RGCL Anniversary Highlights, Day 12
Published on Aug, 23 2022 by RGCL.
Ideas ahead of their time
In 2003, after publishing our work in multiple-choice question generation in the very first BEA workshop, the director of the Centre for Innovation at NBME came to us (see our NBME highlight) with the statement: “I believe that in the future, all test questions will be generated automatically by computers”. At that time, it was a very bold, and head-scratching, statement. Would you want to test humans using questions generated by machines? It is now a reality: the items of Duolingo English Test (an English test that is now used by thousands of Universities) are generated automatically, and of course, reviewed by human experts (see page 13, https://duolingo-papers.s3.amazonaws.com/other/det-technical-manual-current.pdf).
In 2004, as Google Translate was about two years from being published, Ted Marshall, an entrepreneur from Translution, came to RGCL with an idea: what if you can send an email in English to a Romanian customer, and they can read it in their own language? What if such emails are translated with proper terminology? What do we need to make these come true? RGCL then identified several important problems for machine translations back then, chief among them is the problem of translating Named Entities (Google Translate still has some problems with this now). RGCL then implemented a set of NE recognisers and part-of-speech taggers for multiple languages to help Translution solve the problems. It was slightly ahead of its time, as now everything, not only emails, can be automatically translated on the fly.
In 2008, Nokia was still the king of mobile phone market; the iPhone 2G had just been introduced; Siri was two years from being released; and RGCL was involved in a project named QALL-ME, an EU-FP6 project which aimed at the realization of a shared and distributed infrastructure for Question Answering (QA) systems on mobile devices (e.g. mobile phones). Questions are formulated by the users in free natural language input, and the system returns the actual sequence of words which constitutes the answer from a collection of information sources. Is that what Amazon Alexa, Microsoft Cortana, Google Assistant are doing, I hear you ask? Yes, it is, and yes, it was back in 2008 that RGCL was also working on them. Slightly too soon, if you ask us.
Moving forward, RGCL has continued to produce avant-garde and cutting-edge research based on ideas that were ahead of their time. Such examples are our innovative research on computer-aided summarisation; the use of NLP methodology to assist people with autism (see our recent highlights on FIRST and AUTOR); award-winning quality estimation tools for machine translation; and our leading and seminal work on anaphora resolution.