Press Kit -LiMoSINe Project
LiMoSINe Press kit introduces this project that integrates the studies of leading researchers over diverse topics with a view to enable new kinds of language-based technology search. Now we are developing 5 demonstrators: ORMA, ThemeStreams, FlickrDemo, DEESSE and Streamwatchr. http://limosine-project.eu/
Published on: Mar 4, 2016
Transcripts - Press Kit -LiMoSINe Project
Linguistically Motivated Semantic
Co-funded by the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no 288024.
urrently, daily monitoring of social networking sites, e.g. analysis of tweets, is mostly
done manually. However, the spectacular growth of user generated content makes it
inefficient, causing a high demand for tools capable of automatically detecting reputa-
tion alerts. Unfortunately, online reputation management (ORM) applications available on
market still cannot provide professionals with satisfactory solutions.
Some of the research questions related to information access over social media (and among
them, those related to ORM) are addressed by the European Union project LiMoSINe,
Linguistically Motivated Semantic Aggregation Engines, formed by a consortium of four
universities (University of Amsterdam, University of Glasgow, University of Trento, and
Universidad Nacional de Educación a Distancia) and two companies: Fundacio Barcelona
Media Universitat Pompeu Fabra and a communication consultancy firm, LLORENTE &
The main objectives of this project are:
• LiMoSINe integrates the research activities of leading researchers across diverse
topics with a view to enabling new kinds of language-based search technology.
• LiMoSINe´s vision is to allow access to online information from a documentcentric
search paradigm focused on returning disconnected atomic pieces that
understands the user´s questions and intent.
• LiMoSINe´s aggregation engines automatically organize search results inintelligent
and meaningful ways . The objective is to build search and recommendation
systems that will understand a user´s intent, discover and organize facts, identify
opinions, experiences and trends, all from multilingual online sources and open
online sources and open online databases, like Twitter, YouTube, MusicBrainz, Flickr,
The LiMoSINe Project is now developing 5 demonstrators: ORMA, ThemeStreams, Flickr
Demo, DEESSE and Streamwatchr.
ORMA (Online Reputation Monitoring Assistant)
is an interactive annotation tool.
It helps online reputation experts to label tweets
related to a client with information essential for
the analysis of the client’s reputation.
The current version of the assistant works with tweets in English and Spanish. Besides storing
and organising manually introduced data, it can also prompt automatically generated labels
which can be validated or corrected by analysts. The main goal is to facilitate the daily work of
online reputation experts and increase their efficiency.
ThemeStreams is a demonstrator aimed at giving
insight into who started a topic in social media.
It monitors tweets from four groups of people, namely,
politicians, political journalists, lobbyists, and other
influencers and visualizes the dynamics of a given topic
and the discussion around it.
It indexes and visualises tweets as streams of influences,
and shows this as a streamgraph of the political landscape
over time. This shows when somebody said something,
and how many people found that interesting.
ThemeStreams can be adapted to other typed data outside of the political landscape. It is
useful to media analysts for better understanding who said what, and how the discussion
around a topic has evolved among groups of people and over time.
Check it out at http://themestreams.xtas.net
ThemeStreams: visualizing the Stream
of themes discussed In Politics
Who started talking about this issue?
The Flickr Demo allows the user to annotate images related to a large scale social event. On selec-
tion of an image, the user is offered automatic tag recommendations (based on existing tags
added by the user), as well as related tweets and Wikipedia content.
We achieve this by offering the user a number of photo tagging methods:
• Manual Tagging
• Tag Recommendations
• Related Tweets: tweets tagged with the hash
related to the event
• Related Wikipedia content
The entity-Driven Exploratory and sErendipitous
Search SystEm (DEESE) enables a serendipitous
exploration of complex data extracted from Yahoo
Answers, by providing a high-level overview of the
information in the form of an enriched entity network.
Our demo support a serendipitous and exploratory search over Yahoo Answers, giving user an
unprecedented capability of exploring the wealth of data enclosed in it by many different pers-
pectives. Furthermore, it allows to exploit the advantages of the freedom of conversation on
Yahoo Answers, which contains within it opinions, rumors, and social interest and approval.
Streamwatchr is a demonstrator aimed at understan-
ding the world, as it happens. Streamwatchr moni-
tors, interprets and analyzes social media for specific
user activities, e.g, listening to music, watching a
movie, eating, and visualizes these interpretations in
real-time. The interpretation happens via mapping
the text in social media to vertical and horizontal
knowledge bases (e.g., Musicbrainz and Wikipedia).
Currently, Streamwatchr focuses on music-related tweets, and offers a new and playful way to
engage with music. Besides monitoring the latest popular trends and provide recommendations
for interesting new music discoveries, Streamwatchr monitors what people sing along and offers
statistics on the most sung-alongparts of each song.
Streamwatchr is great for both discovering new music and checking charts, and tracking artist
song trends. Check it out at http://streamwatchr.com
LiMoSINe has promoted RepLab, a benchmarking activity in
the area of Online Reputation Management (ORM).
It is a competitive evaluation exercise for ORM systems orga-
nised by UNED, UvA, and LLORENTE & CUENCA that aims at:
• Defining relevant challenges and specifying
appropriate resources for automatic ORM
• Building test collections for benchmarking purposes
• Comparing state-of-art systems and algorithms
Created in 2012 as an activity of Cross Language Evaluation
Forum (CLEF), RepLab has covered such tasks as filtering,
reputation polarity classification, grouping and ranking of
entity related tweets.
RepLab 2014 will focus on Reputation Monitoring in Twitter,
targeting two new tasks: categorization of messages with
respect to standard reputation dimensions and classifica-
tion of Twitter profiles (author profiling). The results will be
discussed at the CLEF 2014 Conference in Sheffield (UK), 15-18
Linguistically Motivated Semantic
LLORENTE & CUENCA
Members of the consortium LiMoSINe Project
If you have any further questions or you want to interview one of the team leaders,
please do not hesitate to contact Vanessa Álvarez or Ana Pitart in these emails:
Universiteit van Amsterdam, The Netherlands
Maarten de Rijke (Scientific Coordination and professor of Information Processing and Internet, University of Amsterdam)
Manos Tsagkias (Postdoctoral researcher in Information Retrieval and Predictive Analytics , University of Amsterdam)
University of Glasgow, United Kingdom
Joemon M Jose (Professor at the Department of Computing Science, University of Glasgow)
Fundacio Barcelona Media Universitat Pompeu Fabra, Spain
Mounia Lalmas (Visiting Principal Scientist, Yahoo! Research Barcelona)
Universita degli studi di Trento, Italy
Alessandro Moschitti (Assistant Professor at the Information Engineering and Computer Science Department, University of Trento)
Llorente & Cuenca Madrid SL, Spain
Adolfo Corujo (Partner and Iberian General Manager, LLORENTE & CUENCA)
Universidad Nacional de Educación a Distancia, Spain
Julio Gonzalo (Assistant Professor and Member of the UNED group in Natural Language Processing and Information Retrieval)