DeSCoVeR: Debiased Semantic Context Prior for Venue Recommendation

Abstract

We present a novel semantic context prior-based venue recommendation system that uses only the title and the abstract of a paper. Based on the intuition that the text in the title and abstract have both semantic and syntactic components, we demonstrate that joint training of a semantic feature extractor and syntactic feature extractor collaboratively leverages meaningful information that helps in recommending venues for paper publication. The proposed methodology that we call DeSCoVeR first elicits these semantic and syntactic features using a Neural Topic Model and text classifier, respectively. The model then executes a transfer learning optimization procedure to perform a contextual transfer between the feature distributions of the Neural Topic Model and the text classifier during the training phase. DeSCoVeR also mitigates the document-level label bias using a Causal back-door path criterion and a sentence-level keyword bias removal technique. Experiments on the DBLP dataset show that DeSCoVeR outperforms the stateof-the-art methods.

A survey on causal inference for recommendation

“…A multitude of studies employ backdoor adjustment to block the backdoor path by directly intervening on the treatment variable. 19 , 65 , 69 , 135 , 136 , 137 , 138 , 139 , 140 , 141 For example, Wang et al. 69 propose the framework called DecRS (deconfounded recommender system) to eliminate bias amplification through intervention on the user representation , which removes the effect of the historical user distribution over item groups on , as Figure 6 C shows.…”

Research on feature extraction methods of academic papers

“…Taking library science and information science as examples, Wang et al [8] used hierarchical clustering to build a two-layer architecture, compared the effects of three classification methods such as SVM, CNN and RNN under different feature combinations through experiments, and selected the most suitable classification algorithm. Rajanala et al [9] extracted semantic and syntactic features from titles and abstracts by using neural topic models and text classifiers, and adopted joint training to optimize the transfer learning process to extract richer semantic and syntactic features. Sethares et al [10] proposed a paper classification method based on SVD, using SVD to extract the features that best represent the class of the paper.…”