This paper presents a framework for extracting contexts and answers from online forum discussions, aiming to create a coherent summary and a valuable QA knowledge base. The proposed approach utilizes Conditional Random Fields (CRFs) to detect relevant information from forum threads. Enhancements to the basic framework include Skip-chain CRFs and 2D CRFs, which are designed to better handle the specific features of forum data. Experimental results demonstrate that these techniques offer promising improvements in performance for context and answer extraction.