Flamenco is part of the interesting BAILANDO initiative, which we talked about in issue 80. Its overall goal is simplifying the access to information. 

Flamenco is devoted to obtaining a user interface that allows the user to explore, search and understand the contents of large collections of information organised (or able to be organised) in hierarchical categories. In other words a big chunk of what is on Internet . 

Lead by Marti Hearst, professor of the University of California in Berkeley and specialist in user interfaces and visualisation for information retrieval, Flamenco tries to answer two specific questions:

• “Can dynamically suggested query previews improve comprehension and findability within domain-specific text and image collections?” 
  
  
• “How can hierarchical, faceted categorical metadata be used effectively within such query previews?”

What lies behind these somewhat cryptic questions is the following idea: maybe the search will be more effective if the user has a flexible way to pre-visualise the results of a query in a way that he/she could refine based not only on querying the database but also by following a list of links “built on the fly” organised according to a metadata hierarchy (dynamically suggested queries).

In order to answer these questions and to test the idea, Flamenco relies on usability studies and user centred design.

Let’s see how it works. An interesting example is the one in the article “Finding the Flow in Web Site Search” (PDF). We’ll make our own search on the architect images database of UC Berkeley in order to find ideas on the use of water for a hypothetical new building. 

We enter the database. What we find first is the typical search rectangle along with a series of links organised hierarchically in different categories grouped in several facets: people, locations, styles, structure types…

From here we can proceed basically in two ways: following the links that lead to predefined query pre-visualisations or issuing new queries in the traditional way. Flamenco combines browsing and interrogation. So we enter “water” as a query and we get a page where we find, on one hand, categories where the word “water” appears and, on the other hand, the same categories seen in the previous page, along with a series of images that satisfy the query. Among them we find houses made of ice or building for the treatment of water.

Click on the image to enlarge it		Click on the image to enlarge it
	FLAMENCO
Click on the image to enlarge it		Three screenshots of the user interface: Upper left: The main page. Combines a conventional search with a view of many predefined queries organised hierarchically into metadata categories al so called "facets". Each facet has many categories that link to particular queries. Upper right: introducing a textual query leads you to a "matrix view", with the matches to your query, some examples of them, the facets and categories that still apply to your query and a scheme of similar predefined queries. Lower left: following a link contained in a facet leads you to a similar matrix view, but it lacks the scheme of similar queries. For a complete understanding of the features see Flamenco's user guide.

At this point we can decide to follow the links or refine the query, possibly restricted within what has already been found. Let’s follow, for example the “waterfalls” link to obtain a new page with buildings and gardens associated to waterfalls. An image is particularly spectacular, the town of Jajce built on the edge of a waterfall. 

Clicking on the image we get a file containing the (many) different ways you can follow to reach the image. The important issue: by following the links included in these paths you can reach similar or related images.

The interest of Flamenco relies on its versatility. A well defined query like “Casa Batllo by Antoni Gaudi” will give good results in many search engines. The interesting questions are those that are not well defined, where you don’t know precisely what you are searching for , where serendipity is a key point since we are looking for ideas.

 Here is where Flamenco seems to be most effective since it combines direct search and browsing using links in a scenario structured in categories and hierarchies where you can reach the same point following many different paths.

Although it’s still in the test phase and with a database that sometimes does not answer, the concept is a very interesting one and promises to provide a more solid interface for the hardworking Internet information searcher. 

Links of this issue: