Abstract
The purpose of this study is to acquire the required knowledge resulting from converting Arabic speech into text. This system passes through some stages. The initial stage is "system training." This stage starts with identifying the lecturer's sound in different frequencies and then saving it upon an audio file and then converting it from analogue waves into digital ones; this is called "the preprocessing path." The properties of this audio file are then extracted and stored within a template. The second stage is "system testing." This stage compares the new sound file with the one previously stored on the system directory by comparing certain features. When the sound characteristics are matched using a similarity scale, the selected sound file has been identified. The third stage is printing the words into a text document. The next stage revises the resulting words according to specific rules identified by the proposed system itself. The final stage extracts the knowledge from the revised text. The proposed system is evaluated using Arabic dataset; it contains Arabic speech collected from different person types and ages. The experiment tests the system under three different recording modes, namely normal, quick, and loud. The results showed that the proposed system is effective in acquiring knowledge from converted Arabic speech to text especially with normal recording mode in Arabic speech recognition and classifying the fields.