Current search engines heavily emphasize on direct
querying which tends to work well only for simple information
needs such as navigational queries. However, direct
querying may not support complex information needs such as
exploratory search well [
We have been developing a new collaborative sur ng
system to enable users to go beyond hyperlinks to browse
exibly for ad hoc information needs. Our main idea is to view
search logs as information footprints left by users in
navigating in the information space and organize these
footprints into a multi-resolution topic map. The map makes
it possible for users to navigate exibly in the information
space by following the footprints left by other users. As
new users use the map for navigation, they leave more
footprints, which can then be used to enrich and re ne the map
dynamically and continuously for the bene t of future users.
Thus, by turning search logs into a topic map, we can
establish a sustainable infrastructure to facilitate users to surf
the information space in a collaborative manner.
Preliminary experiment results show that the topic map is e ective
in helping users to satisfy exploratory information needs [
Figure 1 shows the interface of our system which is implemented as a meta-search engine interacting with Google. The interface has three panes: (1) The top pane is a querying box, where a user can submit a keyword query. (2) The left pane shows a portion of a multi-resolution topic map built based on search logs, where a user can click on a node to navigate into a topic region. (3) The right pane displays information corresponding to a topic region, including the clickthroughs made by previous users when they visit the topic region and the documents covered by the topic region.
These three panes allow a user to navigate in the information space in large, medium, and small steps, respectively. With the query box on the top, a user can make a long distance navigation into any topic region (i.e., \large steps"); with the topic map on the left pane, the user can navigate into related topic regions (i.e., \medium steps"); and with the display of a topic region in the right pane, the user can navigate by following hyperlinks (i.e., \small steps"). A user can take any of these three navigation actions at any time. Thus our system implements a uni ed information seeking model where both querying and browsing are viewed as ways to navigate in the information space.
Inside the system, when a user submits a query, the system would display the most relevant part of the topic map on the left pane and show the search results from Google for the query. When the user navigates on the map to click on a node (corresponding to a topic region), the system would automatically update the right pane to show corresponding search results using a query constructed based on the node selected by the user. In general, the right pane is always synchronized with the left pane to show the documents corresponding to the current node on the map.
The topic map promotes browsing and can naturally support exploratory search. For example, a user who wants to arrange a house can start with a query \table," zoom into \dinning table," zoom out to \dinning,", move horizontally to \kitchen," and further move to \appliance." From \table", this user can also horizontally move to \chair," to \desk,", or to \tablecloth." Another example is \wedding." From \wedding," we can zoom into di erent aspects of wedding such as \wedding dress,' \wedding vows," etc. We can also horizontally move to \vacation," \honeymoon," or \hotels." All these browsing traces can be leveraged to infer users' underlying information needs and better serve users with complex exploratory information needs. The browsing logs can be leveraged to improve the map and further help future users who have similar information needs.
A main technical challenge in developing this system is to
construct topic maps. Currently, the nodes in topic maps
are valid queries in search logs. All queries with the same
number of keywords belong to the same level. The children
of a map node is obtained by adding a keyword into the
current query and the neighbors of the query is by substituting
a keyword in the current query. All these surrounding nodes
are ranked accordingly. Speci cally, we rely on the term
cooccurrence in search logs to construct such a map and all
the technique details can be found in [
From the viewpoint of understanding users and
exploiting user information to provide better search support, our
system implements a strategy of massive implicit feedback
[
We hope our demo can stimulate discussions about many interesting questions related to the workshop: (1) How should we evaluate topic maps? (2) How should we evaluate such an interactive system? (3) How can we formally model a user based on both query logs and browsing logs? (4) How can we leverage maps to clarify user interests?