這是資訊爆炸的時代,幾乎任何資訊都可透過網路取得,因此近期網頁資料擷取領域,相繼有許多研究提出各種的非監督式擷取方法,能夠快速有效擷取資料進行後續加值應用。但是網路環境快速變動的特定,也讓非監督式擷取程式面臨挑戰。非監督式擷取程式在產生擷取規則時,需要相當複雜計算,需要耗費相當的時間,不可能每次擷取資料都重新產生新的擷取規則,因此擷取程式後續驗證與維護機制必然越來越重要。擷取程式會解析網頁內容,產生網頁模板(Template)與資料結構(Schema)來擷取資料。擷取程式驗證機制主要目的是確保時間點t產生的模板(Template)與資料結構結構(Schema)仍能適用於時間點t’的資料擷取。在網頁XML DOM樹狀結構中,資料內容是存在於葉節點位置,因此可運用有限狀態機模型檢測網頁葉節點轉換規律是否與時間點t的網頁模板(Template)與資料結構(Schema)一致。本論文嘗試簡化建立有限狀態機模型過程,使其更為快速有效,並以CSP快速刪減候選狀態節點的數量,加速模型驗證效率,並搭配狀態的序列組合(Sequence Probability)來檢測網頁模板(Template)與資料結構結構(Schema)有效性,以實驗測試其效率及有效性。Wrapper induction is a complex process that takes a considerable amount of time. However, data extraction that requires constant wrapper induction is inefficient for a more sophisticated Web site design. Therefore, wrapper verification and its maintenance are becoming two of a few major subjects for research. This paper focuses on wrapper verification of unsupervised information extraction. The researcher uses the leaf nodes of the dom tree as the source of transformation and obtains a FSM (Finite State Machine) for schema verification. If the new page could pass the verification, it would be considered a simultaneous data layout.This paper attempts to simplify the process of creating the finite state machine model, and proposes Schema Guided Wrapper Verification based on CSP (Constraint Satisfaction Problems) to reduce the number of candidate states to accelerate validation efficiency. The approach this study proposes not only improves validation efficiency, but also finds a better Best State Sequence to improve the accuracy of data extraction.