This study examines the appropriation and usage of voice assistants like Google Assistant or Amazon Alexa in private households. Our research is based on 10 in-depth interviews with users of voice assistants and a follow-up evaluation of their interaction histories. Our results illustrate situations in which the voice assistants were used at home, what strategies the users adopted to interact with them, how the interactions were performed, and what difficulties users experienced. A focus of our study is on misunderstandings, situations where interactions failed partially or completely. Our study shows that the full potential of voice assistants is often not utilized, as complex interactions are often subject to failures and users revert to simple use cases. Our participants used their voice assistant mostly for simple commands, often not even trying new functions. An analysis of their appropriation strategies resulted in implications for the design of supportive tools as well as the further development and optimization of voice interfaces.2020-09-06
Conversational agents (CAs) such as Siri, Alexa, and Google Assistant are increasingly penetrating everyday life. From a Human-Computer Interaction (HCI) perspective, designing CAs that appropriately support the way they are used within daily life is still challenging. While initial design guidelines for human-AI interaction exist, we still know little about how users actually perceive CAs within their daily lives and what aspects motivate their usage of such tools. Within our research, we therefore conducted an interview study with 29 participants to uncover daily positive and negative experiences with CAs. By revealing how users currently perceive CAs, we identify quality criteria that could inform their future design. By evaluating these criteria with respect to existing research discourses about user experience (UX) guidelines for CAs, we contribute to the field by extending these guidelines from an end-user's perspective.2020-09-06
The popularity of conversational agents (CAs) in the form of AI speakers that support ubiquitous smart homes has increased because of their seamless interaction. However, recent studies have revealed that the use of AI speakers decreases over time, which shows that current agents do not fully support smart homes. Because of this problem, the possibility of unobtrusive, invisible intelligence without a physical device has been suggested. To explore CA design direction that enhances the user experience in smart homes, we aimed to understand each feature by comparing an invisible agent with visible ones embedded in stand-alone AI speakers. We conducted a drawing study to examine users' mental models formed through communicating with two different physical entities (i.e., visible and invisible CAs). From the drawings, interviews, and surveys, we identified how users' mental models and interactions differed depending on the presence of a physical entity. We found that a physical entity affected users' perceptions, expectations, and interactions toward the agent.2020-09-04
Increasing number of researchers and designers are envisioning a wide range of novel proactive conversational services for smart speakers such as context-aware reminders and restocking household items. When initiating conversational interactions proactively, smart speakers need to consider users' contexts to minimize disruption. In this work, we aim to broaden our understanding of opportune moments for proactive conversational interactions in domestic contexts. Toward this goal, we built a voice-based experience sampling device and conducted a one-week field study with 40 participants living in university dormitories. From 3,572 in-situ user experience reports, we proposed 19 activity categories to investigate contextual factors related to interruptibility. Our data analysis results show that the key determinants for opportune moments are closely related to both personal contextual factors such as busyness, mood, and resource conflicts for dual-tasking, and the other contextual factors associated with the everyday routines at home, including user mobility and social presence. Based on these findings, we discuss the need for designing context-aware proactive conversation management features that dynamically control conversational interactions based on users' contexts and routines.2020-09-04
Clustering is a popular unsupervised data mining technique. It has been applied in various data mining and big data applications. Efficient clustering algorithms and implementation techniques are keys to cope with the scalability and performance requirements of big data analysis. This paper introduces the design and implementation of a density-based clustering algorithm that can deal with big data efficiently and effectively. We present a parallel Shared Nearest Neighbor (SNN) clustering algorithm using the k-dimensional tree (k-d tree) to reduce search time to improve efficiency. The proposed algorithm is implemented in a distributed environment using the Spark framework. The effectiveness of the proposed algorithm has been evaluated through a case study involving four data sets, Bristol Crime Stats, 911 call, Complex9, and TLC Trip datasets.2020-08-26
The massive amount of data and large variety of data distributions in the big data era call for access methods that are efficient in both query processing and index management, and over both practical and worst-case workloads. To address this need, we revisit two classic multidimensional access methods—the R-tree and the space-filling curve. We propose a novel R-tree packing strategy based on space-filling curves. This strategy produces R-trees with an asymptotically optimal I/O complexity for window queries in the worst case. Experiments show that our R-trees are highly efficient in querying both real and synthetic data of different distributions. The proposed strategy is also simple to parallelize, since it relies only on sorting. We propose a parallel algorithm for R-tree bulk-loading based on the proposed packing strategy and analyze its performance under the massively parallel communication model. To handle dynamic data updates, we further propose index update algorithms that process data insertions and deletions without compromising the optimal query I/O complexity. Experimental results confirm the effectiveness and efficiency of the proposed R-tree bulk-loading and updating algorithms over large data sets.2020-08-26
Wireless sensor networks play an important role in today's world: When measuring physical conditions, the quality of the sensor readings ultimately impacts the quality of various data analytical services. To maintain data correctness and quality, run-time measures such as anomaly detection techniques are gaining significance. In particular, the detection of threatening node anomalies caused by sensor node faults has become a crucial task.
The detection of faulty sensor nodes is a non-trivial task because wireless sensor networks typically consist of low-cost embedded systems with strictly limited resources, especially regarding their energy budget. Thus, efficient and lightweight approaches that meet the requirements of sensor networks are required.
In this SoK paper, we contribute with a novel taxonomy of anomaly detection approaches focused on wireless sensor networks and a meta-survey of related classification schemes. To the best of our knowledge, our taxonomy is a comprehensive super-set of all previously published taxonomies in this field. Based on the taxonomy, we present new insights in node-level anomaly detection approaches and the applicability of immune-inspired techniques, and we lay out related research challenges.2020-08-25
Containers are commonly used to run the data-intensive applications of different tenants in cloud infrastructures. The storage I/O of the colocated tenants is typically handled by the shared system kernel of the container host. When a data-intensive container competes with a noisy neighbor, the kernel I/O services can cause performance variability and slowdown. This is a challenging problem for which several approaches have already been explored. Although the dynamic resource allocation, kernel structure replication, and hardware-level virtualization are helpful, they incur costs of high implementation complexity and execution overhead. As a realistic cost-effective alternative, we isolate the I/O path of each tenant by running dedicated storage systems at user level on reserved resources. We introduce the libservices as a unified user-level storage abstraction to dynamically provision per tenant container root filesystems, application data filesystems and image repositories. We outline several examples of container storage systems whose clients and servers can be composed from libservices. With an early prototype, we successfully demonstrate that the libservices combine the required efficiency and flexibility to build isolated I/O services on multitenant hosts with superior performance over existing user-level or kernel-level systems.2020-08-24
In this paper, we report our recent practice at Tencent for user modeling based on mobile app usage. User behaviors on mobile app usage, including retention, installation, and uninstallation, can be a good indicator for both long-term and short-term interests of users. For example, if a user installs Snapseed recently, she might have a growing interest in photographing. Such information is valuable for numerous downstream applications, including advertising, recommendations, etc. Traditionally, user modeling from mobile app usage heavily relies on handcrafted feature engineering, which requires onerous human work for different downstream applications, and could be sub-optimal without domain experts. However, automatic user modeling based on mobile app usage faces unique challenges, including (1) retention, installation, and uninstallation are heterogeneous but need to be modeled collectively, (2) user behaviors are distributed unevenly over time, and (3) many long-tailed apps suffer from serious sparsity. In this paper, we present a tailored Auto Encoder-coupled Transformer Network (AETN), by which we overcome these challenges and achieve the goals of reducing manual efforts and boosting performance. We have deployed the model at Tencent, and both online/offline experiments from multiple domains of downstream applications have demonstrated the effectiveness of the output user embeddings.2020-08-23