Search

Additional file 1. Details of the semi-supervised learning method and examples of the labeled samples.

Existing conversational systems are mostly agent-centric, which assumes the user utterances would closely follow the system ontology (for NLU or dialogue state tracking). However, in real-world scenarios, it is highly desirable that the users can speak freely in their own way. It is extremely hard, if not impossible, for the users to adapt to the unknown system ontology. In this work, we attempt to build a user-centric dialogue system. As there is no clean mapping for a user's free form utterance to an ontology, we first model the user preferences as estimated distributions over the system ontology and map the users' utterances to such distributions. Learning such a mapping poses new challenges on reasoning over existing knowledge, ranging from factoid knowledge, commonsense knowledge to the users' own situations. To this end, we build a new dataset named NUANCED that focuses on such realistic settings for conversational recommendation. Collected via dialogue simulation and paraphrasing, NUANCED contains 5.1k dialogues, 26k turns of high-quality user responses. We conduct experiments, showing both the usefulness and challenges of our problem setting. We believe NUANCED can serve as a valuable resource to push existing research from the agent-centric system to the user-centric system. The code and data is publicly available at \url{https://github.com/facebookresearch/nuanced}., Comment: Findings of EMNLP 2021

Deep shale gas reservoirs buried underground with depth being more than 3500 m are characterized by high in-situ stress, large horizontal stress difference, complex distribution of bedding and natural cracks, and strong rock plasticity. Thus, during hydraulic fracturing, these reservoirs often reveal difficult fracture extension, low fracture complexity, low stimulated reservoir volume (SRV), low conductivity and fast decline, which hinder greatly the economic and effective development of deep shale gas. In this paper, a specific and feasible technique of volume fracturing of deep shale gas horizontal wells is presented. In addition to planar perforation, multi-scale fracturing, full-scale fracture filling, and control over extension of high-angle natural fractures, some supporting techniques are proposed, including multi-stage alternate injection (of acid fluid, slick water and gel) and the mixed- and small-grained proppant to be injected with variable viscosity and displacement. These techniques help to increase the effective stimulated reservoir volume (ESRV) for deep gas production. Some of the techniques have been successfully used in the fracturing of deep shale gas horizontal wells in Yongchuan, Weiyuan and southern Jiaoshiba blocks in the Sichuan Basin. As a result, Wells YY1HF and WY1HF yielded initially 14.1 × 10 4 m 3 /d and 17.5 × 10 4 m 3 /d after fracturing. The volume fracturing of deep shale gas horizontal well is meaningful in achieving the productivity of 50 × 10 8 m 3 gas from the interval of 3500–4000 m in Phase II development of Fuling and also in commercial production of huge shale gas resources at a vertical depth of less than 6000 m.

Collaborative personalization, such as through learned user representations (embeddings), can improve the prediction accuracy of neural-network-based models significantly. We propose Federated User Representation Learning (FURL), a simple, scalable, privacy-preserving and resource-efficient way to utilize existing neural personalization techniques in the Federated Learning (FL) setting. FURL divides model parameters into federated and private parameters. Private parameters, such as private user embeddings, are trained locally, but unlike federated parameters, they are not transferred to or averaged on the server. We show theoretically that this parameter split does not affect training for most model personalization approaches. Storing user embeddings locally not only preserves user privacy, but also improves memory locality of personalization compared to on-server training. We evaluate FURL on two datasets, demonstrating a significant improvement in model quality with 8% and 51% performance increases, and approximately the same level of performance as centralized training with only 0% and 4% reductions. Furthermore, we show that user embeddings learned in FL and the centralized setting have a very similar structure, indicating that FURL can learn collaboratively through the shared parameters while preserving user privacy.

Federated Learning allows for population level models to be trained without centralizing client data by transmitting the global model to clients, calculating gradients locally, then averaging the gradients. Downloading models and uploading gradients uses the client's bandwidth, so minimizing these transmission costs is important. The data on each client is highly variable, so the benefit of training on different clients may differ dramatically. To exploit this we propose Active Federated Learning, where in each round clients are selected not uniformly at random, but with a probability conditioned on the current model and the data on the client to maximize efficiency. We propose a cheap, simple and intuitive sampling scheme which reduces the number of required training iterations by 20-70% while maintaining the same model accuracy, and which mimics well known resampling techniques under certain conditions.

Interactive user interfaces need to continuously evolve based on the interactions that a user has (or does not have) with the system. This may require constant exploration of various options that the system may have for the user and obtaining signals of user preferences on those. However, such an exploration, especially when the set of available options itself can change frequently, can lead to sub-optimal user experiences. We present Explore-Exploit: a framework designed to collect and utilize user feedback in an interactive and online setting that minimizes regressions in end-user experience. This framework provides a suite of online learning operators for various tasks such as personalization ranking, candidate selection and active learning. We demonstrate how to integrate this framework with run-time services to leverage online and interactive machine learning out-of-the-box. We also present results demonstrating the efficiencies that can be achieved using the Explore-Exploit framework.

Coal resources and water resources play an essential and strategic role in the development of China's social and economic development, being the priority for China's medium and long technological development. As the mining of the coal extraction is increasingly deep, the mine water inrush of high-pressure confined karst water becomes much more a problem. This paper carried out research on the hundred-year old Kailuan coal mine's karst groundwater system. With the help of advanced Visual Modflow software and numerical simulation method, the paper assessed the flow field of karst water area under large-scale exploitation. It also predicted the evolution ofgroundwaterflow field under different mining schemes of Kailuan Corp. The result shows that two cones of depression are formed in the karst flow field of Zhaogezhuang mining area and Tangshan mining area, and the water levels in two cone centers are −270 m and −31 m respectively, and the groundwater generally flows from the northeast to the southwest. Given some potential closed mines in the future, the mine discharge will decrease and the water level of Ordovician limestone will increase slightly. Conversely, given increase of coal yield, the mine drainage will increase, falling depression cone of Ordovician limestone flow field will enlarge. And in Tangshan's urban district, central water level of the depression cone will move slightly towards north due to pumping of a few mines in the north.