Automatic recognition of forms to efficiently utilize form data has become an important requirement in areas such as business and management. Accurately segmenting the table structure is a key step in the automatic table recognition technology. However, when facing complex forms with non-standard layouts, complex backgrounds, and distorted and fuzzy forms, the traditional table structure segmentation methods still have the problem of low accuracy. So to solve the problem, this study proposes a table structure segmentation method based on multi-scale feature enhancement. Firstly, we add a convolutional neural network (CNN) branch based on the framework of Swin-U net semantic segmentation model to extract form image features from three different resolution scales, namely, low, medium, and high; secondly, we introduce a Multi-Scale Feature Attention (MFA) mechanism in the branch of Swin-transformer to achieve the more accurate capture of the three-scale features; and finally, a multi-scale feature fusion module-CTFM is constructed to fuse the extracted features to enhance the segmentation ability of the model on the table structure. We trained our model on one of our own manually collected labeled complex table datasets (CWTD) and validated it with publicly available SciTSR and PubTabNet data, where our method achieves the best Dice_coef scores of 96.72% and 96.52% respectively, proving that our method has more accurate segmentation.

Offensive language is a significant detriment to social media environments. Existing research predominantly assumes monolingual expression, overlooking the prevalent behavior of code-switching (CS). To address this critical knowledge gap, this study identifies and empirically validates the distinct stylometric characteristics of code-switched (CSed) offensive language. Additionally, we developed methods to construct the first social media dataset specifically for CSed offensive content. Our analysis of this dataset reveals that CSed offensive language exhibits unique stylometric characteristics; moreover, these characteristics vary between the language segments involved in the CS. Furthermore, incorporating these features significantly enhances the performance of offensive language detection models. These findings offer significant research and practical implications for social media researchers, platforms, moderators, and users.

Entering text on a smartwatch is challenging due to the difficulty of tapping tiny keys. This study introduces a novel keyboard, Tap’nSwipe, to address the challenge. The keyboard features nine areas, each containing up to four characters. To enter a character, users swipe in a specific direction within the area containing the character, freeing them from precisely tapping on the target key. In addition, Tap’nSwipe leverages word predictions to enter words by allowing users to tap anywhere in the areas containing the target characters. The results of a user experiment show that Tap’nSwipe improves text entry accuracy and reduces error correction efforts when entering random character strings on a small smartwatch screen, with a marginal decrease in tying speed compared to QWERTY. Participants rated Tap’nSwipe higher than QWERTY in perceived character clarity, interaction precision, typing accuracy, typing efficiency, and ease of use. Participants also expressed strong interest in adopting Tap’nSwipe.

Aiming at the problem that the traditional detection method has low efficiency and accuracy due to the fact that there are many key component targets to be inspected by unmanned aerial vehicle (UAV) during power inspection, and the shape difference is large, and the image quality is not high, a transmission line abnormal target detection method based on improved YOLOv8 and super-resolution reconstruction is proposed. Firstly, the super-resolution reconstruction algorithm is used to reconstruct the abnormal image to improve the clarity and enrich the characteristic information contained in the image. On this basis, the improved YOLOv8 network is used to detect the defects in the inspection image. The CBAM attention mechanism is fused in the Bottleneck part of the C2f module to strengthen the model's ability to locate the target; In order to further improve the detection ability of small targets in patrol inspection, a small target detection layer is added to make the network pay more attention to the detection of small targets. Finally, in order to be deployed to the edge devices, the original convolution in the network is modified to a lightweight convolution GhostConv to reduce the number of model parameters. The experimental results show that the proposed method can accurately detect abnormal defects of transmission line components on the basis of improving the quality of inspection images. mAP is improved by 3.7% and the number of model parameters and calculation amount are greatly reduced, which reflects the effectiveness of the algorithm, and it has stronger extraction ability and robustness for subtle defect targets, meeting the detection requirements of power inspection.

Assistive technology is extremely important for maintaining and improving the elderly’s quality of life. Biometrics-based mobile user authentication (MUA) methods have witnessed rapid development in recent years owing to their usability and security benefits. However, there is a lack of a comprehensive review of such methods for the elderly. The primary objective of this research is to analyze the literature on state-of-the-art biometrics-based MUA methods via the lens of elderly users’ accessibility needs. In addition, conducting an MUA user study with elderly participants faces significant challenges, and it remains unclear how the performance of the elderly compares with non-elderly users in biometrics-based MUA. To this end, this research summarizes method design principles for user studies involving elderly participants and reveals the performance of elderly users relative to non-elderly users in biometrics-based MUA. The article also identifies open research issues and provides suggestions for the design of effective and accessible biometrics-based MUA methods for the elderly.

Shoulder-surfing studies in the context of mobile user authentication have focused on evaluating the attackers' performance, yet have paid much less attention to their perception of the shoulder-surfing process. Whether and how the shoulder-surfing setting might affect the attackers' perception remains under-explored. This study aims to investigate the perception of shoulder surfers with two different password-based mobile user authentication methods and three different observation angles. Moreover, this work examines the relationship between the attackers' perception and performance in shoulder surfing and the possible moderating effect of the authentication method for the first time. Based on the data collected from an online experiment, our analysis results reveal the effects of authentication methods and observation angles on the attackers' perception in terms of cognitive workload, observation clarity, and repetitive learning advantage. In addition, the results also show that the relationship between the attackers' cognitive workload and performance in shoulder surfing varies with the mobile user authentication method. Our findings not only deepen the understanding of shoulder-surfing attacks from an attacker's perspective, but also facilitate developing countermeasures for shoulder-surfing attacks.

Despite the widespread recognition that phishing attack is not a one-step process, existing studies on phishing victimization are dominated by binary decisions, which focus on understanding whether or not a user is victimized or susceptible to phishing attacks. There has been little empirical investigation into the stages of phishing victimization. Additionally, the literature presents mixed findings regarding influential factors in phishing victimization, which invite new explanations. Motivated by the potential of stage theorizing for effective anti-phishing recommendations, we propose a stage model for phishing victimization and examine the victimization stage's role in understanding the effects of privacy behavior, technical experience, and user demographics on victimization progression. We develop a series of hypotheses and test them by analyzing data collected from an embedded phishing awareness education program in a large organization. The findings provide strong evidence for the impacts of the stage model by improving our understanding of user characteristics and behaviors concerning different victimization stages and by offering fresh explanations for mixed findings about the effects of user characteristics in the literature. The findings lend themselves to a multitude of recommendations for improving the effectiveness of anti-phishing training.

Abstract Text classification is a common task in data science. Despite the superior performances of deep learning based models in various text classification tasks, their black‐box nature poses significant challenges for wide adoption. The knowledge‐to‐action framework emphasizes several principles concerning the application and use of knowledge, such as ease‐of‐use, customization, and feedback. With the guidance of the above principles and the properties of interpretable machine learning, we identify the design requirements for and propose an interpretable deep learning (IDeL) based framework for text classification models. IDeL comprises three main components: feature penetration, instance aggregation, and feature perturbation. We evaluate our implementation of the framework with two distinct case studies: fake news detection and social question categorization. The experiment results provide evidence for the efficacy of IDeL components in enhancing the interpretability of text classification models. Moreover, the findings are generalizable across binary and multi‐label, multi‐class classification problems. The proposed IDeL framework introduce a unique iField perspective for building trusted models in data science by improving the transparency and access to advanced black‐box models.

The pervasive use of mobile devices exposes users to an elevated risk of shoulder-surfing attacks. Despite the prior work on shoulder-surfing resistance of mobile user authentication methods, there is a lack of empirical studies on textual password authentication methods, particularly the hybrid passwords that integrate textual passwords with biometrics. To fill the literature gap, this research compares two hybrid password methods, touch-gesture- and keystroke-based passwords, with respect to their shoulder-surfing resistance performance. We select a touch-gesture-based password method that deploys multiple shoulder-surfing resistance strategies and a keystroke-based password method that leverages keystroke dynamics. To gain a holistic understanding of these password methods, we examine them under a variety of shoulder-surfing settings by varying interaction mode, observation angle, entry error, and observation effort. Going beyond effectiveness metrics, we also introduce efficiency metrics to assess shoulder-surfing resistance performance more comprehensively. We hypothesize and test the effects of shoulder-surfing settings by conducting both a longitudinal lab experiment and an online experiment with diversified participants. The results of both studies demonstrate the superior performance of the touch-gesture-based password method to the keystroke-based counterpart. The results also provide evidence for the effects of interaction mode, observation angle, and observation effort on shoulder-surfing resistance of hybrid passwords. Our findings offer suggestions for the design and strategies for strengthening the security of password authentication methods.

Domestic violence (DV) can lead to physical, psychological, and/or emotional consequences for its victims. Social media provides a new platform for DV victims to share their personal experiences and seek needed support. The anonymity of social media can potentially provide comfort and safety for victims to disclose their victimization experience. Despite a few efforts in detecting DV from social media, they have focused on differentiating DV-from non-DV-related content, or classifying DV-related content into a few general categories. By conducting an in-depth analysis of the content of DV self-disclosure in social media, this study characterizes DV in multiple aspects for the first time, including victim, perpetrator, relationship, and abuse. Moreover, it identifies the attributes to describe each aspect in detail. Furthermore, we use the social media data generated during the COVID-19 pandemic as a case study to understand the patterns of DV. The research findings of this study have implications for increasing the awareness of DV and designing support for DV victims.

Training on cues to deception is one of the promising ways of addressing humans’ poor performance in deception detection. However, the effect of training may be subject to the context of deception and the design of training. This study aims to investigate the effect of verbal cue training on the performance of phishing email detection by comparing different designs of training and examining the effect of topic familiarity. Based on the results of a lab experiment, we not only confirm the effect of training but also provide suggestions on how to design training to better facilitate the detection of phishing emails. In addition, our results also discover the effect of topic familiarity on phishing detection. The findings of this study have significant implications for the mitigation and intervention of online deception.

Password-based mobile user authentication is vulnerable to shoulder-surfing. Despite the increasing research on user password entry behavior and mobile security, there is limited understanding of how an adversary identifies a password through shoulder-surfing during mobile authentication. This study empirically examines the behaviors and strategies of password identification through shoulder-surfing with multiple observation attempts and from different observation distances. The results of analyzing data collected from a user study reveal the strategies and dynamics of password identification behaviors. The findings have implications for enhancing users’ password security and improving the design of mobile authentication methods.

We present deep learning method that can be used to reconstruct high-quality objects through turbulent media mixed with water and milk. The objects are placed behind turbulent media, and a series of speckle patterns are correspondingly recorded. By using many pairs of the recorded speckle patterns and input object images, a designed convolutional neural network (CNN) is fully trained, and then enables the recorded speckle patterns to be processed in real time. The proposed method is promising for imaging through turbulent media, and it is also believed that the proposed method can be applicable in many areas, e.g., imaging and information optics (such as optical encoding).

Social media has become a powerful and efficient platform for information diffusion. The increasing pervasiveness of social media use, however, has brought about the problems of fraudulent accounts that are intended to diffuse misinformation or malicious contents. Twitter recently released comprehensive archives of fraudulent tweets that are possibly connected to a propaganda effort of Internet Research Agency (IRA) on the 2016 U.S. presidential election. To understand information diffusion in fraudulent networks, we analyze structural properties of the IRA retweet network, and develop deep neural network models to detect fraudulent tweets. The structure analysis reveals key characteristics of the fraudulent network. The experiment results demonstrate the superior performance of the deep learning technique to a traditional classification method in detecting fraudulent tweets. The findings have potential implications for curbing online misinformation.

As it becomes commonplace to use mobile devices to store personal and sensitive data, mobile user authentication (MUA) methods have witnessed significant advancement to improve data and device security. On the other hand, traditional MUA methods such as password (or passcode) are still being widely deployed. Despite the growing body of knowledge on technical strengths and security vulnerabilities of various MUA methods, the perception of mobile users may be different, which can play a decisive role in MUA adoption. Additionally, user preferences for MUA methods may be subject to the influence of their demographic factors and device types. Furthermore, the pervasive use of mobile devices has generated many situations that create new usability and security needs of MUA methods such as support of one-handed and/or sight-free interaction. This study investigates user perception and situational needs of MUA methods using a survey questionnaire. The research findings can guide the design and selection of MUA methods.

Analyzing the disease data from the view of combinatorial features may better characterize the disease phenotype. In this study, a novel method is proposed to construct feature combinations and a classification model (CFC-CM) by mining key feature relationships. CFC-CM iteratively tests for differences in the feature relationship between different groups. To do this, it uses a modified $k$k-top-scoring pair (M-$k$k-TSP) algorithm and then selects the most discriminative feature pairs in the current feature set to infer the combinatorial features and build the classification model. Compared with support vector machines, random forests, least absolute shrinkage and selection operator, elastic net, and M-$k$k-TSP, the superior performance of CFC-CM on nine public gene expression datasets validates its potential for more precise identification of complex diseases. Subsequently, CFC-CM was applied to two metabolomics datasets, it obtained accuracy rates of $88.73\pm 2.06\%$88.73±2.06% and $79.11\pm 2.70\%$79.11±2.70% in distinguishing between hepatocellular carcinoma and hepatic cirrhosis groups and between acute kidney injury (AKI) and non-AKI samples, results superior to those of the other five methods. In summary, the better results of CFC-CM show that in contrast to molecules and combinations constituted by just two features, the combinations inferred by appropriate number of features could better identify the complex diseases.

Despite the increasing use of mobile web in our everyday lives, mobile web search remains a challenging task mainly due to the intensive scrolling inherent in linear presentation of search engine results (SERs) on the small screen of a mobile device. Drawing on the cognitive load theory and information foraging theory, this study aims to improve user performance in mobile web search by proposing two new artifacts—radial visualization of SERs clusters and synchronization of clusters and individual SERs presentations. These artifacts are hypothesized to improve search efficiency and/or navigation efficiency without affecting search effectiveness. This study also examines a possible moderating effect of search task type (open-ended vs. close-ended tasks). We implemented four mobile web search systems that reflect four combinations of different visualization and synchronization settings and empirically evaluated them using a controlled experiment. The results show that the radial visualization improves search efficiency and navigation efficiency, particularly for the open-ended tasks. In addition, synchronous presentation improves navigation efficiency in terms of path similarity.

Cyberbullying is becoming an epidemic problem in adolescents. The accelerated diffusion of information in an online environment exposes bullying messages to a large group of bystanders who witness such incidents. However, bystanders have received much less research attention than the aggressors and victims in the cyberbullying. The current research aims to understand whether adolescents' empathy has an impact on their positive by standing behavior - cyber bullied support behavior. Drawing on related theories and models, we proposed two hypotheses about the effect of empathy. A mixture of survey questionnaire and focus group was used to test the hypotheses. The analysis results of the survey data provided support for the effect of cognitive empathy but not affective empathy. The focus group study revealed that adolescent bystanders preferred offering indirect support to the cyber bullied by reporting to adults over direct intervention. The findings highlight the important roles of empathy training as well as teacher and parent intervention in preventing and curtaining cyberbullying both within and beyond the school environments.

This research aims to compare the performance of automatic deception detection in synchronous computer-mediated communication (CMC) with and without incorporating structural features. In addition, the development of deception detection models draws on the linguistic features that have been widely studied in the online deception literature. The results suggest that structural features can be effective in detecting deception and combining the structural features with linguistic features can improve the performance of detecting deception.

Online deception is fuelled by the escalated penetration of the Internet and social media. As the threat of online deception increases, understanding deception behaviour and underlying strategies is having a greater social impact. The verbal behaviour of online deception has recently been extended to the discourse level; nevertheless, discourse behaviours have been examined in isolation without referring to other behaviours in the discourse. By conceptualising the discourse of online behaviour as a social network (DOBNet), this research investigates possible impacts of deception intent on the central structures of DOBNet at three different levels: the discourse behaviour, subnetwork, and whole network. The empirical results of discourse network analysis and statistical tests provide partial support for each of the hypothesised effects. The findings not only demonstrate the efficacy of discourse in distinguishing deceivers from truth-tellers but also extend deception theories by confirming deception strategies from the perspective of discourse network and by uncovering unique characteristics of online deception strategies.

Understanding text requires not only the extraction of individual concepts, but the identification of semantic relationships among concepts as well. Lexical resources have been applied to analyzing text in a wide range of applications. However, manual compilation of lexical resources is difficult to keep up with the rapid increase of the volume and diversity of user-generated content on the web. Automatic concept hierarchy construction has been considered as one solution to the above problem. Despite extensive effort on automatic construction of concept hierarchies, few studies have focused on the concepts of specific domains. In this study, we propose a comprehensive framework for building a domain-specific concept hierarchy. By synthesizing different types of measurements of relatedness among concepts, we propose an integrated method for building a multi-branch hierarchy of product features from online consumer reviews. The experiment results show that the proposed algorithm successfully reconstructs almost an entire hierarchy except for missing a few concepts and links. Starting from scratch, the algorithm reconstructed about 60% of the manually constructed hierarchy. The proposed method can be used to improve search results by better understanding user queries, and to facilitate personalized recommendations in e-commerce.

Multimedia content in websites poses a variety of problems to mobile web users, such as content incompatible with mobile devices or failing to meet the accessibility needs of users with disabilities. This study proposes and evaluates a user-centered, device-aware approach to multimedia content adaptation for mobile Web to address those problems. It uniquely takes into consideration not only users' preferences and accessibility needs, but also capabilities of multimedia support of individual mobile devices. Evaluation shows that the proposed approach is accurate and efficient. This study provides insights for research on context-aware mobile computing, adaptive interfaces, and mobile HCI.

The explosive growth of mobile handheld devices users and the increasing need of ubiquitous information access make the use of mobile Web a common practice. The effectiveness of mobile web is contingent upon many factors. Among them, visual properties such as colors of web content can significantly influence user experience. However, website designers may not be a usability expert or even be aware of color impacts on the perception and experience of users, particularly on those with color vision deficiency (CVD). As a result, using colors that are difficult to distinguish by CVD people within the same web pages could cause significant problems in recognition and comprehension of web content. Existing color adaptation research has focused on regular web but yet to address the unique challenges of mobile web such as limited computing power and battery capacity. To improve the accessibility of mobile web, this research aims to adapt colors in mobile web pages that are non-distinguishable to CVD users while preserving some properties of original colors. In addition to web color replacement, we propose two methods to improve the efficiency of color adaptation to overcome the resource constraint of mobile devices. One method narrows down colors that need adaptation by identifying foreground and background color pairs. The other reduces the computational cost of assessing a set of replacement colors by introducing differential naturalness and differentiability. The effectiveness of the proposed methods was demonstrated with both theoretical proof and experiment results.

As criminals and terrorist employ social media platforms for planning and executing nefarious activities, understanding the degree of trustworthiness in interactions among actors becomes crucial for detecting their activities. Measuring trust in these environments can benefit analysts who are monitoring web forums to detect criminal or terrorist activities. Previous research proposed a trust model that could enable automatic trust discovery using speech act theory. This paper introduces a new classification method that enriches traditional techniques with contextual information. We conducted experiments to compare the proposed method with traditional approaches. The results show that the proposed method outperforms other alternative methods.

Most of the extant studies about web spam detection either explicitly or implicitly assume that web spam detection is performed on the server side of a search engine. In this paper, we argue that in some scenarios, web spam detection is preferred to be conducted on the client side (e.g., intelligent web browsers). When a page is viewed using an intelligent web browser, an integrated personalized web spam detector can determine whether the page is spam or not specifically tailored to the user's judgements. We propose a framework for implementing a personalized web spam detector. The experimental results obtained from an empirical evaluation confirmed the effectiveness of the proposed personalized web spam detection method.

Semantic similarity is vital to many areas, such as information retrieval. Various methods have been proposed with a focus on comparing unstructured text documents. Several of these have been enhanced with ontology; however, they have not been applied to ontology instances. With the growth in ontology instance data published online through, for example, Linked Open Data, there is an increasing need to apply semantic similarity to ontology instances. Drawing on ontology‐supported polarity mining (OSPM), we propose an algorithm that enhances the computation of semantic similarity with polarity mining techniques. The algorithm is evaluated with online customer review data. The experimental results show that the proposed algorithm outperforms the baseline algorithm in multiple settings.

Logistics network design is a strategic issue for nearly all companies. In the past, a logistics network once established was used for several periods. However, recent years have seen a rapidly changing economic environment. Trends in markets, shifts of customer clusters, changes of cost structures, and increasing competition are some of the aspects compelling companies to periodically redesign their logistics networks. In this paper, a complicated logistics network redesign problem is studied. To ensure the service levels of the important customers and products, weighting factors are employed, and a new service level measurement based on differentiation is proposed. Then a multi-objective model is formulated to redesign the complicated logistics network, and the NSGAII based solution procedure with determinant-code encoding structure is developed to achieve the Pareto solutions. The applicability of the proposed approach is demonstrated with the help of a real world case study from a Chinese automobile enterprise.

Many computer-based communication media offer visual anonymity. As a result, detecting online deception tends to be more difficult relative to traditional non-mediated communication. The state of the art research on online deception has focused on using linear statistical approaches to identifying behavioral differences between deceivers and truth-tellers. However, deception behaviors are not linear because deceivers may adopt dynamic strategies when they are motivated to succeed, and deceivers could disguise themselves to maximize their payoffs. Given such backdrop, this research is aimed to address deception strategies with a game theory approach. The results of an empirical study with a multi-stage game show that deceivers tend to select different strategies from truth-tellers and deceivers may adjust their strategies to avoid detection. These findings provide significant implications for explaining online deception in the full rationality paradigm.

Although speech recognition technology has made remarkable progress, its wide adoption is still restricted by notable effort made and frustration experienced by users while correcting speech recognition errors. One of the promising ways to improve error correction is by providing user support. Although support mechanisms have been proposed for inline speech recognition error correction, how to develop user support for third-party error correction has been under studied. To address the unique challenges associated with third-party error correction, external information that is obtained outside of an erroneous sentence was employed in this research. Specifically, three types of external information were selected, including word alternative hypotheses, noisy context and accurate context, and their impacts on dictation speech recognition error correction were assessed empirically. As expected, results revealed the importance of context information to improving both the performance and perception of error correction. Additionally, the results also provided insights into possible effects of word error rate and sentence length on error correction. These findings have significant implications to interface design for transcript proofreading systems.

Although speech recognition has improved significantly in recent years, its adoption continues to be limited, in part, by the effort and frustration associated with correcting speech recognition errors. Error detection is a particularly challenging issue in third-party error correction where different individuals are responsible for the original dictation and correcting the resulting text. This research aims to address the difficulty experienced in third-party error detection by developing and evaluating a variety of support mechanisms. Drawing on a growing body of literature on human computer interaction and speech recognition, four support mechanisms were designed and evaluated, namely indexed audio, speech summarization, error prediction, and the presentation of alternative hypotheses. A user study assessed the impact of these support mechanisms on both performance and perceptions during error detection tasks. Performance measures included effectiveness and efficiency, and perception measures included confidence, perceived usefulness, and cognitive workload. The results provide strong support for the use of indexed audio in the context of third-party error detection. The results also confirm that consecutive error rate, or the percentage of recognition errors immediately adjacent to another error, has a negative impact on the effectiveness of third-party error detection. Other support mechanisms failed to improve either effectiveness or perceptions, but they did negate the negative impact as consecutive error rate increased. These findings have significant implications for speech recognition error detection research and the design of error detection support solutions.

Abstract Polarity mining provides an in‐depth analysis of semantic orientations of text information. Motivated by its success in the area of topic mining, we propose an ontology‐supported polarity mining (OSPM) approach. The approach aims to enhance polarity mining with ontology by providing detailed topic‐specific information. OSPM was evaluated in the movie review domain using both supervised and unsupervised techniques. Results revealed that OSPM outperformed the baseline method without ontology support. The findings of this study not only advance the state of polarity mining research but also shed light on future research directions.

An analysis of the process and human cognitive model of deception detection (DD) shows that DD is infused with uncertainty, especially in high-stake situations. There is a recent trend toward automating DD in computer-mediated communication. However, extant approaches to automatic DD overlook the importance of representation and reasoning under uncertainty in DD. They represent uncertain cues as crisp values and can only infer whether deception occurs, but not to what extent deception occurs. Based on uncertainty theories and the analyses of uncertainty in DD, we propose a model to represent cues and to reason for DD under uncertainty, and address the uncertainty due to imprecision and vagueness in DD using fuzzy sets and fuzzy logic. Neuro-fuzzy models were developed to discover knowledge for DD. The evaluation results on five data sets showed that the neuro-fuzzy method not only was a good alternative to traditional machine-learning techniques but also offered superior interpretability and reliability. Moreover, the gains of neuro-fuzzy systems over traditional systems became larger as the level of uncertainty associated with DD increased. The findings of this paper have theoretical, methodological, and practical implications to DD and fuzzy systems research.

Online deception is disrupting our daily life, organizational process, and even national security. Existing approaches to online deception detection follow a traditional paradigm by using a set of cues as antecedents for deception detection, which may be hindered by ineffective cue identification. Motivated by the strength of statistical language models (SLMs) in capturing the dependency of words in text without explicit feature extraction, we developed SLMs to detect online deception. We also addressed the data sparsity problem in building SLMs in general and in deception detection in specific using smoothing and vocabulary pruning techniques. The developed SLMs were evaluated empirically with diverse datasets. The results showed that the proposed SLM approach to deception detection outperformed a state-of-the-art text categorization method as well as traditional feature-based methods.

A natural language interface (NLI) enables the ease-of-use of information systems in performing sophisticated human – computer interaction. To address the challenges of mobile devices to user interaction in information management, we propose an NLI as a promising solution. In this paper, we review state-of-the-art NLI technologies and analyse user requirements for managing notable information on mobile devices. To minimize any technical difficulties arising from developing and improving the usability of NLI systems we develop general principles for NLI design, which fills in a gap in the literature. In order to satisfy user requirements for information management on mobile devices, we innovatively design NLI-enabled information management architecture. It is shown from two usage scenarios that the architecture could lead to reduced effort in user navigation and improved efficiency and effectiveness of managing information on mobile devices. We conclude the article with the implications of this study and suggestions for future direction.

Instant messaging (IM) is being widely used in the workplace and personal life at an accelerated rate. In this paper, we empirically investigate an extension of TPB (Theory of Planned Behaviour) model to explain why individuals use punctuation in IM. Conceptually, we examine attitude and behavioural belief and their impacts on punctuation usage. Our research model was tested in an empirical study, which confirmed all of our hypotheses. This paper also explains why people prefer to use punctuation in IM and how the usage is moderated by the communication context.

The importance of research on misinformation has received wide recognition. Two major challenges faced by this research community are the lack of theoretical models and the scarcity of misinformation in support of such research. This paper aims to address the aforementioned challenges by conceptualizing misinformation and enabling the interoperability of misinformation. In particular, a representation and a model of misinformation are proposed through surveying, synthesizing, and explicating existing work in the field. Moreover, ontology is used to represent the proposed model. The ontology-supported misinformation model can not only guide future misinformation research but also lay the foundation for building a digital misinformation library by advancing our knowledge on misinformation and by improving its sharing, management, and reuse. In addition, we present a formal methodology for managing misinformation in a digital library, and suggest future research directions related to the misinformation model.

Web service composition can enhance the efficiency and agility of knowledge management by composing individual Web services together for complex business requirements. There are two main research streams in knowledge representation for Web service composition: the syntactic-based approach and the semantic-based approach. Despite the promises brought by each approach, the two streams are largely separated from each other. In this article, we propose an integrated ontology-supported Web service composition framework, which provides a novel solution to organizational knowledge management. By synergistically leveraging both syntactic-based and semantic-based approaches, this framework provides dual modes to perform service composition. Ontologies are employed to enrich semantics at both the service description and composition levels. The proposed conceptual framework has been implemented in the corporate financial services domain. It is demonstrated that the shared ontology helps to fulfill automated and on-the-fly service composition in particular and knowledge management in general.

As more and more people join online communities, the ability to better understand members' roles becomes critical to preserving and improving the health of those communities. We propose a novel approach to identifying key members and their roles by discovering implicit knowledge from online communities. Viewing an online community as a social network connected by poster-poster relationships, the approach takes advantage of the strengths of social network analysis and weighting schemes from information retrieval in identifying key members. Experimental studies were carried out to empirically evaluate the proposed approach with real-world data collected from a Usenet bulletin board over a one year period. The results showed that the proposed approach can not only identify prominent members whose behaviors are community supportive but also filter chatters whose behaviors are superficial to the online community. The findings have broad implications for online communities by allowing moderators to better support their members and by enabling members to better understand the conversation space.

The efficiency promised by a dictation speech recognition (DSR) system is lessened by the need for correcting recognition errors. Error detection is the precursor of error correction. Developing effective techniques for error detection can thus lead to improved error correction. Current research on error detection has focused mainly on transcription and/or domain-specific speech. Error detection in DSR has been studied less. We propose data mining models for detecting errors in DSR. Instead of relying on internal parameters from DSR systems, we propose a loosely coupled approach to error detection based on features extracted from the DSR output. The features mainly came from two sources: confidence scores and linguistics parsing. Link grammar was innovatively applied to error detection. Three data mining techniques, including Na/spl inodot//spl uml/ve Bayes, neural networks, and Support Vector Machines (SVMs), were evaluated on 5M DSR corpora. The experimental results showed that significant performance was achieved in that F-measures for error detection ranged from 55.3% to 62.5%. This study provided insights into the merit of different data-mining techniques and different types of features in error detection.

With the increase of economic globalization and evolution of information technology, financial data are being generated and accumulated at an unprecedented pace. As a result, there has been a critical need for automated approaches to effective and efficient utilization of massive amount of financial data to support companies and individuals in strategic planning and investment decision-making. Data mining techniques have been used to uncover hidden patterns and predict future trends and behaviors in financial markets. The competitive advantages achieved by data mining include increased revenue, reduced cost, and much improved marketplace responsiveness and awareness. There has been a large body of research and practice focusing on exploring data mining techniques to solve financial problems. In this paper, we describe data mining in the context of financial application from both technical and application perspectives. In addition, we compare different data mining techniques and discuss important data mining issues involved in specific financial applications. Finally, we highlight a number of challenges and trends for future research in this area.

The semantic Web enables intelligent agents to "outsource" knowledge, extending and enhancing their limited knowledge bases. An open question is how agents can efficiently and effectively access the vast knowledge on the inherently open and dynamic semantic Web. The problem is not that of finding a source for desired information, but deciding which among many possibly inconsistent sources is most reliable. We propose an approach to agent knowledge outsourcing inspired by the use trust in human society. Trust is a type of social knowledge and encodes evaluations about which agents can be taken as reliable sources of information or services. We focus on two important practical issues: learning trust and justifying trust. An agent can learn trust relationships by reasoning about its direct interactions with other agents and about public or private reputation information, i.e., the aggregate trust evaluations of other agents. We use the term trust justification to describe the process in which an agent integrates the beliefs of other agents, trust information, and its own beliefs to update its trust model. We describe the results of simulation experiments of the use and evolution of trust in multiagent systems. Our experiments demonstrate that the use of explicit trust knowledge can significantly improve knowledge outsourcing performance. We also describe a collaborative trust justification technique that focuses on reducing search complexity, handling inconsistent knowledge, and avoiding error propagation.

Abstract The role of information retrieval (IR) in support of decision making and knowledge management has become increasingly significant. Confronted by various problems in traditional keyword‐based IR, many researchers have been investigating the potential of natural language processing (NLP) technologies. Despite widespread application of NLP in IR and high expectations that NLP can address the problems of traditional IR, research and development of an NLP component for an IR system still lacks support and guidance from a cohesive framework. In this paper, we propose a theoretical framework called NLPIR that aims at integrating NLP into IR and at generalizing broad application of NLP in IR. Some existing NLP techniques are described to validate the framework, which not only can be applied to current research, but is also envisioned to support future research and development in IR that involve NLP.

This paper introduces a hybrid method for abstracting Chinese text. It integrates the statistical approach with language understanding. Some linguistics heuristics and segmentation are also incorporated into the abstracting process. The prototype system is of a multipurpose type catering for various users with different requirements. Initial responses show that the proposed method contributes much to the flexibility and accuracy of the automatic Chinese abstracting system. In practice, the present work provides a path to developing an intelligent Chinese system for automating the information.

The paper presents a hybrid model integrating an example based (EB) approach with a knowledge base methodology in support of Chinese-English machine translation (MT). The EB approach normally performs translation by referring to similar source sentences in the example base. It can be employed at different stages of MT including source language analysis, source to target transformation and target language generation. The proposed MT model brings together knowledge in two forms: rules and examples. This system is intended for domains that are understood well but not perfectly. Expert and heuristic knowledge in the form of rules are incorporated to provide a skeletal method for solving problems. Examples are then used to flesh out the method by covering idiosyncrasies and special cases that are not anticipated by the rules. In addition to a reasonably accurate and efficient set of rules to serve as a starting point for problem solving, the system is provided with some specific similarity measurement for critical analysis of Chinese sentences. This is based on word grammatical features and integrated with some dynamic mechanism to improve flexibility and robustness during the translation process.