Supercritical water (SCW) is a novel thermal agent that has been recently utilized for the production of heavy oil. However, a lack of knowledge about its recovery mechanisms limits the application of SCW. In this study, pyrolysis and sandpack flooding experiments were performed to investigate the mechanisms and viability of SCW flooding. Then an innovative simulation model was developed for SCW flooding. Finally, sensitivity studies on SCW flooding were conducted by the developed model. The results showed that SCW flooding yielded a 13.99% increase in oil recovery in comparison to steam flooding, indicating that SCW flooding is technically applicable to offshore heavy oil reservoirs. Heavy oil upgrading in SCW can suppress coke formation and plays an important role in oil recovery. A novel numerical model for SCW flooding was established based on a history match of experiments. The simulation results suggested that during SCW flooding, SCW could induce heavy oil upgrading to increase oil mobility, and long-term injection of SCW may cause the formation of coke deposits. Higher injection temperatures and pressures would benefit the production performance of SCW flooding. However, an unlimited increase in temperature would damage formations by significant coke deposits.
When humans converse, what a speaker will say next significantly depends on what he sees. Unfortunately, existing dialogue models generate dialogue utterances only based on preceding textual contexts, and visual contexts are rarely considered. This is due to a lack of a large-scale multi-module dialogue dataset with utterances paired with visual contexts. In this paper, we release {\bf OpenViDial}, a large-scale multi-module dialogue dataset. The dialogue turns and visual contexts are extracted from movies and TV series, where each dialogue turn is paired with the corresponding visual context in which it takes place. OpenViDial contains a total number of 1.1 million dialogue turns, and thus 1.1 million visual contexts stored in images. Based on this dataset, we propose a family of encoder-decoder models leveraging both textual and visual contexts, from coarse-grained image features extracted from CNNs to fine-grained object features extracted from Faster R-CNNs. We observe that visual information significantly improves dialogue generation qualities, verifying the necessity of integrating multi-modal features for dialogue learning. Our work marks an important step towards large-scale multi-modal dialogue learning.
Many NLP tasks such as tagging and machine reading comprehension are faced with the severe data imbalance issue: negative examples significantly outnumber positive examples, and the huge number of easy-negative examples overwhelms the training. The most commonly used cross entropy (CE) criteria is actually an accuracy-oriented objective, and thus creates a discrepancy between training and test: at training time, each training instance contributes equally to the objective function, while at test time F1 score concerns more about positive examples. In this paper, we propose to use dice loss in replacement of the standard cross-entropy objective for data-imbalanced NLP tasks. Dice loss is based on the Sørensen--Dice coefficient or Tversky index , which attaches similar importance to false positives and false negatives, and is more immune to the data-imbalance issue. To further alleviate the dominating influence from easy-negative examples in training, we propose to associate training examples with dynamically adjusted weights to deemphasize easy-negative examples. Theoretical analysis shows that this strategy narrows down the gap between the F1 score in evaluation and the dice loss in training. With the proposed training objective, we observe significant performance boost on a wide range of data imbalanced NLP tasks. Notably, we are able to achieve SOTA results on CTB5, CTB6 and UD1.4 for the part of speech tagging task; SOTA results on CoNLL03, OntoNotes5.0, MSRA and OntoNotes4.0 for the named entity recognition task; along with competitive results on the tasks of machine reading comprehension and paraphrase identification.
The proposed pruning strategy offers merits over weight-based pruning techniques: (1) it avoids irregular memory access since representations and matrices can be squeezed into their smaller but dense counterparts, leading to greater speedup; (2) in a manner of top-down pruning, the proposed method operates from a more global perspective based on training signals in the top layer, and prunes each layer by propagating the effect of global signals through layers, leading to better performances at the same sparsity level. Extensive experiments show that at the same sparsity level, the proposed strategy offers both greater speedup and higher performances than weight-based pruning methods (e.g., magnitude pruning, movement pruning).
Ambient ionization based on liquid extraction is widely used in mass spectrometry<br>imaging (MSI) of molecules in biological samples. The development of nanospray desorption electrospray ionization (nano-DESI) has enabled the robust imaging of tissue sections with high spatial resolution. However, the fabrication of the nano-DESI probe is challenging, which limits its dissemination to the broader scientific community. Herein, we describe the design and performance of an integrated microfluidic probe (iMFP) for nano-DESI MSI. The glass iMFP fabricated using photolithography, wet etching, and polishing shows comparable performance to the capillary-based nano-DESI MSI in terms of stability and sensitivity; the spatial resolution of better than 25 μm was obtained in these first proof-of-principle experiments. The iMFP is easy to operate and align in front of a mass spectrometer, which will facilitate broader use of liquid extraction-based MSI in biological research, drug discovery, and clinical studies.
The development of wind power has brought about increasing challenges in decommissioning, among which DWTBs (decommissioned wind turbine blades) are the most difficult component to deal with. To enable the cost-effective, energy-efficient, and environmentally friendly large-scale utilization of DWTBs, an experimental study on thermogravimetric and pyrolysis characteristics of DWTBs was carried out. A new process involving recycling glass fiber with pyrolysis gas re-combustion and flue gas recirculation as the pyrolysis medium was innovatively proposed, and the simulation calculation was carried out. Thermogravimetric experiments indicated that glass fiber reinforced composite (GFRC) was the main heat-generating part in the heat utilization process of blades, and the blade material could basically complete pyrolysis at 600 °C. As the heating rate increased, the formation temperature, peak concentration, and proportion of combustible gas in the pyrolysis gas also increased. The highest peak concentration of CO gas was observed, with CO2 and C3H6 reaching their peaks at 700 °C. The solid product obtained from pyrolysis at 600 °C could be oxidized at 550 °C for 40 min to obtain clean glass fiber. And the pyrolysis temperature increased with the increase in the proportion of recirculation flue gas. When the proportion of recirculation flue gas was 66%, the pyrolysis temperature could reach 600 °C, meeting the necessary pyrolysis temperature for wind turbine blade materials. The above research provided fundamental data support for further exploration on high-value-added recycling of DWTBs.
Purpose The risk stratification of pediatric anaplastic large cell lymphoma (ALCL) has not been standardized. In this study, new risk factors were included to establish a new risk stratification system for ALCL, and its feasibility in clinical practice was explored.Materials and Methods On the basis of the non-Hodgkin’s lymphoma Berlin–Frankfurt–Munster 95 (NHL-BFM-95) protocol, patients with minimal disseminated disease (MDD), high-risk tumor site (multiple bone, skin, liver, and lung involvement), and small cell/lymphohistiocytic (SC/LH) pathological subtype were enrolled in risk stratification. Patients were treated with a modified NHL-BFM-95 protocol combined with an anaplastic lymphoma kinase inhibitor or vinblastine (VBL).Results A total of 136 patients were enrolled in this study. The median age was 8.8 years. The 3-year event-free survival (EFS) and overall survival of the entire cohort were 77.7% (95% confidence interval [CI], 69.0% to 83.9%) and 92.3% (95% CI, 86.1% to 95.8%), respectively. The 3-year EFS rates of low-risk group (R1), intermediate-risk group (R2), and high-risk group (R3) patients were 100%, 89.5% (95% CI, 76.5% to 95.5%), and 67.9% (95% CI, 55.4% to 77.6%), respectively. The prognosis of patients with MDD (+), stage IV cancer, SC/LH lymphoma, and high-risk sites was poor, and the 3-year EFS rates were 45.3% (95% CI, 68.6% to 19.0%), 65.7% (95% CI, 47.6% to 78.9%), 55.7% (95% CI, 26.2% to 77.5%), and 70.7% (95% CI, 48.6% to 84.6%), respectively. At the end of follow-up, one of the five patients who received maintenance therapy with VBL relapsed, and seven patients receiving anaplastic lymphoma kinase inhibitor maintenance therapy did not experience relapse.Conclusion This study has confirmed the poor prognostic of MDD (+), high-risk site and SC/LH, but patients with SC/LH lymphoma and MDD (+) at diagnosis still need to receive better treatment (ClinicalTrials.gov number, NCT03971305).
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
