反覆練習式（drill - and - practice）CAI是現有各種不同類型CAI中最初普遍普遍應用的技術；然而，傳統反覆練習式CAI軟體受限於選題策略之影響，使得軟體本身無法依據練習者的能力與進步情形做適當的進度調整，造成練習效率不佳。
適性練習系統（Adaptive Drill System, ADS）結合了現代測量理論一項目反應理論（item response theory, IRT）一與類神經網路一倒傳式模型（backpropagation model）一而形成一個全新的架構。在它的兩階段架構中，首先以電腦化適性測驗（computerized adaptive testing, GAT）程序來估計練習者的能力值，再由題庫中選取適合該能力值的目提供練習。而在練習的階段中，以類神經網路技術製作的精熟練習決定器（mastery decision processor, MDP）會預估練習者的未來能力值，據以判斷練習是否應該停止。
本研究所提出的系統架構可適應不同練習者的能力與個別進步情形，並提供適當的練習題目，可獲得較傳統反覆練習式CAI更好的練習效率，為反覆練習式CAI找到一個新的發展方向。

Compared with the current coursewares, the drill and - practice CAI is no doubt the most popualr and fashionable one; however, it is still not satisfactory because of the inefficient performace of item selection proced. Regardless of the discrepancy of the practicer's ability, general item selection procedures always provide same items to them.
Adaptive drill system (ADS) provides a new system architecture that associates with the modern measureement theory, i. e., item response theory (IRT), and artificial neural networks, i. e., backpropagation model. In the two - phase architescture, it first measures practicers' ability level, which is a computerized adaptive testing (CAT) based provedure, and then selects items from the calibrated item bank for practicing based on the measured ability. During practice phase, the mastery decision processor, which is performed by artificial neural network, predicts practicers' future ability to determine whether it should be terminated.
The proposed architecture is not only more efficient than the traditional drill - and - pracitce CAI, but also provide a new direction for developing drill programs.