2007/05/01
Today, I modified my scoring function for getting high precision. In this function, I set a dynamic threshold for filtering unsuitable candidate terms. The results can achieved not bad precision but low recall. Because, the technical terms sometimes occur only once in the paper. For example, we implemented our system in Java. Hence, it is a little difficult to find the Java through my function.
Afterwards, using the Auto-Class for clustering again.
No comments:
Post a Comment