Machine Learning Helps Managing Documents

Keywords are often used to structure a collection of documents. Keywords help humans to search and retrieve records from the collection. Maintaining such a collection involves assigning keywords to new records and updating keywords of old records in the collection. It becomes difficult to update a collection with increasing number of records and keywords. We propose a computer-supported approach to maintain a collection. The approach suggests keywords for individual records using previous keyword assignments. We provide visual explanations to make the suggestions transparent. We implemented the proposed technique for literature collections integrated in an existing literature collection browser.

Please find more information in the paper, which will be published at IEEE VIS 2018 workshop "Machine Learning from User Interaction for Visualization and Analytics".