评测9238论搜索引擎的评测方法

9238论搜索引擎的评测方法

Since then, the unique search engine technology has emerged oneafter another, and now it is obviously in the Warring Statesperiod. However, people' s evaluation methods have not changedmuch. Now, the common evaluation is simply to compare the searchspeed, the number of search results and the accuracy of theirsearch by using a few keywords.

Far not said that in the first quarter of 2001 after the upgradeof the askjeeves, you can play like as a phone in any phone inthe hands of askjeeves phone number, can also be labeled on thepage to enter the online about online speech, using yourcomputer' s microphone and speakers to communicate. Then youjust orally to it a request, it will put your voice into text,and then analyze your request to 7 million standard answers,it and other 2 million multimedia repository and Internet tofind the answer, find and then converted into voice to answeryou.

Imagine, if you ask, "the recent American election is up in theair. What do Americans think?"" After a while, the computer ortelephone to answer you: "according to the latest survey, thelast is if Bush is elected, 80% of Americans will accept himas the legitimate president, if Gore is finally elected, 79%of Americans will accept him as the legitimate president. " Ifyou ask, "who scored in the last World Cup finals?" "It answersyour name and also brings up the audio and video clips of thefinal goal. (of course, the audio and video clips are based onthe fact that you're not using the phone, but the computer) .Although, askjeeves think their speech and search speed has to

be the degree of commercialization, but it will still have manyimmature, if you take a few keywords to test its search speedand precision and recall, and many of the common search engine,it came in where? If it' s in the back, isn' t it a lousy searchengine?

One is evaluatingthe Internet searchengine is averydifficultthing, but a lot of evaluation results are ordinary Internetusers to see, is bound to take the Yahoo, include Sina portal,for them, is just a part of the Internet search, other kindsof search how to do? If you don't count, but the net civilianmuch; if it is, is a mess, where to?

Here, let' s analyze some of the capabilities of severalimportant evaluation elements:

I. recall

Since it is a search engine, first of all, than the search scopeis a matter of course, if this fails, the latter does not seemto participate in the evaluation. Because the number ofincluded pages each search engine announced, can the wholeletter, with a keyword search results is obviously, so thegeneral evaluation on this subject.

But to this date there are still many problems, most decentpoint of the search engine I can find a number of keywords toprove its search results is the most complete. Because althoughthe number of pages indexed in size, but the robot and spiderprogram, index scope and index standards are not the same, thebiggest search engine to be much smaller in the search engine

to search.

Some search engines support "about", "of", "ah", "what" and soon "function word auxiliary word search", and some don'tsupport it. How does this compare?Which evaluation mentioned?Key words besides the content is difficult to choose, also notvery good in the length. Some search engines don't supportsingle character search at all. How do you count it?Generallyonly compare single keyword search, and multi keyword searchit? How long is the search for long sentences? Even searchengines can support any articles or fragments as keywords, sothat the results of the comparison of the search keywords, butnot the same, let alone the function of no comparison. Thesemantic search engines like excite, as well as engines thatsupport fuzzy search, and other search engines that search forvery few or even zero keywords, can ferret out a whole bunchof results, and how do they compare?

Finally, the search engine can optimize the results forspecific keywords. Who will guarantee the fairness of theevaluation? If one of them is evaluated by a search engine inadvance to know the key words used, then the championship isthe only one that can be easily optimized.

Two: search speed

Recall is faster than the search speed, and if there are manypages of search engine index, but search for a second five orsix seconds or longer, please direct it out, there is nosignificance than go down.

The problem of speed, first of all in keyword, single keywordsearch fast, not necessarily more than keywords search fast.Then there is the problemof traffic, which is unfair to a searchengine with more than one hundred million of the daily trafficand a search engine with tens of thousands of visits per day.And the number of pages indexed, a search engine index 1 billion", another search engine index ten million", let them on thesame keywords in the database search results than the searchspeed, so how to convince people?

In addition to optimization problems, some search engines havethe memory search results accelerate the ability to transfera keyword, even the first word search took 10 seconds, secondsearch may be 2 seconds, third times, fourth times, when yougo to the test has always been 0.0001 seconds. In this way, ifyou choose a common word test, it is amazing, if a remote word,perhaps the old half day can not come out, in the end what theelection keywords?How much do you usually use? This is reallya silly sum.

Search engines are not for testing on the local machine in thelab, but for ordinary users, so the search time should includethe search interface and the transmission of search results.A search engine took 0.0001 seconds, but the resulting page took3 seconds, and another search took 0.5 seconds. But it took asecond to transfer the page. Which search engine do you thinkis faster? When you really use it, do you choose the 3.0001seconds later to see the search results, or 1.5 seconds later

to see the search results?

Three: precision

This is pretty important. Even if you search something that isfast and fast, you don' t know how many pages you want to find,So what' s the search result for? This kind of search engine isuseful only when searching for rare things, but to check outrare things, you should go to meta search engines. Why use it?The standard of precision is difficult to determine. It dependson what you check. You have to check a particular website andfind a similar web site. The key to precision is to search whatand what keyword to choose, the judge can decide at random, andthen affect the reliability of the evaluation result.Four: dead link

General search engines always have some search results that aremissing, less than one percent, two, or eight or nine, and thisis often used as one of the criteria for evaluation. But asGoogle uses a web snapshot feature, there are few dead linkproblems, and you can still see Google' s own web pages even ifthe site in the search results is closed. How do you calculatethis kind of dead link?

Five: user burden

Have not seen the domestic search engine evaluation, who usedthis one, but it is to evaluate the advantages and disadvantagesof search engines, including many aspects. Search engines are

for human use. They must be comfortable, convenient and fast.Any user who hinders and delays the user' s access to the finalsearch results is considered a user' s burden.

The first is the search interface, a pure search engineinterface that has a search box. Compared to a portal with adsand a large amount of web content, the search burden they bringto the user is high.

The second is to describe the search results, search resultspage description of the text is long or short, "the textdescription index with keyword part or the beginning of indexedpages indexed pages or a few lines of the main content, keywordsare highlighted by what color is not displayed page address,and the searchresults page layout, the the user' s searchburdenthere is a big difference.

Effect of addition is the user steps, whether can use the mouseto start the search, the search results page shows the numberis only 10, page convenient or not, the search box is two ora, above or below, a search keyword search is still displaycable box, every one of thesewill affect the search efficiency.Six: there are other

Do you want to search in this directory?,

Internet Index database update time,

Stability of search engines,

The support ability and strength of advanced search should alsobe evaluated.

A person is not thoughtful, and there may be other importantevaluation factors I did not mention, if you think of, hope toinform. See here, everyone on the limitations of the currentevaluation methods commonly used search engine must understand,of course, the most ridiculous is that I do not know is ignorantor tricky or special selection criteria, some Chinese searchengine evaluation this year to do not even include Google, aswell as a long list of celebrities can row the violin missedPaganini.

It' s really hard to evaluate search engines.