Susanne Weber about the Language Tech Accelerate Conference in Brussels

cover-image

This conference was all about Language Techs that go beyond academia and research – an industry that is dealing with the social media revolution and the rise of customer experience. It was the mothership of conferences for any Language tech fan.

Day 1

The first day was primarily for the commercial aspect of language tech, such as advertising and other financial focus. The second day, which I attended, was all about semantic tagging and sentiment analysis.

Just to give a few highlights – some of the interesting presentations:

Europeana:

A project set out to digitise and connect the content of galleries, museums, archives. They’re making archives open to members of public, as well as creative industries + “Professionals” archives

They want to digitise Europe’s archives (only 10% have been digitised) and make it open to the public.

LT conference image Credit: LT Accelerate Conference

TILDE - Machine Translation project

Concentrating on Latvian, Lithuanian, Estonian as ‘fringe’ languages. Their focus is on e-Government, real-time content, quick access to web etc. They intend to provide mobile applications for translations, as well as apps for media people. Latvia has EU presidency in 2015 and they’re organising a 3-day conference on multilingual in April 2015 – “Breaking the language barrier”. They state that Google is not good enough. There are security issues with the Cloud, not good for minority languages, not good for large volumes.

News and Publishing using language tech, presented by Jarred McGinnis (Logomachy)

Jarred’s emphasis was all about semantic tagging, disambiguate, identifying relationship. He demonstrated BBC Sport as a prime example for using semantic tagging successfully. He said “We need to create tools for journalists”. The question is how to get journalists to contribute to semantic tagging: e.g. ticking a box in the article.

Social Media Analytics - Diana Maynard (Uni Sheffield)

Diana Maynard presented Sheffield’s research project of social media sentiment analysis. She showed us some interesting stats:

  • 30% of all Americans get their news from Facebook.

  • Of all Twitter users: 390 million users have no followers

  • 40% Twitter users have never sent a Tweet

Benefit of social media sentiment analysis: for election polls: Analysing social media is more efficient than yougov polls. Psephologists can get info: what kind of events will change voters’ minds?

Challenges for sentiment analysis: spelling, grammar, abbreviations, sarcasm – these factors make analysis difficult. They need to develop tools for hashtag analysis – disambiguation. Great example for Twitter ambiguiation: #nowthatcherisdead – this can mean “now that Cher is dead” or “now Thatcher is dead”.

The general consensus of most speakers was this:

Content has shifted from company content to customer content; i.e. customers creating content. Content is King -> customer creates this content – citizens create content. Therefore, knowledge is King. Big Data, citizens’ data - is always multi-lingual.

Images Credit: LT Accelerate Conference


Categories:

Tags: