[目的/意义] 发现海量科技文献中的研究问题及其对应的研究方法，有助于挖掘科学研究中的热点，促进技术方法的创新，探索知识的演化传播规律。[方法/过程] 提出一种融合句法结构和语义关联信息的科技文献“问题—方法”联合抽取模型，模型采用编码器—解码器结构。在编码层，以科技文献的摘要文本为对象，从中抽取SAO三元组句法结构用以表达研究问题和研究方法的关系（即：研究方法—作用于—研究问题），基于SAO三元组构造语义关联图并利用图注意力网络进行编码，再与摘要文本编码融合作为解码器的输入特征；在解码层，通过指针网络基于先抽取的研究方法再抽取研究问题，实现“问题—方法”的联合抽取。[结果/结论] 实验结果表明，模型在测评指标及人工测评中均能取得较好的效果，能够提升从科技文献中抽取核心问题和核心方法的能力。

[Purpose/Significance] Discovering research questions and methods from a vast corpus of scientific literature contributes to uncovering research trends, promoting innovation technical approaches, and exploring patterns of knowledge evolution and dissemination in scientific research. [Method/Process] This paper proposed an integrated model for the joint extraction of “Problem-Method” pairs in scientific literature, combining syntactic structural information and semantic relationships. The model employed an encoder-decoder architecture. At the encoding stage, we focused on the abstract text of scientific literature, extracting Subject-Action-Object (SAO) triplets to represent the relationship between research questions and research methods (i.e., research method - acts on - research question). We constructed a semantic association graph based on SAO triplets and utilized Graph Attention Neural Networks (GAT) for encoding. The resulting encoding, combined with the abstract text, serves as input features for the decoder. At the decoding stage, a pointer network was used to extract research questions based on previously extracted research methods, enabling the joint extraction of “Problem-Method” pairs. [Results/Conclusion] Experiments indicate that our model performs well in terms of evaluation metrics and human assessment, enhancing the ability to extract core research questions and methods from the scientific literature.