Corpus of Annotated Graphs
- A corpus of graphs (simple bar charts, grouped bar charts, and line graphs) have been collected from popular media, along with the articles in which they appear. The graphs have been annotated with their intended message, the paragraph in the article that is deemed most relevant to the graph, what is being measured in the graph, keywords to aid in collecting similar graphs, etc. These graphs will be used as a testbed for extracting graphs from a digital library in response to user queries.
First Collection of Queries
Second Collection of Queries
- Two initial experiments were performed to collect a set of queries for information
graphics. In the first experiment, subjects were presented with a graphic and
asked to write a query for which the graphic might be retrieved. They were
then presented with a
brief description of a domain, asked to write
a query for an information graphic that might be available for that domain,
and then draw an information graphic that would satisfy their query.
In the second experiment,
subjects were presented with a set of four graphs on a particular topic;
for each graph, the subject was asked to write a query that would be
best answered by that graph as opposed to the other three graphs.
-
A final experiment was performed using Mechanical Turk to collect queries on the domains of global economics and higher education. Turkers were asked to write full-sentence interrogative queries on the given domain.
Global Economics Query Set
Higher Education Query Set