Recent advances in deep neural networks have achieved outstanding success in natural language processing tasks. Interpretation methods that provide insight into the decision-making process of these ...models have received an influx of research attention because of the success and the black-box nature of the deep text classification models. Evaluation of these methods has been based on changes in classification accuracy or prediction confidence when removing important words identified by these methods. There are no measurements of the actual difference between the predicted important words and humans’ interpretation of ground truth because of the lack of interpretation ground truth. A large publicly available interpretation ground truth has the potential to advance the development of interpretation methods. Manual labeling important words for each document to create a large interpretation ground truth is very time-consuming. This paper presents (1) IDC, a new benchmark for quantitative evaluation of interpretation methods for deep text classification models, and (2) evaluation of six interpretation methods using the benchmark. The IDC benchmark consists of: (1) Three methods that generate three pseudo-interpretation ground truth datasets. (2) Three performance metrics: interpretation recall, interpretation precision, and Cohen’s kappa inter-agreement. Findings: IDC-generated interpretation ground truth agrees with human annotators on sampled movie reviews. IDC identifies Layer-wise Relevance Propagation and the gradient-by-input methods as the winning interpretation methods in this study.
IEEE 802.16j is an amendment to the IEEE 802.16 broadband wireless access standard to enable the operation of multihop relay stations (RSs). It aims to enhance the coverage, per user throughput of ...IEEE 802.16e. Comparing with a base station (BS), RS does not need a wireline backhaul and has much lower hardware complexity. Moreover, using RSs can significantly reduce the deployment cost of the system. Unfortunately, there are some tradeoffs in the case of multi-hop RSs. Subscriber stations (SSs), located in a long distance, are suffered from the bottleneck of multi-hop, throughput degradation, and increasing of end-to-end delay. This paper proposes a network coding-based relay scheme for multi-hop relay networks, called NC-BR. It allows RSs to combine two wireless backhaul transmissions into one using the network coding technique. This paper also proposes an improved OFDMA frame structure design for the multi-hop relay network. This technique provides higher utilization for the relay zone by reorganizing the RSs transmission sequence. The analysis and simulation results confirm that the proposed scheme can enhance the throughput gain up to 140% and reduce the end-to-end delay by up to 83%. The simulation results also show that the proposed scheme can reduce the jitter by up to 58%.
How political actors choose which politics to focus on helps shape the outcome of the policy process. While the policy agenda of the federal government has received widespread attention, there is ...much less known about the policy agendas of the U.S. states. In this paper, we describe how and why states choose to have similar agendas. We rely on the Twitter activity of every state legislator in America to measure the attention that states pay to the categories developed in the Policy Agenda Project (PAP). We develop machine learning tools to measure the proportion of tweets from every state legislature from 2017 in each of the PAP policy topics. Our results show that states that the public‐facing policy agenda of a state legislature is correlated with the level of legislative professionalism and the partisan and ideological politics of the state. These results further our understanding of state policymaking and agenda setting.
Resumen
La forma en que los actores políticos eligen en qué política enfocarse ayuda a dar forma al resultado del proceso de formulación de políticas. Si bien la agenda de políticas del gobierno federal ha recibido una atención generalizada, se sabe mucho menos sobre las agendas de políticas de los estados de EE. UU. En este documento, describimos cómo y por qué los estados eligen tener agendas similares. Confiamos en la actividad de Twitter de todos los legisladores estatales de Estados Unidos para medir la atención que los estados prestan a las categorías desarrolladas en el Proyecto de Agenda de Políticas (PAP). Desarrollamos herramientas de aprendizaje automático para medir la proporción de tweets de cada legislatura estatal desde 2017 en los temas de política de PAP. Nuestros resultados muestran que los estados que la agenda política de cara al público de una legislatura estatal está correlacionada con el nivel de profesionalismo legislativo y la política ideológica y partidista del estado. Estos resultados amplían nuestra comprensión de la formulación de políticas estatales y el establecimiento de la agenda.
摘要
政治行动者如何对哪些政治加以关注一事会影响政策过程的结果。虽然联邦政府的政策议程受到广泛关注,但很少有研究聚焦于美国各州的政策议程。本文中,我们描述了各州如何以及为何选择相似的议程。我们基于美国各州立法者的推特活动来衡量各州对政策议程计划(PAP)中制定的类别的关注程度。我们开发了机器学习工具来衡量从2017年起各州立法机构的推文在PAP政策主题中的比例。我们的研究结果表明,州立法机构面向公众的政策议程,与立法专业水平、州党派政治以及意识形态政治相关。这些结果加深了我们对州政策制定和议程设置的理解。