Where does Google fail?

greenspun.com : LUSENET : Robot Wisdom : One Thread

Describe as specifically as possible what sort of queries you find Google doesn't do well on, and where you go instead?

-- Jorn (jorn@mcs.com), September 14, 1999

Answers

Google is very variable and I find it hard to categorise queries that don't work well. I was looking for seppuku links this morning and the ones returned by google just weren't right, so I tried northern light and got much better hits. But, this is purely subjective because the google hits were perfectly valid hits, they just weren't the sort of hits I wanted. Google definitely suffers from repeat-hit-itis which it would be nice to fix.

BTW I *never* use yahoo and never have.

L.

-- Lindsay Marshall (Lindsay.Marshall@ncl.ac.uk), September 14, 1999.


When you put your query up yesterday I was looking for information on and sources for Henna. So I stuffed the same few searches ("henna body design" and similar keyword combinations) into HotBot and Google. In the first 20 hits (as deep as I went because I found what I was looking for) Google tended to give me more links to suppliers of "temporary tatoos", body glitter, and more mass-market sorts of sites, HotBot many more sources for actual Henna powder, mixing and catalyst recipes, things like that.

This is consistent with my experience in general, Google gives me the popular sites, but I don't turn to the web for ad brochures, and especially where the terms are ambiguous or in widespread useage I want a more random look. A place where Google shines is if I'm looking for, say, the home page of a specific person, the definitive resource. But usually I'm looking for more esoteric information.

-- Dan Lyke (danlyke@flutterby.com), September 14, 1999.


Google fails completely on the words- "Drunks Against Madd Mothers". Oddly the four words- "where does google fail" don't give them any trouble. They used to be able to handle the query but "somehow" it broke. I brought it to their attention on Aug 29 telling them it was either broke or my site was being censored. In their reply they claimed they don't censor.

-- Joe Sixpack (Joesixpack@damm.org), September 14, 1999.

I'll confess that I don't use google for searching -- I've only used it for ego searching! I've searched on my name, the name of my journal, and on cacophony (for Seattle cacophony society stuff).

I usually use the quick search (typing in the address bar) in IE5, and if that doesn't get what I want, I cycle through all the search providers in the search bar on the side. I'm more likely to try altering my query (required terms, eliminating terms) than go to different *places* to search, though.

Anita of Anita's BOD and Anita's LOL

-- Anita Rowland (anitar@halcyon.com), September 14, 1999.


Google, like most spider-created search indexes, often fails to find useful or relevant information for recent events, new companies, or fresh buzz and gossip. Try searching for "Epinions," and you'll get zero results. This seems to be one of Google's only weaknesses. It not only takes time to spider the Web, it also takes a certain amount of participation on the part of the Web community to create links that Google uses to evaluate the "importance" of a site, a key factor in how it assigns relevance.

On the other hand, Google has emerged as by far the best "people search" engine, leaving former champion HotBot in the dust. Search on "Naval Ravikant," and Google gives you some most interesting leads for finding more about the CEO of epinions.com.

-- Chris Sherman (websearch.guide@about.com), September 14, 1999.



The word "mothers" seems to be missing from the index. Google can't find "mothers against drunk" either.

-- Pete Bevin (moose@bestiary.com), September 14, 1999.

Google seems to fall down for phrase searches for me. Instead of only selecting true phrases (the words appearing in sequence), it will often deliver pages where the words are only close together, if that. I tend to use AltaVista for phrase searches, as it does work right. I suspect that Google was not really designed for phrase searches, especially as they added the feature only some time after opening.

AltaVista also has much richer filtering for various sorts of things than Google does; it is, for example, currently much easier to find links to something with AltaVista, because you can search specifically for text in a link. (And correspondingly it is much easier to exclude a set of pages from the search result, if you know that they will give you lots of false positives.)

-- Chris Siebenmann (cks+rwis@hawkwind.utcs.toronto.edu), September 14, 1999.


I love Google, it is my primary search engine but the more specific your search is, the less likely you are to find what you are looking for. A lot of links are out of date (though the cached pages help most of the time) and Google doesn't seem to cover a large portion of the web. Also, it drops a lot of stop words out of the search automatically, even when the phrase is in double quotes. This annoys me when I'm searching for an exact phrase. Still, it works for me because it is fast, has no ad banners and quite good at finding good material on general topics.

My backup search engine is Fast Search; it is as good as most of the other "portal" sites and has no ad banners, so I can cycle through results quickly. Unfortunately, like most of the other portals, it returns a lot of useless pages and porno sites (which irritates me to no end).

Anyone out there have any searching tips? Maybe I need to have a more sophisticated search technique.

-- Oxnard Montalvo! (astroman@hickory.net), September 14, 1999.


Like Lindsay, I have a hard time classifying the queries that Google doesn't serve well. I've spotted a general trend in that Google does very well with technology-oriented queries, but not well across other subjects. Agreed with other posters that Google is utterly fabulous with people searches. Recently I did a search on "gentrification," and was disappointed with Google's results (it turned up shallow replies). The same word on Northern Light found a fascinating treasure trove of sites. Clearly, Google's failure in this aspect has to do with the number of sites spidered--NL has far more pages indexed than Google does.

-- Peter Merholz (peterme@peterme.com), September 14, 1999.

I recently did a search on 'disestablishmentarianism' and Google choked and refused to search for it. Too big a word? Copied and pasted into all-the-web returned 118 documents.

-- Paul Perry (perry@alamut.com), September 15, 1999.


Google is the main index I use. When it fails, I use alltheweb. Google is not searching the pages it indexes. It searches the index created from the pages it found. Thus, a word not found in the index will not be found on a search. It seems to have a filter for "words" and "garbage" and only "words" enter the index. For instance, Hebrew words are "garbage" for google allthough they are text in web pages. Another example, search for "mothers", it is not there. Google thinks that "mothers" is a "garbage" word. (no pun intended).

When Google fail, I use http://www.alltheweb.com

It will be interesting if the Google company will put the index in a web page for all of us to see.

-- Hanan Cohen (hanan@qglibrary.org.il), October 16, 1999.


urm: -www.google.com query--->googlescout no hits on "googlescout" (term was not used in search) Your search - googlescout - did not match any documents in this database.

allthenet is nicer anyway

-- H.-J. Bosch (bosch@geo.uu.nl), November 05, 1999.


I liked Google for getting the most popular pages on a given topic or keyword -- it filled the gap that Rankdex left -- but there are certain searches they can't do that AltaVista can: limit to a domain; search netnews; search for specific content-type, such as images (AV's image finder is useful for finding all images with say "kerouac" or "eiffel" in its name); search for words only as a specific phrase, with or without additional keywords or phrases; limit to a directory on a website; etc.

A few months ago I noticed they dropped the smart handwritten GOOOOGLE graphics used for multiple pages of output, and replaced them with stuffy just-like-everybody-else graphics. First thing I thought was that if they were so eager to please investors by removing their smart graphics, what else would they compromise?

Alltheweb may have the most files indexed, but AltaVista seems to have the most robust pattern matching language and the most beautiful output (when in text-only mode).

-- Michael Stutz (stutz@dsl.org), November 23, 1999.


google is good when you want to learn a new subject. (i.e. to get an introduction on something).

If you need *specific* info then ALTAVISTA or DEJA are better - they have the larger database.

-- Michael Moser (michael.m@sapiens.com), May 06, 2000.


Moderation questions? read the FAQ