Abstract. To tackle the problem of severe air pollution, China has implemented active clean air policies in recent years. As a consequence, the emissions of major air pollutants have decreased and the air quality has substantially improved. Here, we quantified China's anthropogenic emission trends from 2010 to 2017 and identified the major driving forces of these trends by using a combination of bottom-up emission inventory and index decomposition analysis (IDA) approaches. The relative change rates of China's anthropogenic emissions during 2010–2017 are estimated as follows: −62 % for SO2, −17 % for NOx, +11 % for nonmethane volatile organic compounds (NMVOCs), +1 % for NH3, −27 % for CO, −38 % for PM10, −35 % for PM2.5, −27 % for BC, −35 % for OC, and +16 % for CO2. The IDA results suggest that emission control measures are the main drivers of this reduction, in which the pollution controls on power plants and industries are the most effective mitigation measures. The emission reduction rates markedly accelerated after the year 2013, confirming the effectiveness of China's Clean Air Action that was implemented since 2013. We estimated that during 2013–2017, China's anthropogenic emissions decreased by 59 % for SO2, 21 % for NOx, 23 % for CO, 36 % for PM10, 33 % for PM2.5, 28 % for BC, and 32 % for OC. NMVOC emissions increased and NH3 emissions remained stable during 2010–2017, representing the absence of effective mitigation measures for NMVOCs and NH3 in current policies. The relative contributions of different sectors to emissions have significantly changed after several years' implementation of clean air policies, indicating that it is paramount to introduce new policies to enable further emission reductions in the future.

No abstract

Spoken Language Understanding (SLU), which typically involves intent determination and slot filling, is a core component of spoken dialogue systems. Joint learning has shown to be effective in SLU given that slot tags and intents are supposed to share knowledge with each other. However, most existing joint learning methods only consider joint learning by sharing parameters on surface level rather than semantic level. In this work, we propose a novel self-attentive model with gate mechanism to fully utilize the semantic correlation between slot and intent. Our model first obtains intent-augmented embeddings based on neural network with self-attention mechanism. And then the intent semantic representation is utilized as the gate for labelling slot tags. The objectives of both tasks are optimized simultaneously via joint learning in an end-to-end way. We conduct experiment on popular benchmark ATIS. The results show that our model achieves state-of-the-art and outperforms other popular methods by a large margin in terms of both intent detection error rate and slot filling F1-score. This paper gives a new perspective for research on SLU.