Product bundling, offering a combination of items to customers, is one of the marketing strategies commonly used in online ecommerce and offline retailers. A high-quality bundle generalizes frequent items of interest, and diversity across bundles boosts the user-experience and eventually increases transaction volume. In this paper, we formalize the personalized bundle list recommendation as a structured prediction problem and propose a bundle generation network (BGN), which decomposes the problem into quality/diversity parts by the determinantal point processes (DPPs). BGN uses a typical encoder-decoder framework with a proposed feature-aware softmax to alleviate the inadequate representation of traditional softmax, and integrates the masked beam search and DPP selection to produce high-quality and diversified bundle list with an appropriate bundle size. We conduct extensive experiments on three public datasets and one industrial dataset, including two generated from co-purchase records and the other two extracted from real-world online bundle services. BGN significantly outperforms the state-of-the-art methods in terms of quality, diversity and response time over all datasets. In particular, BGN improves the precision of the best competitors by 16% on average while maintaining the highest diversity on four datasets, and yields a 3.85x improvement of response time over the best competitors in the bundle list recommendation problem.
Customers ask questions, and customer service staffs answer those questions. It is the basic service manner of customer service (CS). The progress of CS is a typical multi-round conversation. However, there are no explicit corresponding relations among conversational utterances. This paper focuses on obtaining explicit alignments of question and answer utterances in CS. It not only is an important task of dialogue analysis, but also able to obtain lots of valuable train data for learning dialogue systems. In this work, we propose end-to-end models for aligning question (Q) and answer (A) utterances in CS conversation with recurrent pointer networks (RPN). On the one hand, RPN-based alignment models are able to model the conversational contexts and the mutual influence of different Q-A alignments. On the other hand, they are able to address the issue of empty and multiple alignments for some utterances in a unified manner. We construct a dataset from an in-house online CS. The experimental results demonstrate that the proposed models are effective to learn the alignments of question and answer utterances.
Introduction: To observe macular microvascular changes in patients with ischemic and non-ischemic central retinal vein occlusion (CRVO) by optical coherence tomography angiography (OCTA), and explore the value of OCTA in differentiating ischemic and non-ischemic CRVO. Methods: Cross sectional study. Fifty patients diagnosed as CRVO with macular edema were included. Macular edema in all patients were regressive after three consecutive anti-VEGF treatment. Patients were divided into ischemic and non-ischemic group according to ultra-wide-angle fundus fluorescein angiography (UWFFA). All patients underwent BCVA, IOP, color fundus photography, UWFFA and OCTA. The following parameters were measured: (1) Vessel density (VD): superficial and deep whole VD (SVD, DVD), superficial and deep central fovea VD (SFVD, DFVD), superficial and deep parafoveal VD (SPFVD, DPFVD); (2) Central foveal retinal thickness (CRT); (3) Area of foveal avascular zone (FAZ), perimeter of FAZ (PERIM), avascular index of FAZ (AI) and VD within a width of 300 microns around the FAZ region (FD-300). Comparison between ischemic and non-ischemic group was performed by two independent sample t-tests. Receiver operating characteristic (ROC) curve analysis was used to measure the area under the curve (AUC) of VD for predicting ischemic CRVO. Results: There were no significant differences in IOP, SFVD, DFVD and CRT between ischemic and non-ischemic group, and significant differences in age, BCVA, SVD, SPFVD, DVD, DPFVD, FAZ area, PERIM, AI and FD-300 between ischemic and non-ischemic group. ROC curve analysis showed AUC of DVD and DPFVD in predicting ischemic CRVO was highest (0.962). the threshold was 38.40%, and the sensitivity was 100%, but the specificity of DVD (92.3%) was significantly higher than that of DPFVD (84.6%). Therefore, DVD ≤ 38.40% can be used as the best threshold for determining ischemic CRVO. Conclusion: OCTA can quantitatively evaluate the macular microvascular structure of CRVO, which is helpful to distinguish ischemic from non-ischemic CRVO.
Retinal vein occlusion (RVO) is a retinal vascular disease that severely impairs the visual function of patients. Observing the changes of retinal blood vessels before and after treatment is of great significance for the prognostic evaluation of RVO. The rapid development and widespread use of fundus imaging technique, especially ultra-wide-angle fundus fluorescein angiography (UWFFA) and optical coherence tomography angiography (OCTA) have made our observation of the retinal blood vessels of RVO more comprehensive and meticulous. In this paper, we reviewed the latest research progress of UWFFA and OCTA in RVO.
Objective To evaluate the aqueous humor levels of vascular endothelial growth factor (VEGF), connective tissue growth factor (CTGF), and tumor necrosis factor α (TNF-α) as biomarkers of the severity of proliferative diabetic retinopathy (PDR) in young and senior patients. Methods This was a prospective clinical study. From October 2020 to June 2021, 37 patients (37 eyes) who were diagnosed with PDR and received pars plana vitrectomy (PPV) at Tianjin Medical University Eye Hospital were recruited and allocated to either the young (16 patients, 16 eyes) or senior subgroup (21 patients, 21 eyes). Twelve patients with cataracts (12 eyes) who underwent phacoemulsification combined with intraocular lens (IOL) implantation during the same period were recruited in the control group. The fibrovascular proliferation (FVP) grade and PDR severity scores were recorded during PPV. Enzyme-linked immunosorbent assay was used to detect the levels of VEGF, CTGF, and TNF-α in the aqueous humor. Results (1) Young patients with PDR had a higher FVP grade and PDR severity score ( P = 0.037, = 0.009); (2) The levels of the three cytokines in the study group were all significantly higher than in the control group (all P < 0.001); (3) The CTGF level in the young subgroup (2239.55 ± 167.32 pg/mL) was significantly higher than that in the senior subgroup (2114.49 ± 102.04 pg/mL) ( P = 0.025). The VEGF level in the young subgroup (311.09 ± 10.74 pg/mL) was significantly lower than that in the senior subgroup (324.85 ± 14.97 pg/mL) ( P = 0.004). The TNF-α level was not statistically different between the two subgroups ( P = 0.382); (4) The CTGF/VEGF ratio in the young subgroup (7.20 ± 0.54) was significantly higher than in the senior subgroup (6.52 ± 0.39) ( P < 0.001); (5) The CTGF/VEGF ratio was positively correlated with the FVP grades (R = 0.377, P = 0.022) and with the PDR severity scores (R = 0.354, P = 0.032) in patients with PDR. Conclusion The aqueous humor CTGF/VEGF ratio was positively correlated with the severity of PDR. A higher CTGF/VEGF ratio in the aqueous humor proved that neovascular fibrosis was more serious in young patients when they received PPV.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.