Promotion of Answer Value Measurement With Domain Effects in Community Question Answering Systems

Abstract

In the area of community question answering (CQA), answer selection and answer ranking are two tasks which are applied to help users quickly access valuable answers. Existing solutions mainly exploit the syntactic or semantic correlation between a question and its related answers (Q&A), where the multifacet domain effects in CQA are still underexplored. In this paper, we propose a unified model, enhanced attentive recurrent neural network (EARNN), for both answer selection and answer ranking tasks by taking full advantages of both Q&A semantics and multifacet domain effects (i.e., topic effects and timeliness). Specifically, we develop a serialized long short-term memory to learn the unified representations of Q&A, where two attention mechanisms at either sentence level or word level are designed for capturing the deep effects of topics. Meanwhile, the emphasis of Q&A can be automatically distinguished. Furthermore, we design a time-sensitive ranking function to model the timeliness in CQA. To effectively train EARNN, a question-dependent pairwise learning strategy is also developed. Finally, we conduct extensive experiments on a realworld dataset from Quora. Experimental results validate the effectiveness and interpretability of our proposed EARNN model.

Learning the Compositional Visual Coherence for Complementary Recommendations

“…With the rapid development of deep learning and computer vision, visual recommendations have raised lots of interests in both academia and industry and benefited lots of applications, such as image recommendations [McAuley et al, 2015b], movie recommendations and fashion recommendations [Han et al, 2017;Hou et al, 2019]. Some previous works treated visual recommendations as special contentaware recommendations incorporating the visual appearance of the items .…”

“…Some previous works treated visual recommendations as special contentaware recommendations incorporating the visual appearance of the items . Along this line, many researchers directly extracted the visual feature by a pretrained CNN model and enhanced traditional recommender systems, such as Matrix Factorization Hou et al, 2019]. For example, proposed a recommendation framework to incorporate the visual signal of the items.…”

“…For example, proposed a recommendation framework to incorporate the visual signal of the items. Recently, many researchers utilized deep learning methods to generate item recommendations [Kang et al, 2017]. To further explore item visual contents, some studies developed deep neural networks to extract the aesthetic information of images Yu et al, 2018].…”

“…Recently, more and more studies based on supervised approaches were proposed to model complementary relationships, which were mainly reflected by users' purchases of the complements or item content similarities [McAuley et al, 2015a;Zhang et al, 2018]. Along this line, since natural complementary relationships in the fashion items Lo et al, 2019], there was a particular interest in understanding the compatibility [Song et al, 2017; of fashion items to generate complementary recommendations [Han et al, 2017;Hsiao and Grauman, 2018;Vasileva et al, 2018]. For example, learned the complicated and heterogeneous relationships between items and enhance fashion recommendations.…”

“…Meanwhile, some studies proposed supervised approaches to learn complementary relationships, such as co-purchased and content compatibility . Recently, many researchers attempted to mine the compatibility of fashion items to better understand the complementary relationships in the clothing recommendations [Han et al, 2017;Hsiao and Grauman, 2018]. In spite of the importance of existing studies, the exploration of compositional associations in the complementary item recommendations is still limited.…”