25 June 2012
The LTi News Roundup - 25th June 2012 (part 1)
Weekly news round-up prepared by the Editorial Staff of LangTechNews for LT-Innovate, the Forum for Europe’s Language Technology Industry.
Top of the Week: Naturalising the interface.
After a classic career path from technology start-up to acquisition by an industry major, Swype (a Nuance company since last year) delivered an interesting Beta upgrade this week to its input technology combining speech technology, clever word statistics and keyboarding.
As the population of mobile users explodes, the interface between user and device is becoming ever more critical. Mobile voice assistants made their debut towards the end of 2011 and the strategic battle for quality and language coverage is on-going. Gesture and touch are now being experimented with or manufactured into a variety of personal and collective (e.g. TV) screens.
This latest Swype product, then, packs swyping (predictive text), voicing (speech input), touching, and typing in a single app. It learns from your emails to predict your typed-in words, and also teaches the voice interface (based on Dragon technology) to better prepare for your typical voice input. This takes us one step further down the road towards personalized devices offering more choice to the end-user, and greater leverage of cloud-based machine learning to automate language capabilities,
New Ways to Smarten up Content: There are two basic entry points to ‘intelligent content’
One is access (what smarts will help me find and use existing content of all kinds?), the other is creation (how can I improve my new content so that it integrates as relevantly as possible into global conversations?).
On the creation dimension, the European Commission’s own Joint Research Centre this week announced the availability of its EuroVoc Indexer that automates the categorisation of parliamentary documents in 22 languages. This makes life easier for content curators in large institutions, who do not have to index documents manually: the machine learns what the document is about so that it is then easier to access.
Similarly, the Dutch start-up Silk has developed an application that automatically prepares semantic content for websites. It adds in the relevant metadata and open-data links that turn a website’s content into ready to go “semantic web data” that can be exploited by other semantic tools in this growing content space. A good example of how complex technology can be packaged into user-friendly solutions for everyday publishers of digital content.
Along the access dimension, the Barcelona company Cognicor has innovated with a new application for automating the processing of low-value complaints. Customer complaints form a genre of user-generated content that is soaring in volume as commercial enterprises of all kinds open up their websites to customer feedback. Cognicor’s idea is to relieve busy contact centre agents by automating the entire process of understanding a complaint, checking its validity and even negotiating some form of compensation. This means using semantics in various interesting ways to interpret the language of the complaint.