Encoding Hierarchical Schema Via Concept Flow for Multifaceted Ideology Detection

Multifaceted ideology detection (MID) aims to detect the ideological leaningsof texts towards multiple facets. Previous studies on ideology detection mainlyfocus on one generic facet and ignore label semantics and explanatorydescriptions of ideologies, which are a kind of instructive information andreveal the specific concepts of ideologies. In this paper, we develop a novelconcept semantics-enhanced framework for the MID task. Specifically, we proposea bidirectional iterative concept flow (BICo) method to encode multifacetedideologies. BICo enables the concepts to flow across levels of the schema treeand enriches concept representations with multi-granularity semantics.Furthermore, we explore concept attentive matching and concept-guidedcontrastive learning strategies to guide the model to capture ideology featureswith the learned concept semantics. Extensive experiments on the benchmarkdataset show that our approach achieves state-of-the-art performance in MID,including in the cross-topic scenario.