Vertical Themes


What are Vertical Themes about?

(Original Text by Ammon Johns) In the quest to return more relevant results from an increasingly packed database many search engines are looking to vertical theme analysis to provide an answer.

At the simplest level themes are how a search engine spider can learn to categorize the subject of a page, section and website. By doing so, a search engine can return pages that match the theme of a query rather than just matching the words.

How can a theme be measured?

Well, the easiest way is to attach a basic lexicon, or thesaurus type database to the spider program. This lexicon would allow a spider to categorize related words under a topical heading.

For example when a page contains the words recipe, cooking, ingredients and oven, the spider may be able to determine that all of these words relate to the category of food. By finding a 'category' for all words (except stop-words) on the page the spider can make a list of possible subject categories for the page as a whole.

By just recording the top 40 or 50 categories that words on the page may apply to, the spider has pretty much mapped out the topics and subjects that the page most likely applies to. It has found the common 'themes' to the words on the page.

By matching page themes across multiple documents, or even across an entire domain, the theme(s) for a whole site can be determined and measured. In this way, search engines can more readily identify sites which are 'specialists' in a particular field and rank them higher than sites where a page simply happens to have the keywords in it.

A simplified example of theme analysis

Imagine that two very different pages, hosted on different sites, both have the words "Jamaican" and "cooking" in the text of the body.

The first page is a tale about the author's love of cricket and has an anecdote about a match played by the Jamaican cricket team which he watched while cooking his lunch one afternoon.

The second page discusses the favourite Jamaican recipies that the author enjoys cooking. It has depth of content about the ingredients and methods involved in cookery in that style.

A human can readily see, even from these short descriptions which page is likely to be the best match. But until now, the search engines were pretty much just counting the times that the words you enter are matched on various pages in the index.

However, with themes, the algorithm is analyzing each word on the page for its 'category' or theme. On the first page it finds more sport related words than any other and few food related words other than the few in that particular anecdote. The main theme of this page is determined to be "Sport".

On the second page, the vertical theme analysis algorithm easily relates the mention of "ingredients", "recipies" and many other words to the over-all category of "Food".

By applying the same 'theme analysis' to the search queries, the algorithm can already be looking for certain 'themes' when the query is made. A search for 'Jamaican cooking' provides possible themes of Travel / Geography / Ethnicity for the word Jamaican, and themes of Food and Drink for the word cooking. The fact that neither term directly relates to 'sport' will reduce the ranking of our first example page against pages with the same matched words but more relevant themes. Since the second page matches not only the words, but also the Food theme, it will rank highly and present a more relevant result to the search.

More about Themes >>

Back to the Help File / Tutorial Directory
Visit our Web Marketing Glossary


©Copyright 2000 - 2005, Association of Internet Marketing Professionals, Inc., All Rights Reserved.
All rights reserved.


Back to Top