Search

We present a computational framework for understanding the social aspects of emotions in Twitter conversations. Using unannotated data and semisupervised machine learning, we look at emotional transitions, emotional influences among the conversation partners, and patterns in the overall emotional exchanges. We find that conversational partners usually express the same emotion, which we name Emotion accommodation, but when they do not, one of the conversational partners tends to respond with a positive emotion. We also show that tweets containing sympathy, apology, and complaint are significant emotion influencers. We verify the emotion classification part of our framework by a human-annotated corpus.

Generating a novel and optimized molecule with desired chemical properties is an essential part of the drug discovery process. Failure to meet one of the required properties can frequently lead to failure in a clinical test which is costly. In addition, optimizing these multiple properties is a challenging task because the optimization of one property is prone to changing other properties. In this paper, we pose this multi-property optimization problem as a sequence translation process and propose a new optimized molecule generator model based on the Transformer with two constraint networks: property prediction and similarity prediction. We further improve the model by incorporating score predictions from these constraint networks in a modified beam search algorithm. The experiments demonstrate that our proposed model, Controlled Molecule Generator (CMG), outperforms state-of-the-art models by a significant margin for optimizing multiple properties simultaneously.

In present study, CHF experiments were conducted to evaluate and improve existing CHF correlations for narrow rectangular channels under a downward flow condition. A test section heated on both sides simulates a subchannel of plate-type fuels in a research reactor. The experiment covers a wide range of flow conditions: 0–2739 kg/m2 s for mass flux, 37–99 °C for inlet temperatures and 120–224 kPa for outlet pressures. The experimental data obtained in the present study are compared with prediction results of existing CHF correlations. From the evaluation, the correlation proposed by Kaminaga et al., (hereafter referred to as Kaminaga correlation) shows the best prediction performance for the experimental data among them. However, it was also found that the Kaminaga correlation still contains a large degree of uncertainty relative to the experimental data. In this study, we attempted to improve the Kaminaga correlation to minimize its prediction uncertainty. For this purpose, we investigated the gap size, mass flux and subcooling effects on the CHF in a narrow rectangular channel based on both the present and Mirshak CHF data. Finally, the Kaminaga correlation was modified to consider the effects of the gap size, mass flux, and subcooling on the CHF.

Automatic evaluation of open-domain dialogue response generation is very challenging because there are many appropriate responses for a given context. Existing evaluation models merely compare the generated response with the ground truth response and rate many of the appropriate responses as inappropriate if they deviate from the ground truth. One approach to resolve this problem is to consider the similarity of the generated response with the conversational context. In this paper, we propose an automatic evaluation model based on that idea and learn the model parameters from an unlabeled conversation corpus. Our approach considers the speakers in defining the different levels of similar context. We use a Twitter conversation corpus that contains many speakers and conversations to test our evaluation model. Experiments show that our model outperforms the other existing evaluation metrics in terms of high correlation with human annotation scores. We also show that our model trained on Twitter can be applied to movie dialogues without any additional training. We provide our code and the learned parameters so that they can be used for automatic evaluation of dialogue response generation models., Accepted at ACL 2020

Rod bundle flow is an important research field related to reactor cooling in nuclear power plants. Owing to the rapid development of computerized performance assessments, interest in coolant flow analysis using computational fluid dynamics has garnered research interest. Rod bundle flow research data compared with experimental results under various conditions are thus needed. To address this, a boiling model verification study was conducted with reference to experiments. This study adopts the Reynolds-averaged Navier&ndash, Stokes equation, a practical analysis method compared to direct numerical simulation and large eddy simulation, including turbulence modeling, to predict the flow of coolant inside a rod bundle. This study also investigates void behavior in low-pressure subcooled flow boiling using a Eulerian approach (two-fluid model). The rod bundle has a length of 0.59 m and a hydraulic diameter of approximately 14.01 mm. At the cross-section at a height of 0.58 m, near the exit, numerical results were compared with the experimental values of the volume fraction of vapor and interfacial area concentration. The simulation results showed good agreement with the experimental data for six different initial conditions with constant density.

Condensation heat transfer is significantly influenced by the geometrical and thermal-hydraulic parameters of the heat exchanger. Therefore, in order to assess their effects, experiments were conducted focused on the steam condensation on the outer surface of a vertical tube in the presence of a non-condensable gas. The experimental data on the heat transfer coefficient (HTC) were produced using three different tubes, with outer diameters of 21.5 mm, 33.6 mm, and 42.4 mm. The pressure, air mass fraction, and wall subcooling in the experiments ranged from 200−600 kPa, 0.05−0.80, and 16.1−55.6 K, respectively. Based on the data set consisting of the experimental data from the present study and the data collected in previous studies, an empirical correlation was developed, which accounts for the effects of the geometrical and thermal-hydraulic parameters. The applicable ranges of the proposed correlation for the air mass fraction, Jakob number, diffusion Reynolds number, and aspect ratio are 0.049−0.890, 8.83 × 10−3−3.41 × 10−2, 1.61−2.37 × 102, and 2.65−17.0, respectively. The maximum error of the correlation was calculated to be 20% with a 95% confidence level, when compared against the experimental data.

We investigated the thermal–hydraulic safety of spent fuel bundles installed in a mast assembly, assuming a lock-up accident of a fuel transfer system. For this, flow analysis with Computational Fluid Dynamics (CFD) for the natural convection and subsequent safety evaluation with 1-D Nuclear Power Plant (NPP) safety analysis code were carried out. Prior to the natural convection analysis for the mast assembly using CFD code, we performed benchmark calculations against two experimental data sets obtained by Betts and Bokhari (2000) and PNL (Pacific Northwest Laboratory, 1980). This was done to select the proper physical models for a natural convection flow analysis and to ensure the reliability of prediction of natural convection flow in the fuel bundle geometry. Finally, we performed a main natural convection analysis for the fuel assembly inside the mast assembly. From this calculation, we observed a stable natural circulation flow between the mast assembly and pool side, and obtained coolant velocity at the inlet of the spent fuel bundle. This flow condition is given as a boundary condition for the 1-D NPP safety analysis code, which is used to predict the Critical Heat Flux and then departure from the nucleate boiling ratio.

Bubble size is a key parameter for an accurate prediction of bubble behaviours in the multi-dimensional two-phase flow. In the current STAR CCM+ CFD code, a mechanistic bubble size model S γ is available for the prediction of bubble size in the flow channel. As another model, Yun model is developed based on DEBORA that is subcooled boiling data in high pressure. In this study, numerical simulation for the gas-liquid two-phase flow was conducted to validate and confirm the performance of S γ model and Yun model, using the commercial CFD code STAR CCM+ ver. 10.02. For this, local bubble models was evaluated against the air-water data from DEDALE experiments (1995) and Hibiki et al. (2001) in the vertical pipe. All numerical results of S γ model predicted reasonably the two-phase flow parameters and Yun model is needed to be improved for the prediction of air-water flow under low pressure condition. Key words : two-phase flow, air-water flow, bubbly flow, bubble size model 1. 서 론

Styles of leaders when they make decisions in groups vary, and the different styles affect the performance of the group. To understand the key words and speakers associated with decisions, we initially formalize the problem as one of predicting leaders’ decisions from discussion with group members. As a dataset, we introduce conversational meeting records from a historical corpus, and develop a hierarchical RNN structure with attention and pre-trained speaker embedding in the form of a, Conversational Decision Making Model (CDMM). The CDMM outperforms other baselines to predict leaders’ final decisions from the data. We explain why CDMM works better than other methods by showing the key words and speakers discovered from the attentions as evidence.

Vector representation of words improves performance in various NLP tasks, but the high dimensional word vectors are very difficult to interpret. We apply several rotation algorithms to the vector representation of words to improve the interpretability. Unlike previous approaches that induce sparsity, the rotated vectors are interpretable while preserving the expressive performance of the original vectors. Furthermore, any prebuilt word vector representation can be rotated for improved interpretability. We apply rotation to skipgrams and glove and compare the expressive power and interpretability with the original vectors and the sparse overcomplete vectors. The results show that the rotated vectors outperform the original and the sparse overcomplete vectors for interpretability and expressiveness tasks.

Conversational partners influence each others' emotions and topics. Using a large dataset of Twitter conversations and an unsupervised machine learning technique, we discover patterns of emotion influence in naturally occurring conversations. We describe our computational framework for automatically classifying emotions, analyzing the emotional transitions, and discovering emotion influence patterns. We found that conversational partners usually express the same emotion (emotion contagion), but when they do not, one of the conversational partners tends to respond with a positive emotion. Also, tweets containing sympathy, apology, and complaint are significant emotion influencers. One of the interesting findings is that expressing a desired emotion is the best strategy to alter partner's emotion.

Self-disclosure, the act of revealing oneself to others, is an important social behavior that contributes positively to intimacy and social support from others. It is a natural behavior, and social scientists have carried out numerous quantitative analyses of it through manual tagging and survey questionnaires. Recently, the flood of data from online social networks (OSN) offers a practical way to observe and analyze self-disclosure behavior at an unprecedented scale. The challenge with such analysis is that OSN data come with no annotations, and it would be impossible to manually annotate the data for a quantitative analysis of self-disclosure. As a solution, we propose a semi-supervised machine learning approach, using a variant of latent Dirichlet allocation for automatically classifying self-disclosure in a massive dataset of Twitter conversations. For measuring the accuracy of our model, we manually annotate a small subset of our dataset, and we show that our model shows significantly higher accuracy and F-measure than various other methods. With the results our model, we uncover a positive and significant relationship between self-disclosure and online conversation frequency over time.