Novel and Applied Algorithms in a Search Engine for Java Code Snippets - Finding Source Code on the Web for Remix and Reuse - page 279

14.5 Retrieving and Presenting Search Results
Juicy has both a general and an advanced user interface, as seen in Fig. 14.4 .In
this section, we will discuss how we use the indexes to obtain matches to the user's
query, rank them, and display them.
14.5.1 Matching and Ranking
The basic user interface consisted of a simple text box, just like a typical search
engine. With the indexes and metadata available, it was necessary to find a combi-
nation that would provide a set of useful results. Furthermore, when using multiple
search indexes, we would need to find a way to combine the results.
We found that matching the search terms to all the metadata fields in our in-
dexes at the same time produced many irrelevant results, but did so quickly. After
experimenting with different combinations, we found that using three of our indexes
produced the largest proportion of good results early in the list. These indexes were
page title, keywords from code snippet, and keywords from text. Page title gives a
general concept of the whole page. Keywords from code snippet contained identifier
names that could explain what the code snippet does. Finally, keywords from text
above code snippets, which proved to be relavant to the code snippet, could also
Fig. 14.4 User interface of Juicy