April 8, 2009 at 14:14Researching the goals of people searching
Daniel Rose (Daniel E. Rose), who worked in AltaVista and Yahoo, investigated the goals of users who encourage them to search in search engines. Together with Danny Levinson in his Understanding User Goals in Web Search, he described the research process and the results.
First, a classification of user goals was developed based on the search query log AltaVista. All requests (with the exception of specific ones) should fall into any of these categories:
Rose says that search engines must take into account the true goals that guide their users and improve their interface accordingly. Unfortunately, automatically detecting these goals is difficult.
Under the cut is a brief description of Rose and Levinson's further work. I have a full version of their abstracts and other articles on human behavior and search on my site.
To take the first steps, the authors developed a special tool for manual classification of queries. They analyzed about 500 randomly selected English search phrases from the AltaVista server logs and manually assigned each of them to a certain category. Here are the results:
It is interesting to note that 40% of the requests were not informational at all, and most of the informational requests came down to finding a specific product or service, rather than finding out facts about it. This contradicts what search engines were originally created for - the search for information in the "electronic library" back in the pre-Internet era. In fact, only 35% of the requests were consistent with this idea.
It is worth noting that the keywords were sampled in several stages at different times and seasons. The distribution of results was similar in all cases.
Unfortunately, the authors do not give any practical recommendations on the application of these results. They say that they have created a framework, a framework with which further research can be conducted to improve the performance of search engines. Rose and Levinson offer readers to try this framework, develop its ideas and highlight practical benefits for themselves.
First, a classification of user goals was developed based on the search query log AltaVista. All requests (with the exception of specific ones) should fall into any of these categories:
Rose says that search engines must take into account the true goals that guide their users and improve their interface accordingly. Unfortunately, automatically detecting these goals is difficult.
Under the cut is a brief description of Rose and Levinson's further work. I have a full version of their abstracts and other articles on human behavior and search on my site.
To take the first steps, the authors developed a special tool for manual classification of queries. They analyzed about 500 randomly selected English search phrases from the AltaVista server logs and manually assigned each of them to a certain category. Here are the results:
It is interesting to note that 40% of the requests were not informational at all, and most of the informational requests came down to finding a specific product or service, rather than finding out facts about it. This contradicts what search engines were originally created for - the search for information in the "electronic library" back in the pre-Internet era. In fact, only 35% of the requests were consistent with this idea.
It is worth noting that the keywords were sampled in several stages at different times and seasons. The distribution of results was similar in all cases.
Unfortunately, the authors do not give any practical recommendations on the application of these results. They say that they have created a framework, a framework with which further research can be conducted to improve the performance of search engines. Rose and Levinson offer readers to try this framework, develop its ideas and highlight practical benefits for themselves.