CrowdSens 2012 : International Workshop on Multimodal Crowd Sensing
Call For Papers
CrowdSens'12 workshop is co-located with the 21st ACM International Conference on Information and Knowledge Management (CIKM 2012), and will take place in Maui, Hawaii, USA.
According to research conducted by the International Data Corporation (IDC), the size of the ‘digital universe’ in 2010 (i.e., the amount of information which is stored digitally) surpassed one Zettabyte (ZB) for the first time in history and it now stands at about 1.8 ZB. This massive expansion in the size of the amount of information appears to be exceeding Moore’s Law. It is also estimated that about 70% of this information is generated by individuals. The ubiquitous availability of computing technology, in particular smartphones, tablets, laptops and other easily portable devices, and the adoption of social networking sites, make it possible to be connected and continuously contribute to this massively distributed information publishing process.
By doing so, users are (unconsciously) acting as social sensors, whose sensor readings are their manually generated data. People document their daily life experiences, report on their physical locations and social interactions with others, express opinions and provide diverse observations on both the physical world (sights, sounds, smells, feelings, etc.) and the online world (news, music, events, etc.). Such massive amounts of ubiquitous social sensors, if wisely utilized, can provide new forms of valuable information that are currently not available by any traditional data collection methods including real physical sensors, and can be used to enhance decision making processes.
It has been shown over and over that reports on real world events, such as the Japan’s Earthquake and Tsunami, the Arab Spring uprisings, and the England’s riots happened in 2011, are much faster propagated within the network of social sensors (e.g. on Twitter) than they are processed by traditional means (e.g. seismic sensor reading analysis, police emergency reports, news media coverage). In these cases, human observers can be exploited to interpret and enrich such integrated sensor-derived information. As an example, both journalists and opinion makers now make increasing usage of massive data collected from social sensors in order to study public opinions, and discover new perspectives of daily stories. As another example, within a smart city scenario, social sensors can contribute important information about the daily city life through various channels, such as social media, SMS, and reports to the city operation center.
Such social sensors can enrich the existing information currently collected by the city physical sensors (e.g. traffic and camera sensors), helping to reduce uncertainty, and leading to a better envision and comprehension of the magnitude of potential problems and situations.
Effective mining, analyzing, fusing, and exploiting information sourced from multimodal physical and social sensor data sources is still an open and exciting challenge. Many factors here add to the complexity of the problem, including the real-time element of the data processing; the heterogeneity of the sources, from physical sensors data to posts on social media; and the ubiquitous and noisy nature of the human-sensor generated information, which can be written in an informal style, duplicated, incomplete or even incorrect.
The 1st International Workshop on Multimodal Crowd Sensing (CrowdSens 2012) will provide an open forum for researchers from various domains such as data management, data mining, information retrieval, and semantic web, for discussing the above challenges.
The main goal of CrowdSens 2012 is to become a major international forum for Researchers and practitioners from different research areas such as Social Web, Semantic Web, Natural Language Processing, Information Extraction, Data Mining, Information Retrieval, User Modelling, Personalization, Stream Processing, and Sensor Networks, who focus their work on user-generated contents.
Our aim is to stimulate discussions about how the knowledge embedded in human sensor data can be collected, extracted, modelled, analysed, integrated, summarized, and finally exploited. Ideas for innovation will extend through different fields, from data mining, user modelling, personalization, recommendation, information retrieval, and business intelligence, to name a few. Different research lines, backgrounds, perspectives, and degrees of expertise will be present at the workshop, and thus very interesting multidisciplinary discussions, collaborations, and work synergies between the workshop attendees are expected as one of the main outcomes of the event.
Haggai Roitman, IBM Research - Haifa, Israel
Ivan Cantador, Universidad Autonoma de Madrid, Spain
Miriam Fernandez, Knowledge Media Institute, UK
Topics of interest
Themes and topics of interest at this workshop include, but are not limited to:
* Data acquisition methods for crowd sensing
- Physical world crowd data capture
- Multimedia crowd data capture (e.g. SMS, MMS, CDRs, transcripts)
- Real-time data acquisition methods
- Massive scale social sensor monitoring and crawling
- Predictive models for social data acquisition
- Scheduling, prioritization and sampling methods
* Data models for crowd sensing
- Social sensor event models
- Social sensor data representation
- Social sensor context representation
- Spatio-temporal models for crowd sensing
- Multimodal data models for crowd sensing
- Semantic models for crowd sensing
- Uncertainty models for incomplete and noisy social sensors data
- Trust and authorization models for crowd sensing
- Privacy in crowd sensing
* Novel data processing, analysis, and classification methods
- Data cleansing for crowd sensing (e.g. real-time duplicates detection)
- Feature extraction, Entity analytics and novel NLP methods
- Context extraction and prediction using multimodal sources
- Uncertainty estimation and predictive analytics
- Data mining methods under incomplete and noisy data (e.g. online clustering, categorization, classification)
- Opinion mining, sentiment analysis methods for crowd sensing
- Trends, bursts, anomalies and outliers detection over large scale social sensor data
- Network analysis, information propagation and influence detection methods for crowd sensing
- Crowd behavioural analysis and prediction
- Real-time community detection and analysis
- Social stream processing methods (e.g. top-k querying, filtering, sampling)
* Event detection, fusion, and summarization methods
- Event detection methods (under uncertainty, incomplete or noisy settings)
- Event story detection
- Detection of developing events
- Event uncertainty estimation
- Event time and location estimation
- Methods for event data delivery
- Methods for event data reporting, summarization or visualization
- Pattern recognition methods
- Multimodal data fusion methods
* Evaluation methods for crowd-sensing
- Quality metrics and key performance indicators for crowd sensing
- Benchmarks and evaluation methodologies for crowd sensing
* Applications of crowd sensing
- News mining from social sensors (e.g. emerging story detection)
- Infotainment (e.g. event discovery and recommendation)
- Disaster management (e.g. weather monitoring, disaster prediction)
- Public safety (e.g. prediction of developing situation and sentiments)
- Public health (e.g. epidemic monitoring, infectious disease outbreak detection)
- Transportation (e.g. prediction of traffic loads, detection of hazards)
- Finance (e.g. market monitoring)
- Cyber security (e.g. Counter terrorism, dark web monitoring)
- Government and Politics (e.g. Voice of Citizen, opinion mining)
- Retail and consumer products (e.g. Voice of Customer, demand sensing)
We invite two main types of contributions: full papers (6-8 pages) and short papers (2-4 pages). Both types of contributions could be new research ideas, position statements, critiques of existing approaches, or experiment reports.
Submitted papers will be evaluated according to their originality, technical content, style, clarity, and relevance to the workshop. Each paper will be reviewed by at least three independent referees.
Manuscripts should be submitted electronically, in PDF format and formatted using the ACM camera-ready templates available at:
All submissions will be done electronically via the CrowdSens 2012 Web submission system:
Accepted workshop papers will also be published in the CIKM workshop proceedings, which will be printed on CD only and indexed in the ACM digital library, together with the main CIKM 2012 proceedings.
At least one author of each accepted paper must register for the conference. Information about registration is provided at CIKM 2012 Web page: