Aspect sentiment classification (ASC) is a fundamental task in sentiment analysis. Given an aspect/target and a sentence, the task classifies the sentiment polarity expressed on the target in the sentence. Memory networks (MNs) have been used for this task recently and have achieved state-of-the-art results. In MNs, attention mechanism plays a crucial role in detecting the sentiment context for the given target. However, we found an important problem with the current MNs in performing the ASC task. Simply improving the attention mechanism will not solve it. The problem is referred to as target-sensitive sentiment, which means that the sentiment polarity of the (detected) context is dependent on the given target and it cannot be inferred from the context alone. To tackle this problem, we propose the targetsensitive memory networks (TMNs). Several alternative techniques are designed for the implementation of TMNs and their effectiveness is experimentally evaluated.

Background: Recent studies have shown that early diagnosis and intervention promote the patient's good prognosis. For patients who underwent cardiac surgery and require extracorporeal circulation support, the incidence of postoperative cognitive dysfunction (POCD) is higher than in other types of surgery due to greater changes in brain perfusion compared with normal physiological conditions. Recent studies have confirmed that the use of ulinastatin or dexmedetomidine in the perioperative period effectively reduces the incidence of POCD. In this study, ulinastatin was combined with dexmedetomidine to assess whether the combination of the two drugs could reduce the incidence of POCD. Methods: One hundred and eighty patients with heart valve replacement surgery undergoing cardiopulmonary bypass from August 2017 to December 2018 were enrolled, with age 60-80 years, American Society of Anesthesiologists (ASA) grades I-III, education level above elementary school, and either gender. According to the random number table method, patients were grouped into ulinastatin + dexmedetomidine (U+D) group, ulinastatin (U) group, dexmedetomidine (D) group, and normal saline (N) control group. Group U was pumped 20,000 UI/kg immediately after induction and the first day after surgery, group D continued to pump 0.4 µg/kg/h from induction to 2 h before extubation, group U+D dexmedetomidine 0.4 µg/kg/h + ulinastatin 20,000 UI/kg, and group N equal volume of physiological saline. The patients were enrolled with Mini-Mental State Examination (MMSE) before surgery. The cognitive function was assessed by Montreal Cognitive Assessment (MoCA) on the first day before surgery and on the seventh day after surgery. Inflammatory factors, such as S100β protein, interleukin (IL)-6, matrix metalloproteinase (MMP)-9, and tumor necrosis factor (TNF)-α, were detected in peripheral blood before anesthesia (T0), immediately after surgery (T1), and immediately after extubation (T2). Results: One hundred and fifty-four patients enrolled in this study. Compared with group N, the incidence of POCD in group U+D was the lowest (P < 0.05), followed by group U and group D. Group U+D had the lowest concentration of inflammatory factors at the T1 and T2 time points, followed by group U and group D. Zhou et al. Ulinastatin and Dexmedetomidine for Postoperative Cognitive Dysfunction Conclusions: Both ulinastatin and dexmedetomidine can reduce the perioperative inflammatory response and the incidence of POCD in patients with heart valve surgery, and their combination can better reduce the incidence of POCD.

The serum levels of pro-inflammatory marker IL-6 and S-100β protein increased after total hip-replacement in elderly patients, and such increase may serve as predicting parameters for the occurrence of POCD.

To reveal information hiding in link space of bibliographical networks, link analysis has been studied from different perspectives in recent years. In this paper, we address a novel problem namely citation prediction, that is: given information about authors, topics, target publication venues as well as time of certain research paper, finding and predicting the citation relationship between a query paper and a set of previous papers. Considering the gigantic size of relevant papers, the loosely connected citation network structure as well as the highly skewed citation relation distribution, citation prediction is more challenging than other link prediction problems which have been studied before. By building a meta-path based prediction model on a topic discriminative search space, we here propose a two-phase citation probability learning approach, in order to predict citation relationship effectively and efficiently. Experiments are performed on real-world dataset with comprehensive measurements, which demonstrate that our framework has substantial advantages over commonly used link prediction approaches in predicting citation relations in bibliographical networks.