Command-Line Prediction Interface Using Prediction Rules Based on History

Abstract

In this paper, we propose and implement a user support system for UNIX interface based on prediction of command-line. The proposed system can predict a command and the arguments independently using prediction rules extracted from the user's command-line history in real time. In order to predict next command-line, the proposed system pays attention to contexts of the history and relations depending on command-to-arguments. This system needs no knowledge about commands in advance, because it can extract the prediction rules from the history. The prediction process is implemented as the repetition of the following phases: input phase, prediction phase, computation phase, and updating phase. By repeating the process, the system can extract rules for the user and can adapt itself to various situations. We have compared the proposed system with FRQ and LRU method using actual command-line history, and obtained excellent results.