About Topper

For more information on Topper please read our white paper: topper.pdf

Topper FAQ

What is Topper Search?
Why do I want to personalize search results?
How does Topper personalize search results?
Can I use someone else's preferences to sort my results?
What about privacy?
What do the positive/negative buttons do?
What do the 'Sort by Rank' and 'Sort by Score' buttons do?
How does it work?
What features about the search result does Topper use?
Isn't this just Active Learning?
What do the numbers to the left and right of the result mean?
After I reloaded the page my 'hat' seems to have disappeared?
Why did I get a 403 error?
Can Topper be applied to anything other search results (e.g. blogs, Digg, Reddit)?
How can I get involved in the this or another Topper application?
  • What is Topper Search?
    Topper is a personalization tool that actually learns your preferences. Topper Search is a personalization of web search. Topper learns what you like from a single search and then applies what it learned to any search that you make. Topper is not a web search engine but rather a tool that sits on top of search results. This version currently sits on top of Yahoo Search results. Topper can be easily be adapted for Google or any other search engine.

  • Why do I want to personalize search results?
    Search engines work well when they return the most popular results to a large general audience. Many queries, however, are ambiguous in intent. As a result the most popular results will likely cover a large range of topics that suit the variety of user intents. For example the query 'lions' will typically return sites related to both the Detroit Lions and a large member of the cat family. In order to find what you really want search engines suggest you further qualify your query such as 'detroit lions' or 'african lions'. Wouldn't it be nice if the search engine knew that you were 'sports fan' and filtered the results to your personal taste? We can take this one step further. What if you could tell the search engine at different times that you were either a 'sports fan' or 'looking for information for a science report'? Topper Search was developed with this idea that people should be able to wear a Topper 'hat' or mental model that expressed their interests at any given time. With Topper you can build a 'sports fan' hat or ' looking for information for a science report' hat that will help you filter search results.

  • How does Topper personalize search results?
    Topper Search allows you to create or use an existing Topper 'hat' to sort search results to your taste. Creating a Topper hat is simple - simply click on the results you like and don't like, watch Topper score your results. These results can then be sorted using a blend of your 'hat' score and the original rank of the result. As you click each result Topper automatically builds a new mental model for you and re-scores all of the results. Multiple hats can be created, modified and reused that represent your different tastes.

  • Can I use someone else's preferences to sort my results?
    Of course. That's really the whole point. Topper allows people with expertise in some area to easily share their knowledge with others. For example an expert mechanic can create a hat that helps evaluate relevant topics, products and sites for buying a car. This hat and knowledge can then be shared transparently with anyone such that searches are seen through the eyes of an expert mechanic. With Topper you can search through the eyes of anyone that is willing to share his or her hat.

  • What about privacy?
    Topper does all of its learning right there in your browser. No communication with any backend servers is necessary to build a Topper hat. No software installation is necessary either. Create Topper hats to sort your results and no information will ever recorded about your personal tastes. You can even use existing Topper hats, created by other users just like you, without any concern that your actions are recorded. On the other hand if you want to reuse or share your hat you can save the model on our servers for future use.

  • What do the positive/negative buttons do?
    These buttons allow you to pick which results you like or don't like. As soon as you click at least one positive and negative example Topper goes to work building a hat. Once the hat is built it is used to classify or score all of the results on your page. Any other searches that you perform will also be scored by the same hat. You can add or remove examples from hat using the same buttons to further improve and refine your model.

  • What do the 'Sort by Rank' and 'Sort by Score' buttons do?
    These buttons allow you to sort your results using a blend of both the original search ranking and the Topper score. If the button on the far left is clicked the results will be sorted according to the original search rank. If you click the button on the far right the results will be sorted solely by the Topper score.

  • How does it work?
    Topper is a machine-learning library written entirely in Javascript. Topper takes features from search results that you have identified and learns, right there in your browser, how to distinguish what you like and don't like. Topper uses this machine-learned model or 'hat' to score results that you have not seen. Because Topper learns from you, the more you add to the model the more capable it becomes.

  • What features about the search result does Topper use?
    Topper uses the text contained in the URL, title and summary of the search result. Future versions may use other data as well such as the text from the site that the URL points to as well as other data.

  • Isn't this just Active Learning?
    Yes, Topper uses active learning which is a form of supervised machine learning. Topper builds a new model as each example (i.e. result) is added or removed from the training set. If you sort the results according to Topper score you will be able to easily see where the model is making mistakes. By sorting, both normal and reverse, you will be able to label the small set of examples that improve your model most.

  • What do the numbers to the left and right of the result mean?
    The number on the left is the original search rank. The number on the right is the Topper score.

  • After I reloaded the page my 'hat' seems to have disappeared?
    This is intended. Reloading a page wipes out your current Topper hat unless you saved it. Topper hats are only built locally in your web browser for privacy reasons. We will only record your preference information if you explicitly save your Topper hat.

  • Why did I get a 403 error?
    Currently Topper Search runs on top of the Yahoo Search service which is rate limited. When a rate limit has been exceeded, the HTTP request will receive an error code of 403. We can increase the limit if necessary. Alternatively we are working to build Topper on top of other web services in search and other areas of interest.

  • Can Topper be applied to anything other search results?
    Yes. Topper is designed to work with any kind of list. Topper provides powerful learning algorithms to anywhere your web browser can go. Topper does not require any special software or extensions to download and install. We are currently working on Topper for blogs and newsfeeds. If you are interested in putting Topper on top of your lists please contact us at info@toppersearch.com

  • How can I get involved in the this or another Topper application?
    If you are interested in Topper Search, Topper Blogs or you own idea about where to apply Topper please contact us at info@toppersearch.com


  • Copyright © 2008 Topper. All Rights Reserved