The dissemination of information by individuals, which has recently gained widespread attention, significantly affects multitudes of people regardless of the accuracy of the information. There is a vast amount of information available, making it difficult to identify baseless rumors, when the information is unverified. This study presents a bot called “Chillmo, ” which has been developed to alert users of such rumors. The proposed system aims to raise users' awareness of the reliability of information by actively alerting the users regarding the rumors receiving attention and providing the accurate information about such rumors through quizzes. Based on the experimental results, it was confirmed that Chillmo can make the users skeptical about rumors, promote the verification of authenticity, and raise their interest in the reliability of the information. In the future, we intend to conduct experiments with a larger number of people using the system to investigate trends by age and changes in the system usage over time.

The Institute of Humanities at Shinshu University has instituted an e-learning system to provide mathematics and statistics pre-enrollment education (PE) to applicants who passed the 2020 examination for candidates recommended for the faculty of engineering. This study presents this PE's results. PE can be categorized into individual work and group work. In individual work, accepted applicants answer mathematics and statistics problems provided by the University e-Learning Association. In group work, they search solutions to such problems by consulting with the faculty of engineering's undergraduate students and other accepted applicants; a group representative submits the answers. We show that accepted applicants can maintain self-efficacy even in perfectly distributed asynchronous PEs.

By exchanging information on objects on the road detected by in-vehicle sensors through inter-vehicle communication, vehicles can detect objects that cannot be detected directly by their own sensors. This enables the connected vehicles (CVs) to provide more appropriate safe driving support and automatic driving even in the environment where connected and non-connected vehicles are mixed. Such a system is called a Collective Perception System, and has been actively researched in recent years. However, if individual CVs simply transmit information on the objects they have detected frequently, the wireless communication channel will be congested, and the data originally intended to be sent will not reach the destination, making it difficult for each vehicle to detect other objects quickly. Therefore, appropriate congestion control technology for sensor information transmission is necessary. This paper introduces recent techniques for congestion control in the transmission of vehicle and sensor information in inter-vehicle communication. In addition, we introduce a congestion control technique we have designed based on the relative positions of vehicles.

Reliable driving intention inference is an essential issue in the mixed automation traffic system. To improve traffic safety and efficiency, this study develops an accurate and efficient driving intention inference framework named FES-XGB, which is short for Feature Extraction and Selection based eXtreme Gradient Boosting (XGBoost) algorithm. In contrast with conventional approaches, which only consider motion information of the subject and neighboring vehicles, this study includes a new kind of decision variables into driving intention inference for the first time, i.e., the local and global traffic environment information assumed to be obtained from vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) technology. The high-precision NGSim trajectory dataset is employed to learn the relationship between traffic environment information and driving intentions and evaluate the proposed framework. According to the experiment results, by taking the environment information as additional input, the accuracy of the conventional XGBoost model can increase from 89.42% to 92.86%, indicating the environment information has a close relationship with the driving intention. By employing the proposed FES-XGB framework, the accuracy can be further increased to 94.09%, while the training and online inference cost can be reduced by 94.03% and 65.25% respectively. With the traffic environment information as additional input, the proposed FES-XGB framework can be integrated into advanced driver-assistance systems (ADAS) for a safer and more efficient traffic system.

This paper presents a statistical method combined with a neural network for efficient traffic prediction from a limited amount of training data. The traffic prediction during a large-scale event is essential to maintain the safety of event participants. The conventional methods for predicting traffic time series, however, cannot be utilized because the rare nature of the large-scale events prevents us from preparing a sufficient amount of training data. To efficiently train traffic prediction from a limited amount of training data, we propose a pattern-aware regression method that reduces the number of model parameters by interpreting traffic data as a weighted sum of latent behavior patterns. The proposed method trains a neural regression model to predict the weights of these patterns from the event information instead of directly predicting the traffic time series. The behavior patterns are jointly estimated during the training in a Bayesian manner to avoid overfitting. We performed experiments with foot traffic data recorded at a real soccer stadium and show that the proposed method outperforms the conventional direct regression methods. We also demonstrate an application of our method for predicting travel time from the stadium to the nearest highway interchange, which outperforms a popular commercial service.

As smartphones are much used over a wide area, the pedestrian navigation systems are greatly utilized in our daily lives. In general, navigation systems use GPS (Global Positioning System) for the user's positioning but its accuracy tends to decrease in indoor environments. A pedestrian dead reckoning method, or PDR method in short, is one of the positioning methods in indoor environments, which estimates the user's positions by using sensors such as acceleration and angular velocity sensors. These PDR methods do not always use external infrastructures and hence they can be implemented with a lower cost. When we consider using a smartphone as a PDR sensor device, attention must be paid to the fact that there are various carrying modes such as holding it directly and carrying it inside a pocket. How to deal with these various carrying modes is of great concern in PDR when using a smartphone. In this paper, we propose a PDR method based on a combination of a smartphone and a smartwatch. By synchronizing the smartphone and smartwatch sensors effectively, the proposed method can successfully reduce drift errors and thus estimate accurately the user's positions, compared to just using a smartphone. Furthermore, even when the user carries his/her smartphone in various carrying modes, the proposed method still realizes accurate PDR. The experimental results demonstrate that the positioning errors are reduced by approximately 82.1% on average compared to the existing method.

As people spend more time indoors owing to the COVID-19 global pandemic, the automatic detection of indoor human activity has increasingly become of interest to researchers and consumers. Conventional Wi-Fi Channel State Information (CSI)-based detection provides adequate accuracy; however, they have a deployment constraint owing to specific hardware and software for full CSI acquisition. This study exploits the Compressed Beamforming Report (CBR), which is a default form of CSI in IEEE 802.11ac and 11ax, to address the constraint in Wi-Fi CSI-based methods. The CBRs are shared among most IEEE 802.11ac compliant devices and are easily obtained with outer sniffers. Our CBR-based Activity Count Estimator (CBR-ACE) is a novel wireless sensing system using CBRs. The CBR-ACE provides a Raspberry Pi-based tool to easily deploy a new wireless sensing system into existing networks, and utilizes the CBR irregularity for automatic detection. From experiments in real-dwelling environments, the proposed CBR-ACE achieves average estimation errors of 0.97 in the best case.

Hashcash, which is a Proof of Work (PoW) of bitcoin, is based on a preimage problem of hash functions of SHA-2 and RIPEMD. Since these hash functions employ the Merkle-Damgard (MD) construction, a preimage can be found with a negligible amount of memory. It is well known that such calculations can be speeded up by ASIC, and this causes a serious problem from the so-called 51% attack by dedicated ASIC mining pools. To address this issue, we propose a new PoW scheme based on a preimage problem of variants of SHA-3. Unlike SHA-2 and RIPEMD, SHA-3 adopts a sponge construction as an underlying domain extension algorithm. This difference allows us to make the problem of finding a preimage very memory-consuming calculations by properly choosing parameters of sponge functions. As a result, our scheme can achieve ASIC resistance by using SHA-3 for Hashcash.

It is well known that the following two conditions should be satisfied in the control-point based geometric correction of historical maps: (a) Conversion from a historical map into a present map is a homeomorphism and (b) The straightness of designated specific line segments is maintained. In this paper, a new method for the control-point based geometric correction of historical maps, which simultaneously satisfies both the above conditions, is proposed. The correction process is modeled as a phenomenon in a three-dimensional vector field. Each point in a historical map is connected with the corresponding point in a present map by a streamline of the field. Since a unique streamline passes through any point in the vector field having no zero-vector point, the above connection relationship becomes a homeomorphism. As a result, Condition (a) is satisfied. On the other hand, the straightness of designated line segments is maintained because streamlines intersecting with the line segments in the historical map are formed so as to necessarily intersect with the corresponding line segments in the present map. Consequently, Condition (b) is satisfied. The experimental results demonstrate the effectiveness of the proposed method.

Recently, as digital machine tools have become popular, even ordinary people have become involved in personal fabrication. However, optical elements such as lenses are difficult to be manufactured because they require post-output processing such as polishing. To solve this problem, we focus on an ultraviolet (UV) printer, which is getting popular in Makerspaces and Fab Labs. We propose a lens forming method that does not require post-output processing using the UV printer. The transparent ink is first laminated to form a shape. The UV printer then smooths its surface by filling the layer roughness with glossy printing. We implemented tools to design the lens shape by inputting the lens diameter, focal length, and so on. In this paper, we introduce the concept, implementation, and applications of our lens forming method. We also explain the results of performance evaluation and user study.

Beginners must learn correct swing motion in golf. Systems are available that provide visual, auditory, or haptic feedback to the user. Such systems, however are only one modality and do not explore combinations of different modalities. We developed a wearable device using vibro-transducers that can provide auditory and haptic feedback separately or simultaneously. We also conducted user studies to compare how each modality improves the common flaws in golf swings, such as sway, separated elbow, and head lift. The results demonstrated that the combination of auditory and haptic feedback is significantly effective in correcting these faults.

Disaster information-sharing among emergency personnel and residents helps facilitate quick and safe evacuation from disaster areas. However, it is difficult to communicate with people in disaster areas when communications infrastructure is not available. To tackle this problem, we have been developing a heterogeneous delay/disruption tolerant network (DTN)-based disaster information-sharing system that uses long-range narrowband links (e.g., LoRa) and short-range broadband links (e.g., Wi-Fi). Our disaster information-sharing system disseminates evacuation recommendations, evacuation routes, and other vital information to residents in disaster-stricken areas by the store-carry-forward method using residents' mobile devices and relay nodes installed at roadsides, fire departments, and shelters. Since evacuee movements and information dissemination in heterogeneous DTN affect each other, and since understanding those interactions will facilitate the development of more efficient disaster information distribution strategies, we developed a simple cellular-automaton-based simulation model to investigate data dissemination in our heterogeneous-DTN system and analyzed those interactions. Our simulation results revealed a particular complication that we have named the “leaving-behind phenomenon, ” in which people who have obtained information on damaged roads take detours to a shelter before providing the same information to the evacuees that will approach the damaged road. Based on our simulation results, we report on a way to alleviate this problem by installing fixed relay nodes at the most suitable locations.

Context recognition has attracted attention for various daily life applications. Many existing approaches use micro-electromechanical systems (MEMS) sensors which require additional silicon chips to process and transmit the sensor data. The energy consumption of such components is relatively large, requiring maintenance for charging or replacing batteries. In this paper, we propose BAAS: a novel concept using Backscatter As A Sensor. BAAS recognizes contexts using a frequency shift backscatter tag with ultra-low power consumption. The key components of the backscatter tag are an oscillator and a motion switch. The state of the motion switch changes according to the movement of humans or the change of the situation of things. While the motion switch is on, the energy is supplied to the oscillator, and the frequency of the backscattered signal shifts according to the oscillation frequency of the oscillator. Context recognition is achieved by observing the existence and absence of the frequency shift. To demonstrate the feasibility of context recognition using the backscatter tag, we have implemented a prototype and evaluated its performance. Our results show that we can detect the frequency shift by BAAS within 3m, backscattering BLE signal from an exciter implemented by a commodity device.

Multi/many-core processors are being increasingly used to reduce power consumption and improve performance. The use of model-based development for embedded systems has also been increasing. Relative to these trends, the model-based parallelizer or MBP has an essential role in parallelizing applications at the model level. MBP maps Simulink blocks to cores using various types of information such as block characteristics, C code, and multi/many-core hardware implementation. However, MBP does not consider many-core hardware with cluster structures such as Kalray MPPA2-256 processor which contains 16 clusters of 16 cores for 256 general-purpose cores in total. This paper proposes an algorithm that determines core allocations by considering cluster structures. The proposed algorithm combines two other algorithms: one algorithm uses the core allocation of MBP and path analysis at the cluster-level and considers effects from communication contention when determining cluster allocations, and the other algorithm uses the results from MBP and remaps cluster allocations. The proposed algorithm produces better results than its component algorithms could produce separately. Evaluations demonstrate that the proposed algorithm obtained the best results among four methods in terms of execution time on Simulink models.

We propose a privacy-preserving scheme to outsource zero-knowledge proof generation to a party that we call a worker. Our scheme can be applied to zk-SNARKs with a trusted setup, zero-knowledge proofs deployed in many applications. Compared to known privacy-preserving outsourcing schemes, our scheme is more practical in the sense that the computational and memory load on the worker is almost the same as that on the prover in cases where the provers generate proofs on their own.

We propose an approach called automatic optimize-time validation for binary optimizers. Our approach does not involve executing the whole program for validation but selecting a small part of code (1 to 100 instructions) for the target test code. It executes the target code and its optimized code with several input data during binary optimization. One benefit is that we can test a small part of an actual customer's code during binary optimization. Our approach can be used to test several input data not included in the target code, which is the most beneficial aspect of the approach. If the results are the same after execution, we will use the optimized code for the final output code. If the results differ, we can consider a couple of option, e.g., while developing a binary optimizer, we can abort the compilation with an error message to easily detect a bug. After a binary optimizer becomes generally available, we can use the input code for the final output code to maintain compatibility. Our goal is for the output binary code to be compatible, fast, and small. We focused on how to improve compatibility in this study. We implemented our approach in our binary optimizer and successfully detected one new bug. We used a very small binary program to observe the worst case of increased compilation time and output binary file size. Our implementation showed that our approach increases optimization time by only 0.02% and output binary file size by 8%.

GPU with the Single Instruction Multiple Data (SIMD) execution model enables a program to work efficiently. However, the efficiency may decrease because of branch divergence that occurs when SIMD threads follow different paths in some branches. Once the divergence occurs, some threads must wait until completion of the execution of the others. Thus, it is important to reduce branch divergence to improve the efficiency of GPU programs. On the other hand, branch divergence may be increased by some traditional code optimizations based on code motion such as partial redundancy elimination (PRE) and scalar replacement (SR). These methods insert some expressions into some paths, on which insertion points may be included in divergent branches. That is, the insertion may increase branch divergence, which may result in the decrease of execution efficiency of GPU programs. In this paper, we propose a new SR approach, called Speculative SR based on Question Propagation (SSRQP), which not only removes redundant memory accesses but also reduces branch divergence. SSRQP achieves SR based on speculative code motion, which not only eliminates inter-iteration redundant memory accesses without increasing branch divergence but also decreases branch divergence that originally exists through hoisting memory accesses in true and false sides of a divergent branch out of it. To prove the effectiveness of our method, we have conducted experiments through applying it to some benchmarks with divergent branches. The experimental results show that it can improve the efficiency about 40% in the best case in comparison with traditional techniques.

Low-latency I/O devices are connected to the peripheral component interconnect express bus on a database server. Most practical database systems are built as a high availability system to avoid a single point of failure. Therefore, we evaluated a high availability database system configured with servers using low-latency I/O devices. We have shown that the performance overhead of the high availability configuration using low-latency solid state drives is 12% compared to a single server configuration, in a primitive update test case. The result of a mixed-workload benchmark indicated that the database system configuration using low-latency I/O devices was up to 6.1 times faster than the performance using traditional external storage when the allocated database buffer was 5% of the database size.

Overtourism has a negative impact on tourist sites all over the world. Serious problems are environmental issues, such as littering, caused by the rush of too many visitors. It is important to change people's mindset to be more environmentally aware for improving this situation. In particular, if we can find people with a high awareness about environments, we can work effectively to promote eco-friendly behavior by taking them as the start. However, grasping individual awareness is inherently difficult. For this challenge, we utilize SNS data, which are available in large volume, with a hypothesis that people's subconsciousness influences their posts. In this paper, we address two research topics for grasping such awareness. First, we propose a classification task, in which a system is given users' SNS posts about tourist sites, and classifies them into types of their focuses. Experimental results show widely-used classifiers can solve the task at about 0.84 of accuracy using our created dataset. Second, we investigate the relation of the focuses and such awareness with a questionnaire survey targeting over 2,700 people, and show that users' awareness influences focuses of SNS posts with both of a statistical analysis and an analysis using real-world data.

Soft-bodied animals move by using the anisotropic friction between their body and the environment. Inspired by these bodily structures, we propose a wriggle soft-bodied robot with an anisotropic frictional skin to support its locomotion. We combine soft and rigid materials in 3D printers to make different patterns of frictional anisotropy. We build a simulation model to predict the locomotion speed of the robot with different anisotropic frictional patterns. By using these patterns as the ventral side, we design a wriggle soft-bodied robot that can move 2.8 times faster than a robot with an omnidirectional frictional ventral surface. The fabrication time is less than one hour, and the locomotion speed is 0.17 body-lengths per second.

Achieving differential privacy and utilizing secure multiparty computation are the two primary approaches used for ensuring privacy in privacy-preserving machine learning. However, the privacy guarantee by existing integration protocols of both approaches for collaborative learning weakens when more participants join the protocols. In this work, we present Secure and Private Gradient Computation (SPGC), a novel collaborative learning framework with a strong privacy guarantee independent of the number of participants while still providing high accuracy. The main idea of SPGC is to create noise for the differential privacy within secure multiparty computation. We also created an implementation of SPGC and used it in experiments to measure its accuracy and training time. The results show that SPGC is more accurate than a naive protocol based on local differential privacy by up to 5.6%. We experimentally show that the training time increases in proportion to the noise generation and then demonstrate that the privacy guarantee is independent of the number of participants as well as the accuracy evaluation.

We propose a high-speed packet forwarding architecture on general-purpose servers, in which a Network Interface Card (NIC) drives packet forwarding by direct packet transfer to other NICs via a PCIe switch. The demand for high-speed packet forwarding technology on general-purpose servers is increasing with the spread of networking concepts such as Network Function Virtualization (NFV). However, the current architecture, which processes packets by CPU, cannot achieve the similar degree of performance that hardware routers can provide because the processing capacity of the CPU and the bandwidth of the main memory constrain the performance. Our proposed method, called P2PNIC, overcomes this constraint by eliminating the CPU and the main memory from the entire packet forwarding. In the P2PNIC architecture, a NIC determines to which NIC to forward the packets and directly transfers the packets to the NIC over the PCIe. We evaluate the P2PNIC architecture by comparing it with the DPDK applications as examples of the current architecture. The evaluation shows that the P2PNIC architecture achieves 3.44 times higher throughput and up to 79% lower latency than the DPDK applications. This study offers a new approach in software-based network infrastructure for achieving comparable performance with hardware routers in the future.

Collaborative learning practices foster the ability to solve creative problems in collaboration with other learners. The collaboration enables learners to learn new ideas from other learners and enhances the social ability of the learners through interaction with other learners. Although the learning science field now uses qualitative analysis to analyze the effects of the collaborative discourse, qualitative analysis requires much human and time costs to analyze the collaborative discourse with dozens of students. This study proposes Sensor-based Regulation Profiler to reduce the analysis costs. The proposed scheme consists of the business card-type sensors that acquire sensor data from each learner with a precise time synchronization as well as learning analysis methods that analyze the collaborative discourse from the acquired sensor data. Experimental evaluations using the proposed scheme showed that the proposed business card-type sensors realized a time synchronization error of 7.7µs on average across the sensors. In addition, the proposed learning analysis could extract and visualize the collaborative activity of each learner in the collaborative discourse through the social graph extraction, learning phase extraction, speaker identification, and activity estimation by using the sensor data from the proposed business card-type sensors.

This paper describes the development of an electric appliance controller using an IoT system. The IoT system can configure and control its edge devices which are placed behind a Network Address Translator (NAT) using Wiki pages. These operations were realized by Bot Computing, a framework for Internet of Things (IoT). Any electric appliance can be controlled using the combination of Bot Computing and edge devices with an Infra-Red (IR) transmitter, if the appliance has the IR remote controlled function. We can program the power on/off time of any electric appliances, by writing a script on a Wiki page on the Internet, using Bot Computing. We can change the program anytime, anywhere. We also can control turning on or off the appliance anytime, anywhere.

The widespread use of cloud computing has made it easier for service providers to develop new features and handle increased access. However, the network dependencies among components in distributed applications deployed in the cloud are becoming more complex because the number and types of components are increasing. When system administrators make changes to a system, they cannot specify the impact of the changes, which may lead to larger failures than expected. Current methods of automatically discovering dependencies trace network flows included in TCP/UDP sockets in the Linux kernel on all hosts deployed in distributed applications. However, as the rate of communication increases, the number of flows transferred from the kernel space to user space increases, which increases CPU usage for tracing. We propose a low-overhead method of bundling multiple flows with the same network service into a single flow in a kernel to discover dependencies. The proposed method reduces the number of transferred flows to the user space, thus reducing CPU usage. Experimental results from evaluating our method indicate that the method maintains a CPU overhead below 2.2% when the number of flows increases.

User-level threading or task-parallel systems have been developed over decades to provide efficient and flexible threading features missing from kernel-level threading for both parallel and concurrent programming. Some of the existing state-of-the-art user-level threading libraries provide interfaces to customize the implementation of thread scheduling to adapt to different workloads from both applications and upper-level systems. However, most of them are typically built as huge sets of monolithic components which achieve customizability with additional costs via concrete C APIs. We have noticed that the zero-overhead abstraction of C++ is beneficial for assembling flexible user-level threading in a clearer manner. To demonstrate our ideas, we have implemented a new user-level threading library ComposableThreads which provides customizability while minimizing the interfacing costs. We show that the users can pick up, insert, or replace the individual classes of ComposableThreads for their own purposes. ComposableThreads offers several characteristic abstractions to build high-level constructs of user-level threading including suspended threads (one-shot continuations) and lock delegators. We evaluate both the customizability and performance of our runtime system through the microbenchmark and application results.

WLANs have been used in a variety of places. The MAC protocol is an important item of WLANs, and directly affects the transmission efficiency. The distributed MAC protocol has the advantage of not requiring infrastructure such as access points, but it also has the problem that the total throughput decreases significantly when traffic is overloaded due to hidden node problems. In this paper, we focus on the MAC protocol and propose a new MAC protocol that provides a dynamically optimal back-off process in multi-hop wireless networks. To improve hidden node problems, we first conducted a theoretical analysis and found that the average idle slot spacing is a relevant indicator for traffic load. By using the average idle slot spacing and the number of neighbor nodes, the optimal CW (Contention Window) required to achieve high throughput can be configured. This paper compares the simulation results with those of conventional methods and evaluates them in terms of throughput, retransmission attempts, fairness, delay time, and number of collisions. The overall evaluation shows that the MAC protocol proposed in this paper has a better performance than the conventional method.

End-to-end test automation for web applications is important in order to release software quickly in accordance with market changes. However, the cost of implementing and maintaining test scripts is a major obstacle to the introduction of test automation. In addition, many testing activities, such as exploratory testing, user-interface testing, and usability testing, rely on human resources. We propose an approach to generate effective test scripts from manual testing, which is indispensable in software development. Manual testing activities are recorded by our tool. The generated test scripts leverage the page-object pattern, which improves the maintainability of test scripts. To generate page objects, our approach extracts operations as methods useful for test automation from the test logs. Our approach also generates test cases that cover features of an application by analyzing its page transitions. We evaluated whether our approach could generate complete test scripts from test logs obtained from four testers. Our experimental results indicate that our approach can generate a greater number of complete methods in page objects than a current page-object generation approach. We also conducted an empirical evaluation of whether our approach can reduce the cost of implementing test scripts for real systems. The result showed that our approach reduces about 48% of the time required to implement test scripts compared with manual implementation.

We introduce an edge routing decision problem called the police officer patrol problem (POPP), which is related to the vertex cover problem. A vertex cover of a graph can be regarded as the placement of police officers or fixed surveillance cameras so that each street of a neighborhood represented by the graph can be confirmed visually without moving from their position. In the edge routing problem we consider, a single police officer must confirm all the streets. The officer is allowed to move, but can confirm any street visually from an incident intersection without traversing it. In this paper, we show that the POPP on mixed graphs is NP-complete.

Flashcard systems are effective tools for learning words but have their limitations in teaching word usage. To overcome this problem, we suggest that a flashcard system shows a new example sentence on each repetition. This extension requires high-quality example sentences, automatically extracted from a huge corpus. To do this, we use a Determinantal Point Process which scales well to large data and allows us to naturally represent sentence similarity and quality as features. Our human evaluation experiment on the Japanese language indicates that the proposed method successfully extracted high-quality example sentences.

LSM-tree-based key-value stores (KVSs) with KV separation minimizes the write latency and write amplification of LSM-trees by writing the KV-pairs sequentially into a log file and indexing these KV pairs in an LSM-tree. As a tradeoff, the storage device must handle the parallel writes directly from the client, which can be detrimental to some storage devices such as non-volatile memory (NVM). Such tasks can easily overload the underlying memory buffers of NVM, especially when these KVSs are configured without a filesystem to manage IO congestion and minimize the latency during parallel workloads. In this paper, we discuss how to maximize the performance of NVM in LSM-trees with KV separation called Non-Volatile KVS (NVKVS). This approach features several optimizations such as asynchronous multithreading that decouples the client and write threads and reusable write-ahead log (WAL). With these optimizations, NVKVS has three benefits. First, it minimizes the write latency by using NVM. Second, it can saturate the NVM throughput through a user-configurable worker pool and optimizations. Third, it leverages the endurance of NVM to maximize the lifespan of the SSD. Our experimental testings show that NVKVS produces better latencies and over double the throughput of other the tested KVS.

Using existing resources to create teaching materials can save effort and achieve the desired quality easily. Yet while some resources can be used freely for educational purposes, others such as textbooks or online course materials cannot. This is a particular problem during a pandemic when much teaching has gone online and the risk of teachers violating copyright is even higher. Therefore, a solution that facilitates the usage of copyright-restricted resources for generating teaching materials with royalty sharing is needed. Our work exploits the advantage of blockchain technology and proposes a system to bond participants with a smart contract; it securely registers records of multiple authorships and contribution distribution of a teaching material that reuses in part, existing resources. Such records can be used as authorship evidence to claim economic benefits when a material is used. We implement TMchain on Ethereum-Remix IDE with a core smart contract. To lower the cost of using blockchain, the material files are stored off-chain and tied to the word processing system for the final authorship and contribution share determination when a material is completed. Furthermore, we test TMchain with teaching material creation scenarios to demonstrate its effective and practical potential.

Assessing whether an ungraded second language learner can read a given text quickly is important for supporting learners of diverse backgrounds. Second language acquisition (SLA) studies have tackled such assessment tasks wherein only a single short vocabulary test result is available to assess a learner. Such studies have shown that the text-coverage or namely the percentage of words the learner knows in the text, is the key assessment measure. Currently, count-based percentages are used, in which each word in the given text is classified as being known/unknown to the learner, and the words classified as known are then simply counted. When each word is classified, we can also obtain an uncertainty value as to how likely each word is known to the learner. However, how to leverage these informative values to guarantee their use as an assessment measure that is comparable to that of the previous values remains unclear. We propose a novel framework that allows assessment methods to be uncertainty-aware while guaranteeing comparability to the text-coverage threshold. Such methods involve a computationally complex problem for which we also propose a practical algorithm. In our evaluation using newly created crowdsourcing-based dataset, our best method under our framework outperformed conventional methods.

The number of investors holding risky assets in Japan is much lower than that in western countries even though it is an effective way for building investor assets. Although Japanese investment companies offer a service to invest in points through coalition loyalty programs instead of actual currency to address this situation, the problem still persists. One reason for this is that novice investors do not know in which stocks to invest. One possible solution is recommending stocks; however, we still face the cold-start problem because there is no transaction history for novice investors. In this study, we propose a novel content-based recommendation approach that utilizes touchpoint information, e.g., payment and app usage data, on smartphones in daily life. This approach employs user-weighted recency, frequency, and monetary, called UW-RFM and a complementary module to comply with Japanese guidelines that prohibit presenting only a small number of companies and establishing a minimum number of companies to be presented. We conduct an online evaluation to validate the effectiveness of the proposed approach in an actual investment service. The evaluation results show that the proposed approach motivates users to invest more, i.e., 0.352 more clicks on the recommendation area and 3, 016 points (yen), than the baseline method that does not consider touchpoint information.

The growth of user-generated content service platforms has led to people relying on user-generated content (UGC) rather than search engines when searching for and accessing information on the web. Attackers can also use UGC on a UGC service platform to disseminate web-based social engineering (SE) attacks to a large number of people. In this paper, we focus on an event-synced navigation attack, a type of web-based SE attack that generates UGC with links to malicious websites and distributes it synced with a real-life event at a specific time. To understand the attacks in the wild, we propose a three-step system to detect event-synced navigation attacks in real time by capturing the inevitable footprints left by attackers. We evaluate each step of the proposed system and determine that the proposed system can classify malicious and non-malicious UGC with 97% accuracy. In addition, we performed a comprehensive measurement study on event-synced navigation attacks spread from popular UGC platforms. We found that 34.1% of the fully qualified domain names of malicious websites associated with the event-synced navigation attack were spread from two or more UGC platforms. Finally, we also found that 87.8% of FQDN associated with well-known type of malicious websites (i.e., information theft, survey scams, suspicious browser plugin installations, etc.) survive for more than 100 days and that countermeasures taken by the UGC platform only covered 31.0% of the malicious UGC we detected in this study even though the malicious websites were accessed frequently.

Video-guided machine translation (VMT) is a type of multimodal machine translation that uses information from videos to guide translation. However, in the VMT 2020 challenge, adding videos only marginally improved the performance of VMT models compared to their text-only baselines. In this study, we systematically analyze why videos did not guide translation. Specifically, we evaluate the models in input degradation and visual sensitivity experiments and compare the results with a human evaluation using VATEX, which is the dataset used in the VMT 2020 challenge. The results indicate that short and straightforward video descriptions in VATEX are sufficient to perform the translations, which renders the videos redundant in the process. Based on our findings, we provide suggestions on the design of future VMT datasets. Code and human-evaluated data are publicly available for future research.

Deep learning has been widely used in natural language processing (NLP) such as document classification. For example, self-attention has achieved significant improvement in NLP. However, it has been pointed out that although deep learning accurately classifies documents, it is difficult for users to interpret the basis of the decision. In this paper, we focus on the task of classifying open-data news documents by their theme with a deep neural network with self-attention. We then propose methods for providing the interpretability for these classifications. First, we classify news documents by LSTM with a self-attention mechanism and then show that the network can classify documents highly accurately. Second, we propose five methods for providing the basis of the decision by focusing on various values, e.g., attention, the gradient between input and output values of a neural network, and classification results of a document with one word. Finally, we evaluate the performance of these methods in four evaluating ways and show that these methods can present interpretability suitably. In particular, the methods based on documents with one word can provide interpretability, which is extracting the words that have a strong influence on the classification results.

In this work we construct a database of the Japanese Imperial Diet. The Imperial Diet was established in 1890 after the promulgation of the Constitution of the Empire of Japan and its historical analysis is crucial to understand the actual functioning of the Japanese Diet. Since the minutes of the Imperial Diet were publicly available only in image format, textization is an imperative process for further analysis. Following the recent advancement of the deep neural networks (DNNs), especially in the character recognition, we apply DNNs to construct a text database of the Imperial Diet. In the course of textization, we trained DNNs using multiple datasets while introducing a novel approach of applying separate batch normalization to datasets. The results of the tentative analysis show a significant potential to deepen our understanding of the development of parliamentary democracy in Japan.

Information and communication technology (ICT) has become a potent tool with practical applications in various sectors, including security, transportation, agriculture, and fisheries. This paper describes the implementation and potential application of ICT to support sustainable fisheries management in sardine fishery in Bali, Indonesia. Generally, the adaptive management framework in fisheries management consists of data collection, stock assessment, management procedures, management measure development and implementation, monitoring, and evaluation. In every step of the adaptive management process, ICT can be applied. A conceptual framework and its application in the Bali Strait reveal the successful utilization of ICT to generate digital capture fisheries data through participatory data collection using the MICT-L (Marine ICT-Landing) digital catch landing data recorder, IoT (Internet of Things) GPS tracking devices, and a newly established online database. Digitalizing capture fisheries data is important to achieve robust stock assessment and to guide policy recommendations and management measures based on key indicators, such as total allowable catch (TAC) and spatial catch productivity. Furthermore, to emphasize the practical implications of our results, we describe how the system increases engagement by fisheries managers and other stakeholders and improves understanding and awareness about sustainable fish resource utilization by displaying indicators in a smart dashboard for daily monitoring. We also describe technical constraints and issues during system implementation such as GPS tracker transmission problems as well as non-technical considerations such as the willingness of fishers to participate, which is important to achieve successful implementation.

Jaw deformity is a growth and developmental disorder that indicates malocclusion accompanied by deformity of the upper and lower jaw, which both oral and maxillofacial surgery concern. Planning and evaluation of orthognathic surgery have based on hard tissue such as jaw bones and teeth, because they were suitable for incorporating tomographic imaging technologies. But the changes in the soft tissues that cover the repositioned jaw bones and teeth vary among individuals and have been left to the empirical decisions of the surgeons. Possible problems in the facial appearance after surgery, however, throw even greater negative impact on the patients' quality of life. Soft tissues need more accurate treatment to be considered. Silhouettes of patients' profile faces used for evaluating soft tissues in the past do not necessarily show cheek bulges that are critical to considering the factor of personality. This research proposes a computer-assisted image analysis method that is capable of quantitatively measuring the effect of orthognathic surgery over soft tissues. It reads a cephalometric image of the patient's profile face before and after repositioning jaw bones by surgery, locates the cheek region, and measures both the amount and the direction of cheek due to changes of soft tissue. The first experiment quantified the results of orthognathic surgery in the cases of both mandible (lower jaw) deficiency (Class II) and maxilla (upper jaw) deficiency (Class III) and visualized the changes in both regions of cheek and mouth using the proposed method. The second experiment calculated the changes in both cheek and mouth regions and compared those averages over patients between two surgey conditions, i.e. one conducted genioplasty for chin reposition in maxilla adjustment for treating Class III and the other did not. The proposed method quantitatively clarified how the soft tissues of cheek and mouth changed after surgery when the maxilla had moved in the treatment and the result was consistent with what oral surgeons have recognized only by looking in the past.