HTML ( help OR glossary OR tutorial ) search
I would do
HTML ~help search
So then why doesn't google when it sees "widgets" do a "widget OR widgets" search?
| 8:17 pm on Aug 6, 2003 (gmt 0)|
~ operator is case sensitive
| 8:23 pm on Aug 6, 2003 (gmt 0)|
I searched for ~Apple and it listed Windows as one of the synonyms.
|Please Be Gentle|
| 8:30 pm on Aug 6, 2003 (gmt 0)|
Can anyone confirm whether the tilde is in fact case-sensitive please? I did ~MARATHON upper and lower case and got the same results. Is it really case-sensitive, if so it may defeat the purpose.
Just a couple of points
You said "let's make a "cool little fuzzy concept operator" and give it a geeky name like tilde (~);".In all fairness to Google they didn't invent the term tilde, there has always been a tilde above the Spanish "enye" ie n with a tilde eg in the spanish word for Spain, and Portuguese has a tilde above its a,e,o and u. As far as I know it exists in eastern European languages. Personally I prefer to call it a squiggle or a fuzzy wave. To highlight its existence maybe Google could launch a competitiion to name the new operator, and in the process highlight its functionality. A couple of lava lamps and a beanbag may just do the trick.
Seriously though,they should have led upto its introduction with fanfare, built up a bit of suspense as it is worthy of attention. With regards to Nielsen, I agree with part of what he said:
"the emphasis should be on increasing users' success on the first attempt.".
Absolutely, time is precious and speed is of the essence. However the next part perturbed me:"Nielsen suggested they should not be offered advanced search options by default. "
Surely that would not be conducive to enriching the user's search experience. I hope that Google has not adhered to the accessibility guru's maxim on that one. Admittedly, neither Google nor Googleguy is likely to admit to this if it is the case-" Yeah, well actually we're about to passwordprotect the advanced search, and users will have to answer questions on quantum physics before using it - only genii need apply..". I don't think it is likely but it would explain certain things. For instance with the combined brainpower of Google they could have come up with a selection of friendlier user interfaces and means of engaging the user if they really wanted to. Also a look at
the Google API forum would bolster your theory. Google could place a higher premium on search if they wanted to. I mean certain users would be encouraged to enhance their search skills to save time or even money. Others are motivated by "a little gold star on the copy book" - Google could introduce search skill qualifications, similar to the basic exams of certificates such as ECDL (european computer driving license). Who knows, if one were applying for a job as a librarian or researcher, such qualifications may be coveted. Once again I digress.
To be is to be perceived. Google needs to be seen to pay more attention to the normal user's needs. With the threat of Microsoft and Yahoo looming this is more pertinent than ever. Microsoft prides itself on user interfaces (although I still have nightmares where I am chased by animated
paperclips) and Yahoo also concentrates on personalization and portals etc. Google needs to compete by leveraging the strengths of the opposition in my humble opinion. The war
for the end user starts on the query input page - he/she is not necessarily aware of the sophistication of the search algorithms.
ps Giacomo, you said you had time to spare while you awaited the next Google update. I am no expert ( I don't have a website so strictly speaking these updates don't really affect me) but I am not certain that there will definitely be another update. Anyway best of luck with it,but don't hold your breath
|brotherhood of LAN|
| 8:34 pm on Aug 6, 2003 (gmt 0)|
So anyone have a brave theory as to how it works?
Justa hunch for me.....its based on IDF, so when you type "scottish" its looking at all the pages with high frequency for "scottish" and are added up alongside other high frequency words. They are paired alongside other documents that contain the keyword "scottish".
So when I do a synonym search for "scottish", its coming up with "bbc","scotland","edinburgh"....words that are typically high in frequency in an on topic document.
i.e. the bbc, with one of the largest scottish sites, will have a high frequency of "bbc" and "scottish" over a large number of pages.
So even though they are not synonyms they appear as synonyms in the search. Maybe a search guru could explain it better but to me it seems like it works along those lines.
anyway, there are already dictionary/thesaurus' that could have been used for this function?
| 8:35 pm on Aug 6, 2003 (gmt 0)|
in clarification of case sensitive, i mean that 'run' is different from 'Run' as can be seen from
and it could be that it is specific to the words that a ~ operator pulls for its lists of synonymic-words.
| 8:54 pm on Aug 6, 2003 (gmt 0)|
I've not had a chance to check yet..but its possible the synonymic words could be the same ones that appear on the adwords search tool. If not I'd really like a tool to show them all (as oppossed to having to search through lots of pages!)
| 10:47 pm on Aug 6, 2003 (gmt 0)|
Do you mean like word proximity, or is it semantic spacing. The problem then would be that it can go too far. For instance search for ~swimming. The fourth result contains the word pool, but clearly it is unrelated whereas others contain the word pool as it relates to swimming.
| This 73 message thread spans 3 pages: < < 73 ( 1  3 ) > > |