As the world is flooded with deluge of data, the demand for mining data to gain insights is increasing. One effective technique to deal with the problem is to model the data as networks (graphs) and then apply graph mining techniques to uncover useful patterns. Several graph mining techniques have been studied in the literature, and graphlet-based analysis is gaining popularity due to its power in exposing hidden structure and interaction within the networks. The concept of graphlets for basic (undirected) networks was introduced around 2004 by Pržulj, et. al. [14]. Subsequently, graphlet based network analysis gained attraction when Pržulj added the concept of graphlet orbits and applied to biological networks [15]. A decade later, Sarajlić, et. al. introduced graphlets and graphlet orbits for directed networks, illustrating its application to fields beyond biology such as world trade networks, brain networks, communication networks, etc. [19]. Hence, directed graphlets are found to be more powerful in exposing hidden structures of the network than undirected graphlets of same size, due to added information on the edges. Taking this approach further, more recently, graphlets and orbits for signed networks have been introduced by Dale [3]. This paper presents a simple algorithm to enumerate signed graphlets and orbits. It then demonstrates an application of signed graphlets and orbits to a metabolic network.
{"title":"Algorithm and Application for Signed Graphlets","authors":"Apratim Das, A. Aravind, Mark Dale","doi":"10.1145/3341161.3343692","DOIUrl":"https://doi.org/10.1145/3341161.3343692","url":null,"abstract":"As the world is flooded with deluge of data, the demand for mining data to gain insights is increasing. One effective technique to deal with the problem is to model the data as networks (graphs) and then apply graph mining techniques to uncover useful patterns. Several graph mining techniques have been studied in the literature, and graphlet-based analysis is gaining popularity due to its power in exposing hidden structure and interaction within the networks. The concept of graphlets for basic (undirected) networks was introduced around 2004 by Pržulj, et. al. [14]. Subsequently, graphlet based network analysis gained attraction when Pržulj added the concept of graphlet orbits and applied to biological networks [15]. A decade later, Sarajlić, et. al. introduced graphlets and graphlet orbits for directed networks, illustrating its application to fields beyond biology such as world trade networks, brain networks, communication networks, etc. [19]. Hence, directed graphlets are found to be more powerful in exposing hidden structures of the network than undirected graphlets of same size, due to added information on the edges. Taking this approach further, more recently, graphlets and orbits for signed networks have been introduced by Dale [3]. This paper presents a simple algorithm to enumerate signed graphlets and orbits. It then demonstrates an application of signed graphlets and orbits to a metabolic network.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121919143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Justin Song, Valerie Spicer, Andrew J. Park, Herbert H. Tsang, P. Brantingham
The structure of the urban setting determines the crime patterns. This research explores the street profile analysis which is a new method for analyzing crime in relation to street networks. Street profile analysis can be used to identify crime surges or heavy concentrations of crime along roadways. In this study, the street profile technique is combined with a discrete calculus approach to locate the boundaries of small criminal spaces in the City of Vancouver, British Columbia, Canada. This experimental technique utilizes open source property crime data from the Vancouver Police Department to analyze crime patterns within Vancouver. This computational crime analysis technique is described in detail and the utility of this technique explored. The new technique is a valuable tool for the intelligence and security informatics communities.
{"title":"Computational Method for Identifying the Boundaries of Crime with Street Profile and Discrete Calculus","authors":"Justin Song, Valerie Spicer, Andrew J. Park, Herbert H. Tsang, P. Brantingham","doi":"10.1145/3341161.3343537","DOIUrl":"https://doi.org/10.1145/3341161.3343537","url":null,"abstract":"The structure of the urban setting determines the crime patterns. This research explores the street profile analysis which is a new method for analyzing crime in relation to street networks. Street profile analysis can be used to identify crime surges or heavy concentrations of crime along roadways. In this study, the street profile technique is combined with a discrete calculus approach to locate the boundaries of small criminal spaces in the City of Vancouver, British Columbia, Canada. This experimental technique utilizes open source property crime data from the Vancouver Police Department to analyze crime patterns within Vancouver. This computational crime analysis technique is described in detail and the utility of this technique explored. The new technique is a valuable tool for the intelligence and security informatics communities.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116761422","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Amir Pouran Ben Veyseh, M. Thai, Thien Huu Nguyen, D. Dou
Fake news and rumors constitute a major problem in social networks recently. Due to the fast information propagation in social networks, it is inefficient to use human labor to detect suspicious news. Automatic rumor detection is thus necessary to prevent devastating effects of rumors on the individuals and society. Previous work has shown that in addition to the content of the news/posts and their contexts (i.e., replies), the relations or connections among those components are important to boost the rumor detection performance. In order to induce such relations between posts and contexts, the prior work has mainly relied on the inherent structures of the social networks (e.g., direct replies), ignoring the potential semantic connections between those objects. In this work, we demonstrate that such semantic relations are also helpful as they can reveal the implicit structures to better capture the patterns in the contexts for rumor detection. We propose to employ the self-attention mechanism in neural text modeling to achieve the semantic structure induction for this problem. In addition, we introduce a novel method to preserve the important information of the main news/posts in the final representations of the entire threads to further improve the performance for rumor detection. Our method matches the main post representations and the thread representations by ensuring that they predict the same latent labels in a multitask learning framework. The extensive experiments demonstrate the effectiveness of the proposed model for rumor detection, yielding the state-of-the-art performance on recent datasets for this problem.
{"title":"Rumor Detection in Social Networks via Deep Contextual Modeling","authors":"Amir Pouran Ben Veyseh, M. Thai, Thien Huu Nguyen, D. Dou","doi":"10.1145/3341161.3342896","DOIUrl":"https://doi.org/10.1145/3341161.3342896","url":null,"abstract":"Fake news and rumors constitute a major problem in social networks recently. Due to the fast information propagation in social networks, it is inefficient to use human labor to detect suspicious news. Automatic rumor detection is thus necessary to prevent devastating effects of rumors on the individuals and society. Previous work has shown that in addition to the content of the news/posts and their contexts (i.e., replies), the relations or connections among those components are important to boost the rumor detection performance. In order to induce such relations between posts and contexts, the prior work has mainly relied on the inherent structures of the social networks (e.g., direct replies), ignoring the potential semantic connections between those objects. In this work, we demonstrate that such semantic relations are also helpful as they can reveal the implicit structures to better capture the patterns in the contexts for rumor detection. We propose to employ the self-attention mechanism in neural text modeling to achieve the semantic structure induction for this problem. In addition, we introduce a novel method to preserve the important information of the main news/posts in the final representations of the entire threads to further improve the performance for rumor detection. Our method matches the main post representations and the thread representations by ensuring that they predict the same latent labels in a multitask learning framework. The extensive experiments demonstrate the effectiveness of the proposed model for rumor detection, yielding the state-of-the-art performance on recent datasets for this problem.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130186993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Detecting diseases at early stage can help to overcome and treat them accurately. Identifying the appropriate treatment depends on the method that is used in diagnosing the diseases. A Clinical Decision Support System (CDS) can greatly help in identifying diseases and methods of treatment. In this paper we propose a CDS framework that can integrate heterogeneous health data from different sources, such as laboratory test results, basic information of patients, and health records. Using the electronic health medical data so collected, innovative machine learning and deep learning approaches are employed to implement a set of services to recommend a list of diseases and thus assist physicians in diagnosing or treating their patients health issues more efficiently.
{"title":"A clinical decision support framework for automatic disease diagnoses","authors":"C. Comito, Agostino Forestiero, Giuseppe Papuzzo","doi":"10.1145/3341161.3343509","DOIUrl":"https://doi.org/10.1145/3341161.3343509","url":null,"abstract":"Detecting diseases at early stage can help to overcome and treat them accurately. Identifying the appropriate treatment depends on the method that is used in diagnosing the diseases. A Clinical Decision Support System (CDS) can greatly help in identifying diseases and methods of treatment. In this paper we propose a CDS framework that can integrate heterogeneous health data from different sources, such as laboratory test results, basic information of patients, and health records. Using the electronic health medical data so collected, innovative machine learning and deep learning approaches are employed to implement a set of services to recommend a list of diseases and thus assist physicians in diagnosing or treating their patients health issues more efficiently.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"69 12","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131470999","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Dionisis Margaris, D. Spiliotopoulos, C. Vassilakis
Online advertising benefits by recommender systems since the latter analyse reviews and rating of products, providing useful insight of the buyer perception of products and services. When traditional recommender system information is enriched with social network information, more successful recommendations are produced, since more users' aspects are taken into consideration. However, social network information may be unavailable since some users may not have social network accounts or may not consent to their use for recommendations, while rating data may be unavailable due to the cold start phenomenon. In this paper, we propose an algorithm that combines limited collaborative filtering information, comprised only of users' ratings on items, with limited social network information, comprised only of users' social relations, in order to improve (1) prediction accuracy and (2) prediction coverage in collaborative filtering recommender systems, at the same time. The proposed algorithm considerably improves rating prediction accuracy and coverage, while it can be easily integrated in recommender systems.
{"title":"Social Relations versus Near Neighbours: Reliable Recommenders in Limited Information Social Network Collaborative Filtering for Online Advertising","authors":"Dionisis Margaris, D. Spiliotopoulos, C. Vassilakis","doi":"10.1145/3341161.3345620","DOIUrl":"https://doi.org/10.1145/3341161.3345620","url":null,"abstract":"Online advertising benefits by recommender systems since the latter analyse reviews and rating of products, providing useful insight of the buyer perception of products and services. When traditional recommender system information is enriched with social network information, more successful recommendations are produced, since more users' aspects are taken into consideration. However, social network information may be unavailable since some users may not have social network accounts or may not consent to their use for recommendations, while rating data may be unavailable due to the cold start phenomenon. In this paper, we propose an algorithm that combines limited collaborative filtering information, comprised only of users' ratings on items, with limited social network information, comprised only of users' social relations, in order to improve (1) prediction accuracy and (2) prediction coverage in collaborative filtering recommender systems, at the same time. The proposed algorithm considerably improves rating prediction accuracy and coverage, while it can be easily integrated in recommender systems.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130279755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Sheung Yat Law, D. Kasthurirathna, Piraveenan Mahendra
Social systems are increasingly being modelled as complex networks, and the interactions and decision making of individuals in such systems can be modelled using game theory. Therefore, networked game theory can be effectively used to model social dynamics. Individuals can use pure or mixed strategies in their decision making, and recent research has shown that there is a connection between the topological placement of an individual within a social network and the best strategy they can choose to maximise their returns. Therefore, if certain individuals have a preference to employ a certain strategy, they can be swapped or moved around within the social network to more desirable topological locations where their chosen strategies will be more effective. To this end, it has been shown that to increase the overall public good, the cooperators should be placed at the hubs, and the defectors should be placed at the peripheral nodes. In this paper, we tackle a related question, which is the time (or number of swaps) it takes for individuals who are randomly placed within the network to move to optimal topological locations which ensure that the public utility satisfies a certain utility threshold. We show that this time depends on the topology of the social network, and we analyse this topological dependence in terms of topological metrics such as scale-free exponent, assortativity, clustering coefficient, and Shannon information content. We show that the higher the scale-free exponent, the quicker the public utility threshold can be reached by swapping individuals from an initial random allocation. On the other hand, we find that assortativity has negative correlation with the time it takes to reach the public utility threshold. We find also that in terms of the correlation between information content and the time it takes to reach a public utility threshold from a random initial assignment, there is a bifurcation: one class of networks show a positive correlation, while another shows a negative correlation. Our results highlight that by designing networks with appropriate topological properties, one can minimise the need for the movement of individuals within a network before a certain public good threshold is achieved. This result has obvious implications for defence strategies in particular.
{"title":"Placement matters in making good decisions sooner: the influence of topology in reaching public utility thresholds","authors":"Sheung Yat Law, D. Kasthurirathna, Piraveenan Mahendra","doi":"10.1145/3341161.3343674","DOIUrl":"https://doi.org/10.1145/3341161.3343674","url":null,"abstract":"Social systems are increasingly being modelled as complex networks, and the interactions and decision making of individuals in such systems can be modelled using game theory. Therefore, networked game theory can be effectively used to model social dynamics. Individuals can use pure or mixed strategies in their decision making, and recent research has shown that there is a connection between the topological placement of an individual within a social network and the best strategy they can choose to maximise their returns. Therefore, if certain individuals have a preference to employ a certain strategy, they can be swapped or moved around within the social network to more desirable topological locations where their chosen strategies will be more effective. To this end, it has been shown that to increase the overall public good, the cooperators should be placed at the hubs, and the defectors should be placed at the peripheral nodes. In this paper, we tackle a related question, which is the time (or number of swaps) it takes for individuals who are randomly placed within the network to move to optimal topological locations which ensure that the public utility satisfies a certain utility threshold. We show that this time depends on the topology of the social network, and we analyse this topological dependence in terms of topological metrics such as scale-free exponent, assortativity, clustering coefficient, and Shannon information content. We show that the higher the scale-free exponent, the quicker the public utility threshold can be reached by swapping individuals from an initial random allocation. On the other hand, we find that assortativity has negative correlation with the time it takes to reach the public utility threshold. We find also that in terms of the correlation between information content and the time it takes to reach a public utility threshold from a random initial assignment, there is a bifurcation: one class of networks show a positive correlation, while another shows a negative correlation. Our results highlight that by designing networks with appropriate topological properties, one can minimise the need for the movement of individuals within a network before a certain public good threshold is achieved. This result has obvious implications for defence strategies in particular.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"1993 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131145538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The United Nations, in their annual World Drug Report in 2018, reported that the production of Opium, Cocaine, Cannabis, etc. all observed record highs, which indicates the ever-growing demand of these drugs. Social networks of individuals associated with Drug Trafficking Organizations (DTO) have been created and studied by various research groups to capture key individuals, in order to disrupt operations of a DTO. With drug offenses increasing globally, the list of suspect individuals has also been growing over the past decade. As it takes significant amount of technical and human resources to monitor a suspect, an increasing list entails higher resource requirements on the part of law enforcement agencies. Monitoring all the suspects soon becomes an impossible task. In this paper, we present a novel methodology which ensures reduction in resources on the part of law enforcement authorities, without compromising the ability to uniquely identify a suspect, when they become “active” in drug related activities. Our approach utilizes the mathematical notion of Identifying Codes, which generates unique identification for all the nodes in a network. We find that just monitoring important individuals in the network leads to a wastage in resources and show how our approach overcomes this shortcoming. Finally, we evaluate the efficacy of our approach on real world datasets.
{"title":"Monitoring Individuals in Drug Trafficking Organizations: A Social Network Analysis","authors":"K. Basu, Arunabha Sen","doi":"10.1145/3341161.3342938","DOIUrl":"https://doi.org/10.1145/3341161.3342938","url":null,"abstract":"The United Nations, in their annual World Drug Report in 2018, reported that the production of Opium, Cocaine, Cannabis, etc. all observed record highs, which indicates the ever-growing demand of these drugs. Social networks of individuals associated with Drug Trafficking Organizations (DTO) have been created and studied by various research groups to capture key individuals, in order to disrupt operations of a DTO. With drug offenses increasing globally, the list of suspect individuals has also been growing over the past decade. As it takes significant amount of technical and human resources to monitor a suspect, an increasing list entails higher resource requirements on the part of law enforcement agencies. Monitoring all the suspects soon becomes an impossible task. In this paper, we present a novel methodology which ensures reduction in resources on the part of law enforcement authorities, without compromising the ability to uniquely identify a suspect, when they become “active” in drug related activities. Our approach utilizes the mathematical notion of Identifying Codes, which generates unique identification for all the nodes in a network. We find that just monitoring important individuals in the network leads to a wastage in resources and show how our approach overcomes this shortcoming. Finally, we evaluate the efficacy of our approach on real world datasets.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"272 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133852291","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hankyu Jang, Samuel Justice, P. Polgreen, Alberto Maria Segre, Daniel K. Sewell, S. Pemmaraju
This paper presents a high-fidelity agent-based simulation of the spread of methicillin-resistant Staphylococcus aureus (MRSA), a serious hospital acquired infection, within the dialysis unit at the University of Iowa Hospitals and Clinics (UIHC). The simulation is based on ten days of fine-grained healthcare worker (HCW) movement and interaction data collected from a sensor mote instrumentation of the dialysis unit by our research group in the fall of 2013. The simulation layers a detailed model of MRSA pathogen transfer, die-off, shedding, and infection on top of agent interactions obtained from data. The specific question this paper focuses on is whether there are simple, inexpensive architectural or process changes one can make in the dialysis unit to reduce the spread of MRSA? We evaluate two architectural changes of the nurses' station: (i) splitting the central nurses' station into two smaller distinct nurses' stations, and (ii) doubling the surface area of the nursing station. The first architectural change is modeled as a graph partitioning problem on a HCW contact network obtained from our HCW movement data. Somewhat counter-intuitively, our results suggest that the first architectural modification and the resulting reduction in HCW-HCW contacts has little to noeffect on the spread of MRSA and may in fact lead to an increase in MRSA infection counts in some cases. In contrast, the second modification leads to a substantial reduction - between 12% and 22% for simulations with different parameters - in the number of patients infected by MRSA. These results suggest that the dynamics of an environmentally mediated infection such as MRSA may be quite different from that of infections whose spread is not substantially affected by the environment (e.g., respiratory infections or influenza).
{"title":"Evaluating Architectural Changes to Alter Pathogen Dynamics in a Dialysis Unit","authors":"Hankyu Jang, Samuel Justice, P. Polgreen, Alberto Maria Segre, Daniel K. Sewell, S. Pemmaraju","doi":"10.1145/3341161.3343515","DOIUrl":"https://doi.org/10.1145/3341161.3343515","url":null,"abstract":"This paper presents a high-fidelity agent-based simulation of the spread of methicillin-resistant Staphylococcus aureus (MRSA), a serious hospital acquired infection, within the dialysis unit at the University of Iowa Hospitals and Clinics (UIHC). The simulation is based on ten days of fine-grained healthcare worker (HCW) movement and interaction data collected from a sensor mote instrumentation of the dialysis unit by our research group in the fall of 2013. The simulation layers a detailed model of MRSA pathogen transfer, die-off, shedding, and infection on top of agent interactions obtained from data. The specific question this paper focuses on is whether there are simple, inexpensive architectural or process changes one can make in the dialysis unit to reduce the spread of MRSA? We evaluate two architectural changes of the nurses' station: (i) splitting the central nurses' station into two smaller distinct nurses' stations, and (ii) doubling the surface area of the nursing station. The first architectural change is modeled as a graph partitioning problem on a HCW contact network obtained from our HCW movement data. Somewhat counter-intuitively, our results suggest that the first architectural modification and the resulting reduction in HCW-HCW contacts has little to noeffect on the spread of MRSA and may in fact lead to an increase in MRSA infection counts in some cases. In contrast, the second modification leads to a substantial reduction - between 12% and 22% for simulations with different parameters - in the number of patients infected by MRSA. These results suggest that the dynamics of an environmentally mediated infection such as MRSA may be quite different from that of infections whose spread is not substantially affected by the environment (e.g., respiratory infections or influenza).","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"382 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117158435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Shreya Jain, D. Niranjan, Hemank Lamba, Neil Shah, P. Kumaraguru
Livestreaming platforms enable content producers, or streamers, to broadcast creative content to a potentially large viewer base. Chatrooms form an integral part of such platforms, enabling viewers to interact both with the streamer, and amongst themselves. Streams with high engagement (many viewers and active chatters) are typically considered engaging, and often promoted to end users by means of recommendation algorithms, and exposed to better monetization opportunities via revenue share from platform advertising, viewer donations, and third-party sponsorships. Given such incentives, some streamers make use of fraudulent means to increase perceived engagement by simulating chatter via fake “chatbots” which can be purchased from shady online marketplaces. This inauthentic engagement can negatively influence recommendation, hurt streamer and viewer trust in the platform, and harm monetization for honest streamers. In this paper, we tackle the novel problem of automating detection of chatbots on livestreaming platforms. To this end, we first formalize the livestreaming chatbot detection problem and characterize differences between botted and genuine chatter behavior observed from a real-world livestreaming chatter dataset collected from Twitch.tv. We then propose Sherlock, which posits a two-stage approach of detecting chatbotted streams, and subsequently detecting the constituent chatbots. Finally, we demonstrate effectiveness on both real and synthetic data: to this end, we propose a novel strategy for collecting labeled, synthetic chatter dataset (typically unavailable) from such platforms, enabling evaluation of proposed detection approaches against chatbot behaviors with varying signatures. Our approach achieves .97 precision/recall on the real-world dataset, and .80+ F1 scores across most simulated attack settings.
{"title":"Characterizing and Detecting Livestreaming Chatbots","authors":"Shreya Jain, D. Niranjan, Hemank Lamba, Neil Shah, P. Kumaraguru","doi":"10.1145/3341161.3345308","DOIUrl":"https://doi.org/10.1145/3341161.3345308","url":null,"abstract":"Livestreaming platforms enable content producers, or streamers, to broadcast creative content to a potentially large viewer base. Chatrooms form an integral part of such platforms, enabling viewers to interact both with the streamer, and amongst themselves. Streams with high engagement (many viewers and active chatters) are typically considered engaging, and often promoted to end users by means of recommendation algorithms, and exposed to better monetization opportunities via revenue share from platform advertising, viewer donations, and third-party sponsorships. Given such incentives, some streamers make use of fraudulent means to increase perceived engagement by simulating chatter via fake “chatbots” which can be purchased from shady online marketplaces. This inauthentic engagement can negatively influence recommendation, hurt streamer and viewer trust in the platform, and harm monetization for honest streamers. In this paper, we tackle the novel problem of automating detection of chatbots on livestreaming platforms. To this end, we first formalize the livestreaming chatbot detection problem and characterize differences between botted and genuine chatter behavior observed from a real-world livestreaming chatter dataset collected from Twitch.tv. We then propose Sherlock, which posits a two-stage approach of detecting chatbotted streams, and subsequently detecting the constituent chatbots. Finally, we demonstrate effectiveness on both real and synthetic data: to this end, we propose a novel strategy for collecting labeled, synthetic chatter dataset (typically unavailable) from such platforms, enabling evaluation of proposed detection approaches against chatbot behaviors with varying signatures. Our approach achieves .97 precision/recall on the real-world dataset, and .80+ F1 scores across most simulated attack settings.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114325389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper examines the effectiveness of a range of pre-trained language representations in order to determine the informativeness and information type of social media in the event of natural or man-made disasters. Within the context of disaster tweet analysis, we aim to accurately analyse tweets while minimising both false positive and false negatives in the automated information analysis. The investigation is performed across a number of well known disaster-related twitter datasets. Models that are built from pre-trained word embeddings from Word2Vec, GloVe, ELMo and BERT are used for performance evaluation. Given the relative ubiquity of BERT as a standout language representation in recent times it was expected that BERT dominates results. However, results are more diverse, with classical Word2Vec and GloVe both displaying strong results. As part of the analysis, we discuss some challenges related to automated twitter analysis including the fine-tuning of language models to disaster-related scenarios.
{"title":"Estimating Distributed Representation Performance in Disaster-Related Social Media Classification","authors":"P. Jain, R. Ross, Bianca Schoen-Phelan","doi":"10.1145/3341161.3343680","DOIUrl":"https://doi.org/10.1145/3341161.3343680","url":null,"abstract":"This paper examines the effectiveness of a range of pre-trained language representations in order to determine the informativeness and information type of social media in the event of natural or man-made disasters. Within the context of disaster tweet analysis, we aim to accurately analyse tweets while minimising both false positive and false negatives in the automated information analysis. The investigation is performed across a number of well known disaster-related twitter datasets. Models that are built from pre-trained word embeddings from Word2Vec, GloVe, ELMo and BERT are used for performance evaluation. Given the relative ubiquity of BERT as a standout language representation in recent times it was expected that BERT dominates results. However, results are more diverse, with classical Word2Vec and GloVe both displaying strong results. As part of the analysis, we discuss some challenges related to automated twitter analysis including the fine-tuning of language models to disaster-related scenarios.","PeriodicalId":403360,"journal":{"name":"2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)","volume":"261 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124795084","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：