第三代定序技術速度較快、序列較長,但錯誤率較高,因此進行基因體組裝前,須先進行錯誤校正工作。我們所開發的第三代定序錯誤校正軟體 (FILEC),和目前最好的錯誤校正軟體 (Canu) 相比,雖然錯誤校正速度較快,且組裝完整度較高,但校正準確率卻有所下降。因此在本論文研究中,我們首先觀察容易發生校正錯誤的區域,通常位於低覆蓋率之重複序列區域 (Repeat Regions) 以及串聯重複序列 (Tandem Repeats)。我們設計出方法能識別這些區域,並針對這些區域特性改良原有錯誤校正演算法。實驗結果指出調整原本非序列比對校正法 (Alignment-free Correction) 能些許改善正確性。但意外發現傳統序列比對校正法 (Alignment-based Correction) 能大幅改善正確性。因此我們結果顯示若要同時改善校正速度與準確性,或許必須均衡二種錯誤校正法之使用時機。 Third-generation sequencing technologies are able to generate longer reads within shorter turnaround time, but they come at the cost of higher sequencing error rates. Therefore, prior to genome assembly, error correction is required to reduce the errors presented in the sequencing reads. The error correction and assembly software that we developed (called FILEC) has improved the speed and contiguity of a leading genome assembler called Canu; however, the assembly accuracy of FILEC is lower than that of Canu. In this thesis, we first investigated the regions FILEC tend to wrongly corrected, and observed that they are regions containing low-coverage repeats and tandem repeats. Subsequently, we develop new methods for identifying and for improving the correction algorithms specifically for these regions. The experimental results indicated that the accuracies can be slightly improved by improving the original alignment-free correction algorithm. But surprisingly, the accuracies can be greatly improved by the slower alignment-based correction using dynamic programming. Our results imply a good balance of alignment-free and alignment-based correction algorithms is crucial for improving both assembly speed and accuracy.