|
. |
|
|
|
|
A query starts with selecting a Data Set. The current options are the CiteSeer 300 topics or the NIPs 10 topic. Then one has to select one out of the possible Selections: Author, Topic, Document, Word or Word Vector. Here we show and example of choosing Author. We typed the Name Pazzani_M (see up at right). The result - two tables. A table with the high ranked topics given that author, along with the probability for each topic given the author, and a list of all documents in the data set that belong to that author (as shown on #documents Pazzani_M has 70 documents). Mouse-clicking on one of the topics (e.g., the data mining topic as shown in the figure) produces the screen display to the left. The most likely words for this topic and the most likely authors given a word from this topic are then displayed. We have found this to be a useful technique for interactively exploring topics and authors, e.g., which authors are active in a particular research area.
Similarly, one can click on a particular paper (e.g., the paper “A Learning Agent for Wireless News Access”) as shown in the lower screenshot and the display in the panel to the right is then produced. This display shows the words in the documents and their counts, the probability distribution over topics for the paper given the word counts (ranked by highest probability first), and a probability distribution over authors, based on the proportion of words assigned by the model to each topic and author respectively. |
