Efficiently Spotting Social Media Opinions Across Modalities and Languages
Keeping track of a daily volume of up to 100 million relevant social media comments and customer opinions in different media, languages and formats in order to obtain valuable insights for customers requires usually a great amount of manual effort.
Social media contributions have to be read, watched and categorized in order to meet the need of deep and high qualitative understanding of the data associated with many service and market research use cases.
xLiMe technology supports the automatization of this process in VICO Analytics by clustering and structuring data across media and across languages, this way significantly reducing manual effort.
This enables us to deliver time more efficently and at the same time well-grounded research services.
In Zattoo’s web application, users can browse TV shows via semantic linkings.
Based on speech-to-text transcriptions of the actual TV shows, xLiMe technology provides “related shows” recommendations that go beyond traditional similarity recommendation approaches which are usually mostly based on curated TV listings data or collaborative filtering approaches.
True understanding of the actual audio-visual content allows for highly relevant recommendations.
Two xLiMe demonstrators showing the capabilities of xLiMe technology to interlink media items in different languages and modalities.
Search and explore the media items in our 'live' showcase and do the same using a dataset from June 2016 focusing on the EU (Brexit) referendum.
Lei Zhang, Achim Rettinger, Ji Zhang.
A Knowledge Base Approach to Cross-lingual Keyword Query Interpretation.
The 15th International Semantic Web Conference (ISWC'16), Springer, October, 2016
Lei Zhang, Michael Färber, Achim Rettinger.
XKnowSearch! Exploiting Knowledge Bases for Entity-based Cross-lingual Information Retrieval.
The 25th ACM International on Conference on Information and Knowledge Management (CIKM'16), ACM, October, 2016
Semantic image and video annotation tools aim to automatically extract high level information from visual content.
The xLiMe semantic visual annotator can be applied to detect and localize within an image a specific object of interest or a brand logo, thus enriching image and/or video metadata.
The xLiMe semantic visual annotator allows you to follow your favourite brand or to monitor a topic of interest both on social network or TV mainstream.
ELES (pronounced "Alice") is a lightweight combination of entity linking and entity summarization. In the demo, we use an entity linking service to analyze text and link fragments to entities of the DBpedia knowledge base. The LinkSUM summarizer (interfaced via the SUMMA API definition) produces fact-based summaries of DBpedia entities. The two applications are combined on the client side through the "Internationalization Tag Set 2.0" W3C recommendation and lightweight jQuery-based interfaces.
Cross-channel and Cross-media Recommender
xLiMe semantic integrator leverages implicit or latent semantics in the textual data present in different languages to provide search, cross-channel recommendations and analytic's.
In Event Registry users can identify correlations between different data types,
such as mentions of an entity in the media or its popularity on social media, stock prices of companies, number of sold items, etc.
Event Registry is a media monitoring platform that collects news published globally. One of the features it provides is also tracking of how frequently a particular entity is mentioned in the media. This allows one to obtain a time-series data of mentions over several years. Along with mentions of entities, Event Registry also stores time-series information about other data types, such as stock prices of companies, trading volumes, exchange rates, etc. With several millions of time-series, Event Registry can then identify for a given time-series what are other time-series that correlate the most with it. This can be used to identify simple relations between various data types as well as potential causation.