Analyzing the structure of complex sentences, comprising multiple clauses within a single sentence, has been recognized as a challenging aspect of dependency parsing. This challenge arises from the syntactic ambiguity inherent in clausal relations, wherein the main verb of each clause may have multiple candidate dependency heads. Minami's Scope Preference Theory (1964, 1974) hypothesized that clausal relations are determined by preferences between pairs of subordinate clauses. Building upon this theoretical foundation, we introduce a neural model fine-tuned to resolve clausal relations, with a focus on the verbs within each clause. To address the challenges in dependency parsing, we propose a forest reranking approach that enables our reranking model to grasp global context. Our approach utilizes our model as a reranker on dependency forests by leveraging cube-pruning for efficient tree enumeration. Through a series of experiments and analyses conducted on the Penn Treebank (PTB) and Penn Chinese Treebank (CTB), we find that our approach is effective in resolving the ambiguity inherent in complex sentence structures.