Chapter Four:

 

In this lesson we will discuss how search engines work in general terms, not all possible scenarios (or search algorithms!).

 

What is a search engine, and how does it work?

What we think of as a search engine is really a team effort. There are 3 "members" of the team -- a  mechanism that identifies web pages to be included in the database, a  mechanism that indexes the sites and a searching mechanism with an interface, which scans, for keywords within the index. Users search the index (and hence, the database or web documents) through a query box or a template.  Documents in which the search terms occur are presented as "hits." 

Although some facilities incorporate "natural language" searching (searching by asking a question "Where are the doughnuts?"), most search tools retrieve "hits" or "matches" by seeking occurrences of your search terms within its database and by attempting to match the terms (converted to a "string" of data bits) against its index. Because the terms are converted to a digital string, the search engine must somehow be instructed to include plurals and alternate forms of a term 
 
 
Note:
Although some search tools automatically include plurals, many do not. If you are interested in "dogs," search for "dog or dogs" or use a wildcard such as * 
(A wildcard is a typed symbol which simply means "put any character here").
Some search engines also allow "stemming."
(This involves using  a special character symbol which simply means "put any ending here after this point".) An example: the term comput&  (where &=stemming symbol) would bring up hits from the following words: computer, computers, computing, computation etc.

 

What's a 'bot?

A 'bot, otherwise known as an intelligent agent, spider, crawler, robot, or worm, is an automated device (software) which may be programmed to search for terms (data  "strings") matching certain criteria. In terms of web search engines, a 'bot identifies and notes the url's of web pages to be included in the database. Later, another 'bot comes along and works on the interiors of the web documents, recording occurrences of  words and their position within the text. This information is used to create a huge index. 'Bots travel along the links of a web site, that is, they crawl or traverse from one hypertext link to another. 

 

 

What's the index for?

The index is how the search engine locates the url's which match your request. The web documents containing the query keywords are presented as a listing which may include a brief summary of the site. A simple way to understand the index is to think of it as a computerized book index. To discover where a topic occurs in a book, we would look up the word in the index which would indicate the page number(s) where the term occurs. Now imagine that every single word is included in the book index. A computerized version might be represented like this: 
 
 
Keyword
Number of times keyword occurs in book
Position(s) in book of keyword
Page number(s)
Apple
175
 title page, page 1: first paragraph word #5, page 2: first paragraph word #20, second paragraph word #15, page 5: 2nd paragraph word 21,...etc. etc.,  in summary title page, table of contents, 
pages: 1,2,5 etc.,12,25.
Orange
22
 table of contents, page 3, first paragraph word #3, page 17; first paragraph word #30, page 21 etc. table of contents, 
pages 3,17,21 etc.
Grape
3
page 50, 2nd paragraph, word #18, page 52, 1st paragraph word #41, page 53, 1st paragraph word #4
 pages 50, 52, 53

Some immediate observations might include:

  • a) the word apple occurs a lot in the database
  • b) the word apple occurs in the title
  • c) the words apple and orange occur in the table of contents
  • d) the word grape does not occur in the title or table of contents.
A search engine uses its index to retrieve web documents in which your search terms occur. The index lists the term and where it occurs (the url or address of the web page) much like a book index. 
 
 
Remember: a search engine returns hits only from its own database, that is, web pages which it has indexed. So if the site you are looking for has not yet been indexed, it won't be in the results listing no matter how magnificent your search strategy or statement.

 

How does a search engine decide how to list web sites matching the search terms I use ?

Each search engine uses a different algorithm or method to calculate something called a "relevance" which it "ranks." Have you ever noticed the numbers which sometimes appear next to the url's in a listing of search results? This is the "relevance ranking." Relevance means the probability that the "hit" or "match" is on-target with your query. The creators of search engines change the way they calculate relevance and do not tell us mere users their methodology; being high in the major search engines' rankings on a topic means big business. 

Sometimes Web site owners try to skew the odds of appearing on the first page of "results" for folks searching specific keywords. Being on the first page or in the top results increases the likelihood that the site will be seen and hence selected by the user. Unscrupulous folks "spam" the search engine to try to improve their rankings (and hence, their Web-based business) in a variety of methods including using "invisible text" (where text is colored the same as a background) or repeatedly using keywords in "meta-tags" (descriptive information not usually seen by the user unless when viewing the "page source" -- seel below). 

Perhaps most unsettling is the rising trend of some search engines which  in effect, sell higher ratings to companies willing to pay for the privilege. Most users will be unaware that the set of search results has in effect been manipulated to boost these companies ratings artificially. 
 
 

Exactly how relevance is calculated is protected, proprietary information but it is important to be aware that search engine providers may have alliances or agreements with other businesses (reciprocal and/or financial) which may affect  search results.

In general however, relevance is calculated by noting where the term occurs within the text and assigning this position a "weight" or level of importance. Some search utilties also include a popularity element in calculating the relevance algorithm; that is, the more a site is linked to or used, the higher the rating. Search terms occurring in the title, summary, in key positions within a paragraph or appearing several times within a paragraph usually carry more "weight" because there is a higher probability that terms in these positions indicate significant material on the topic. 

This is very similar to our book index example above; because the term apple occurs many times and in key positions (title, table of contents, beginning of paragraphs) there is a high probability that the document contains significant information about apple. Note that orange also occurs in the table of contents, an indication of the term's relative importance (it is a significant topic, but not as important as apple). The algorithm of the search engine and the methodology it uses to calculate relevance emulate the observations and judgments we make based on our experience.  A search engine will return the terms in our book index as hits when the search terms apple and grape are requested whereas a human might judge that although the two terms occur within the document, there is no significant relationship between them and is hence irrelevant. 

Some search engines look only in certain fields to index documents such as the title field, first paragraph and in something called "meta-tags."  Meta-tags allow the creator of a web site to add descriptive keywords which are not displayed in the actual web documents; they are specifically to enhance retrieval of the document. As people "spam" the search engine (for example, by repeating terms over and over again) meta-tags are decreasing in importance because the folks that program the 'bots train them to overlook repetitions and other clues to "spamming." 
 
Note:
Because each search engine assigns relevancy rankings differently, if you execute exactly the same search in several search engines you will have different results in terms of how and where the url's are listed (even if the database contents are identical).

 

What's the best search engine?

This is going to disappoint a lot of folks by giving the answer "the best search engine is the one that fits the task" instead of recommending a particular utility.  Until you have some experience with knowledge seeking tools and importantly, with identifying your real information need (for example, a query on "Leonardo di Vinci's Mona Lisa" is likely to be more successful than "that lady with the smile by a Renaissance artist" ( or simply "di Vinci") or "dosage and usage guidelines for St. John's Wort" as opposed to "St. John's Wort") it may be difficult to ascertain which tool is best for your purpose. But the good news is, you will make better choices with experience. 

 

What are simple ways to make my search more effective?

A very effective way to increase the relevance or precision of "hits" is to search as a phrase. In most cases it simply means putting quotation marks around the search terms. "Red Fox" is a different search than red fox in most search engines. What you are actually doing by searching as a phrase is using the concept of proximity which concerns the terms' physical closeness to one another (that is, their proximity).  A document with red fox occurring close or next to each other are more likely to be on target than a document with red in the title and fox buried in the text. 

Another way to increase your search effectiveness is to be as specific as possible; that is including as many terms and synonyms as you can think of to fully describe your topic. Instead of 

women and computers
try 
(woman or women) and (technology or computer) and (training or professional development) and (barriers or problems)

Note: search utilities may not support the use of parentheses (called nesting) in basic searches although many support them in their "advanced" searches. 

So to recap, phrase searching and specificity are two simple ways to increase precision in searching.

 

What are the most popular and useful search utilities? (the "major" search engines)

Ok folks. We are looking at a sampling of search engines and describing generalities; we are not attempting to create a definitive listing. For example, we'll be discussing meta search engines in Chapter 7, so you won't find them listed here.
 
  • Alta Vista (http://www.altavista.com)
    Originally  developed by Digital Equipment Corporation, Alta Vista searches the Web and Usenet. 
    In its very large database, both simple and advanced searching are supported with the ability to limit searches to select portions of web documents. For example, it is possible to limit searches to title, domains, images and links within Web documents and by particular newsgroups or subjects in Usenet. Also, ability to browse by subject (although this is rather slow). 
     
  • Excite(http://www.excite.com/)
    Search site featuring a very large database and a lot of "extras" such as: Excite Channels (guide to sites by subject), stock quotes, news, tv and searching of  Newsgroups. Offers concept searching. 
     
  • HotBot (http://www.hotbot.com/)
    Voted no. 1 among search engines by PC Magazine, Hot Bot offers a sophisticated interface with a vast array of options such as: searching by dates, by certain domains in the U.S. (e.g. .com, .org, .edu, .gov), by media type (e.g. image, audio, video). Also, a huge database, powerful advanced searching options, access to other search tools by type and a subject guide.
 There are more "major search engines" for you to evaluate in Assignments

 

Specialized Search Engines and Collections:

Specialized search engines are most often programmed to "collect" web documents along a topical theme. For example, in the Arts, Science, Health-related topics or even more specialized subjects such as Ancient History of the Mediterranean. 

Also fitting in this category are "search tools" that really calculate rather than retrieve information (such as those fitting in the "distance between two points" or "salary differential" categories). Since it is impossible to list specific tools here, the following are sites which group or list subject specific search engines or tools: 

  • Beaucoup
    (http://www.beaucoup.com)
    Beaucoup is a collection of approximately 1000 search engines, directories and indices from all over the world, organized into categories such as: General Searchers, Reviewed Sites/What's New, Software, Reference, Education, Art/Graphics, Social/Environmental/Political Concerns, and Consumer Medicine. Good starting point for popular subjects. 
  • Internet Search Engine Collections
    (http://library.albany.edu/internet/engines.html#collections)
    from the University of Albany by Laura Cohen.