How to Design a Search Engine to Actually Search

Many search engines force you to sort through dozens of useless results, SPAM and pop-up ads in order to find the web pages that match your original query. So, if you've had enough of wading through too much unwanted online info, maybe it's time to develop your own powerful search engine.

Difficulty: Challenging

Instructions

1. Split the page into related sections. Searching through all the text on a page is cumbersome and time consuming. If paragraph searching proves difficult, create a limited search bracket (30 words.)

2. Stop connecting searches to advertisers. Many search engines front-load marginally related pages to suit advertisers. This lowers the number of relevant matches.

3. Search literal words and characters. Automatically screening out articles and prepositions insults user intelligence and increases potentially useless matches. You can provide the user with the option to screen words if they want.

4. Give users credit; don't underestimate their intelligence. People are smart enough to advance beyond today's low-tech applications.

5. Piggyback searches. Once pages are found focus the search on pinpointing the section with the search text. This allows the user to skip past ads and unrelated sections.

6. Build from prototypes. Start with a simple prototype that converts lines of text into words to match the search parameters. Build from that to a program that keeps track of matches across multiple lines.

Tips & Warnings

• Don't use current search engines as your model. The objective shouldn't be to find as many possible pages as possible, but to pinpoint specific information.
• Work with "NEAR" keyword. Experiments with NEAR can show off the advantages of searching paragraphs over pages.
• Stop SPAM piggybacking. Low-tech search engines that look through entire pages promote spam pages using search strings to match pages with thousands of unrelated words.