Alibaba Group

About: Alibaba Group is a company organization based out in Hangzhou, China. It is known for research contribution in the topics: Computer science & Terminal (electronics). The organization has 6810 authors who have published 7389 publications receiving 55653 citations. The organization is also known as: Alibaba Group Holding Limited & Alibaba Group (Cayman Islands).

Meta Relational Learning for Few-Shot Link Prediction in Knowledge Graphs

Abstract: Link prediction is an important way to complete knowledge graphs (KGs), while embedding-based methods, effective for link prediction in KGs, perform poorly on relations that only have a few associative triples. In this work, we propose a Meta Relational Learning (MetaR) framework to do the common but challenging few-shot link prediction in KGs, namely predicting new triples about a relation by only observing a few associative triples. We solve few-shot link prediction by focusing on transferring relation-specific meta information to make model learn the most important knowledge and learn faster, corresponding to relation meta and gradient meta respectively in MetaR. Empirically, our model achieves state-of-the-art results on few-shot link prediction KG benchmarks.

Identification and characterization of circRNAs encoded by MERS-CoV, SARS-CoV-1 and SARS-CoV-2.

Abstract: The life-threatening coronaviruses MERS-CoV, SARS-CoV-1 and SARS-CoV-2 (SARS-CoV-1/2) have caused and will continue to cause enormous morbidity and mortality to humans. Virus-encoded noncoding RNAs are poorly understood in coronaviruses. Data mining of viral-infection-related RNA-sequencing data has resulted in the identification of 28 754, 720 and 3437 circRNAs encoded by MERS-CoV, SARS-CoV-1 and SARS-CoV-2, respectively. MERS-CoV exhibits much more prominent ability to encode circRNAs in all genomic regions than those of SARS-CoV-1/2. Viral circRNAs typically exhibit low expression levels. Moreover, majority of the viral circRNAs exhibit expressions only in the late stage of viral infection. Analysis of the competitive interactions of viral circRNAs, human miRNAs and mRNAs in MERS-CoV infections reveals that viral circRNAs up-regulated genes related to mRNA splicing and processing in the early stage of viral infection, and regulated genes involved in diverse functions including cancer, metabolism, autophagy, viral infection in the late stage of viral infection. Similar analysis in SARS-CoV-2 infections reveals that its viral circRNAs down-regulated genes associated with metabolic processes of cholesterol, alcohol, fatty acid and up-regulated genes associated with cellular responses to oxidative stress in the late stage of viral infection. A few genes regulated by viral circRNAs from both MERS-CoV and SARS-CoV-2 were enriched in several biological processes such as response to reactive oxygen and centrosome localization. This study provides the first glimpse into viral circRNAs in three deadly coronaviruses and would serve as a valuable resource for further studies of circRNAs in coronaviruses.

Visible processing method, apparatus and system for web page access behavior of users

Abstract: The invention discloses a method for visually processing user behaviors of web page access. The method comprises: acquiring click data of the user side; calculating the access frequency of users on various regions in the web page according to the acquired click data; matching the access frequency of the users on the various regions in the web page with corresponding regions; and displaying the matching result. The invention simultaneously discloses a device and a system for visually processing the user behaviors of web page access. The embodiment of the invention can closely connect the relations between the attention degree of the users on content in the web page and corresponding content in the web page, and intuitively and clearly display the attention degree of the users on the content in the web page.

Method and system utilizing communication terminal to pay

Abstract: The invention relates to a method utilizing a communication terminal to pay, comprising the following steps: (1), providing an intermediate platform; (2), judging whether the number of the communication terminal is the number of a mobile telephone or not when an incoming call request transmitted by the communication terminal is received by the intermediate platform; if not, prompting and receiving the number of the mobile telephone input by a user and bound with a payment mode by the intermediate platform; (3), finding out account information bound with the number of the mobile telephone from an account subsystem and receiving an payment request transmitted by the communication terminal of the user; (4), constructing payment alternation between the communication terminal of the user and the subsystem of a third party by the intermediate platform and confirming payment sum; and (5), constructing a relation between the intermediate platform and the user by a short message or speech external call mode, withholding the bound account on the intermediate platform only after a response that the user confirms the payment is received, and returning a withholding treatment result to the subsystem of the third party and the user. The payment method has safe payment and convenient operation and has no need of additionally increasing the cost.

Semantic Human Matting

Abstract: Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications. Since the matting problem is severely under-constrained, most previous methods require user interactions to take user designated trimaps or scribbles as constraints. This user-in-the-loop nature makes them difficult to be applied to large scale data or time-sensitive scenarios. In this paper, instead of using explicit user input constraints, we employ implicit semantic constraints learned from data and propose an automatic human matting algorithm (SHM). SHM is the first algorithm that learns to jointly fit both semantic information and high quality details with deep networks. In practice, simultaneously learning both coarse semantics and fine details is challenging. We propose a novel fusion strategy which naturally gives a probabilistic estimation of the alpha matte. We also construct a very large dataset with high quality annotations consisting of 35,513 unique foregrounds to facilitate the learning and evaluation of human matting. Extensive experiments on this dataset and plenty of real images show that SHM achieves comparable results with state-of-the-art interactive matting methods.