Help Questions

• How does the URL search work?
• How does the keyword search work? Is it different from Google and Yahoo?
• What is "Surprise Me"? Is it different from "General Search"?
• What are bookmark similarity, novelty, and prestige?

How does the URL search work?

At GiveALink, we collect bookmark files from people willing to donate them. These bookmark files reveal patterns and related sites that are identified using a novel semantic similarity measure. Our measure exploits the hierarchical structure of bookmark files. For example, if many users put two URLs in the same folder in their bookmark files, then the two URLs are probably very similar. However if the two URLs are always in different folders, then they are probably not related.

How does the keyword search work? Is it different from Google and Yahoo?

Our keyword search looks a lot like Google and Yahoo: you can enter words describing the subject you are searching for and we will suggest relevant sites! However, instead of crawling the Web to find pages containing the search words, we match search words against titles and descriptions in our bookmark collection. We only recommend sites from our bookmark collection--sites that a human has bookmarked as relevant to what you are looking for!

What is "Surprise Me"? Is it different from "General" search?

Sometimes sites can be related to each other, but not in an obvious way. For example, in the Simpsons, Bart and Moe are not directly related to each other, but they are similar in a way because they both spend a lot of time with Homer. Our "Suggest Novel URLs" attempts to identify sites with connections that may not be obvious.

"Surprise Me" keyword search is very similar to "Surprise Me" URL search: it suggests sites that are relevant to your search words but in a way that is unexpected and not obvious. The difference is that the first expects you to enter search words while the latter expects you to enter a valid URL.

What are bookmark similarity, novelty, and prestige?

You can sort your search results using three measures: bookmark similarity, novelty, and prestige. You can also sort by the product of any combination of the measures. For example, if you select novelty and prestige, the top result will be one with high (novelty * prestige) rank, i.e. a site that has high novelty and high prestige.

• Similarity

  Bookmark similarity is a novel semantic similarity measure that exploits the hierarchical structure of bookmark files. We use Lin's measure to calculate the similarity of two URLs in the bookmark collection of one user. To obtain a similarity value for one pair of URLs, we sum up the similarity that our users report.

• Novelty

  The novelty of a recommended site is the ratio of indirect similarity over direct similarity to the search URL. For example, in the Simpsons, Bart and Moe are not directly related to each other, but they are similar because they both spend time with Homer. Thus if you are searching with keyword "Moe", the recommendation "Bart" will have high novelty because the two have high indirect similarity and low direct similarity.

• Prestige

  Prestige is a global ranking measure that is independant of the query URL and keywords. It generates scores for the URLs in the GiveALink database similar to the way PageRank does. In simple terms, it identifies Web pages that are central to out database, that a lot of users donated, and that have high average similarity to other bookmarked Web pages. In mathematical language, prestige orders URLs according to their corresponding values in the principle eigenvector of the similarity matrix.