Microposts 2015 - 5th Making Sense of Microposts Workshop

The #Microposts workshops aim to bring together researchers from multiple disciplines to debate current, leading edge effort toward analysing and understanding Microposts - "information published on the Web that is small in size and requires minimal effort to publish (e.g. a Tweet, Facebook share, Instagram like, Google +1)". Although individual Microposts are small and therefore typically focus on a single thought, message or theme, collectively they provide a rich source of information and opinion about a range of topics. The workshop aims to continue to provide a forum to enable discussion and hence, improve understanding of social and cultural phenomena that influence the publication and reuse of Microposts; to assess different approaches to gleaning the information content of Micropost data; and discuss application of this knowledge content in a variety of contexts, including emergency response, crowd and event tracking, mass communication, opinion mining and sentiment analysis. Enabling the understanding and application of Microposts requires techniques and tools that function at scale, and that are able to handle the very high rate at which Microposts are published.
Despite advances in the development of tools to tackle the specific challenges inherent in Micropost data, applications and approaches for analysing Microposts for different tasks still rely on the use of third party text extraction tools. Such tools are typically applied to extract entities and concepts contained in Microposts. One issue here is the lack of formal evaluation of the accuracy of text extraction tools specifically for Micropost data; instead, comparative assessment using corpora of well-formed, normal length, natural language documents is reported. To address this issue, starting with #MSM2013, the workshop hosted an entity extraction challenge in which participants detected named entities typed with corresponding concepts (e.g. 'Barack Obama' is a Person). In 2014, the challenge was extended to require also the linking of entities extracted to relevant DBpedia sources. For #Microposts2015 we propose to go one step further and extend the challenge by not only testing the accuracy of entrants' systems in extracting entities and linking them with DBpedia, but also assessing the runtime efficiency of the submitted systems. Evolution of the challenge each year addresses a current need of researchers and others who rely on the output of text extraction tools, where reliability and computation time are important when dealing with large-scale datasets.
TOPICS OF INTEREST
---
#Microposts2015 will focus on topics including, but not exclusive to, the three areas below:
UNDERSTAND: With a focus on the human in Micropost data generation and analysis, we encourage submissions that look at understanding how situation and context drive individual and collective generation of Microposts, whether targeted at the general public, a specific person or other entity, e.g. a ruling government or a cause. We particularly encourage interdisciplinary work and that driven by research in the Social, Information & Web Sciences, that lead to deeper understanding of the content of Microposts, and how this content influences the contribution of Micropost data to, among others:
- Collective awareness
- Education & citizen empowerment, data & citizen journalism
- Civil action, media & politics
- Political and polemical aspects of Microposts
- Ethics, legal and privacy issues
- Psychological profiling and psychological aspects of Micropost-based interaction
- Cultural, generational and regional differences in access and use
- Impact of effortless posting and wearable devices on communication
DISCOVER: The extraction of information content from Micropost data and its subsequent analysis contribute to the discovery of patterns and trends in the data. This information is key to further knowledge discovery and application, using a number of approaches including:
- Emergent semantics
- Data mining from Microposts
- Opinion mining, sentiment and sentic analysis
- Network analysis and community detection
- Influence detection and social contagion modelling
- Prediction approaches
- Linking Microposts into the Web of Linked Data (i.e. entity extraction and URI disambiguation)
APPLY: Applications papers and case studies describing systems related, but not limited, to:
- Collective intelligence, user profiling, personalisation & recommendation
- Business analytics & market intelligence
- Event & topic detection and tendency tracking
- Microposts as second screen to television
- Geo-localised, Micropost-based services
- Public consensus & citizen participation
- Security, emergency response & health
- Linking social and physical signals for, e.g., crowd tracking
SOCIAL SCIENCES TRACK
---
To foster collaboration between Computer Science and Social Sciences, and continue to encourage contribution from the latter domain to improve on 'Making Sense of Microposts', there will be a special track dedicated to Social Science papers. This track will be chaired by Katrin Weller (GESIS, Germany) and Danica Radovanović (University of Novi Sad, Serbia), long-time members of the #Microposts workshop program committee. Details of this track will be sent out in a separate call for papers.
NAMED ENTITY RECOGNITION AND LINKING CHALLENGE
---
On the back of two very successful challenges in 2013 and 2014, that saw widespread interest, we have already obtained sponsorship from SPAZIODATI for a financial award for the highest ranking submission. Each year the challenge has allowed a valuable overview of the state of the art and has received expressions of interest and submissions from both industry and academia. We have also seen continued interest after publication of the results in each year. We expect this to continue in 2015, with the extension of the challenge to recognize entities, and type them before linking to the corresponding DBpedia entities. We will also measure the runtime computation efficiency of submitted systems. In line with the overall workshop goals, we envisage that the outcomes of the challenge submissions will continue to advance work in the domains of named entity recognition and entity disambiguation, with a specific focus on function over the short length information snippets in Microposts. A detailed description of the challenge will be published in a separate call.
WORKSHOP STRUCTURE
---
A keynote address from an invited speaker will open the day, and followed by paper presentations. We will hold a poster and demo session to trigger further, in-depth interaction between workshop participants. The last set of presentations will be brief overviews of selected submissions to the Challenge. The workshop will close with the presentation of awards.