View Categorization Decisions
For each story the system classifies, a user can view the exact path taken through the category tree. A detailed report explains why some categories were chosen as matches while others were not, and why certain subtrees of the taxonomy were traversed or ruled out.
In addition, the software can mark words and phrases in the story that contributed to its being categorized into a specific category. An information expert can examine the highlighted words and see if they are being given too much or too little priority. He can also identify important words that were not picked up by the software at all. Using a different tool, a user can then modify the list of features used by the category classifier, adding or removing words or changing the weight of discriminating features (words or phrases typical to a certain category).
Optimize Training Documents
Training documents are the basis of automatic categorization: The software uses these examples to learn the concept of each category. GammaWare can help information experts find good examples, by providing input on the quality of training documents (or in other words, how closely they reflect the concept of the category).
Optimize Your Taxonomy
GammaWare help information experts analyze and optimize your taxonomy. The software examines the taxonomy and reports on its functionality. For example, GammaWare can detect related categories that should be connected by a semantic link, or overlapping categories that should be united.
In addition, the software can generate reports detailing how the taxonomy's structure affects the route stories take through the taxonomy tree, as they are categorized. For example, GammaWare can provide early warning that stories are welling up at a high level category instead of penetrating to more specific subcategories, due to non-optimal taxonomy structure.
