No abstract

Abstract. In online social networks, most relationships are lack of meaning labels (e.g., "colleague" and "intimate friends"), simply because users do not take the time to label them. An interesting question is: can we automatically infer the type of social relationships in a large network? what are the fundamental factors that imply the type of social relationships? In this work, we formalize the problem of social relationship learning into a semi-supervised framework, and propose a Partially-labeled Pairwise Factor Graph Model (PLP-FGM) for learning to infer the type of social ties. We tested the model on three different genres of data sets: Publication, Email and Mobile. Experimental results demonstrate that the proposed PLP-FGM model can accurately infer 92.7% of advisoradvisee relationships from the coauthor network (Publication), 88.0% of manager-subordinate relationships from the email network (Email), and 83.1% of the friendships from the mobile network (Mobile). Finally, we develop a distributed learning algorithm to scale up the model to real large networks.

No abstract

Despite the recent success of multi-task learning and transfer learning for natural language processing (NLP), few works have systematically studied the effect of scaling up the number of tasks during pre-training. Towards this goal, this paper introduces EXMIX (Extreme Mixture): a massive collection of 107 supervised NLP tasks across diverse domains and task-families. Using EXMIX, we study the effect of multi-task pre-training at the largest scale to date, and analyze cotraining transfer amongst common families of tasks. Through this analysis, we show that manually curating an ideal set of tasks for multi-task pre-training is not straightforward, and that multi-task scaling can vastly improve models on its own. Finally, we propose EXT5: a model pre-trained using a multi-task objective of self-supervised span denoising and supervised EXMIX. Via extensive experiments, we show that EXT5 outperforms strong T5 baselines on SuperGLUE, GEM, Rainbow, Closed-Book QA tasks, and several tasks outside of EXMIX. EXT5 also significantly improves sample efficiency while pre-training. * Google AI Resident. † Equal contribution. Sebastian is now at Google Research. Sanket returned to CMU.

No abstract