Impact case study database

Impact 1: The team’s research was critical to the development of the keyboard app SwiftKey, which was acquired by Microsoft in 2016 for USD 250 million:

Research by Briscoe’s NLP group on language modelling and classification and on compression of such models [R4] provided the foundational ML techniques that were used in the development of SwiftKey. The SwiftKey App was designed to pick up the patterns in people’s language use, and through Machine Learning (ML) techniques, to be very accurate in predicting what people were most likely to write next. Language data was gathered by Briscoe and his colleagues as part of an STFC grant [E1] and iLexIR, a technology and knowledge transfer company cofounded by Briscoe, provided the data to the 2009 start-up (co-founded by ex-Cambridge students Ben Medlock and John Reynolds) that became SwiftKey. This data was used to train the language prediction models on SwiftKey’s new keyboard application and – as the co-founder and CTO of Switfkey confirms - “made a huge difference at a critical early stage” [E1].

SwiftKey was first released as an exclusive for Android Market in July 2010, followed by an iOS release in September 2014 after Apple allowed third-party keyboard support. By June 2014, users were estimated to have saved half a trillion keystrokes, and SwiftKey had been the top paid Android app for two years running [E2]. By 2016 SwiftKey software was installed on more than 300 million devices and it was estimated that its users had saved nearly 10 trillion keystrokes, across 100 languages, saving more than 100,000 years in combined typing time [E3].

The success of SwiftKey led to its acquisition by Microsoft in 2016 for USD250 million [E4]. The software was incorporated into Microsoft’s flagship mobile app, SwiftKey Keyboard for iPhone and Android (500 million installs by 2020), which are still popular today [E4].

Impact 2: The team’s research on automatic feedback and assessment of texts for “English for Speakers of Other Languages” has helped over 500,000 learners and reduced the workload of teachers.

The NLP group’s knowhow and the ALTA prototypes for automated writing assessment were the basis of the 2014 University spin-out “English Language iTutoring” ( ELiT).

The first production version of the assessment software was released by ELiT as a free cloud-based service “Write & Improve”, in September 2016. Within a year of its launch, it was used by over 650,000 people in 225 countries with well over a million pieces of writing submitted and checked [E5].

The ML technology enables Write & Improve to assess the writing and estimate a proficiency level, to identify errors and make suggestions, and to give supportive feedback - all in under 15 seconds [E6]. The learner is able to make corrections and resubmit their writing repeatedly, and the suggested improvements will change with each submission because the feedback is scaffolded.

Testimonials from students provide evidence of the success of Write & Improve in creating a free, easy-to-use, low-stakes, friendly, online environment that encourages students to practise, reflect on, revise and improve their writing independently. “The best thing is that we have the opportunity to edit our work whenever we want. It's good because we have time to think about our mistakes and then correct them.” “It's clear, easy to understand, but makes you think as well. I was pleasantly surprised when the system gave me the derivation of the word I used incorrectly.” [E7]

Teachers also recognise the benefit for themselves and their students. A review article on the ELT [English Language Teaching] news website explains: "I definitely recommend “Write&Improve” to all teachers and students. It is not only free and easy to use but also it helps students redraft and correct in a pleasant and stress free environment, something that most of them have never done before and this is what really makes the difference." [E8]

Write & Improve also helps to reduce a teacher’s workload. A Senior Research Manager at Cambridge Assessment described the tool as a way “to keep the learner motivated and engaged, and to help them identify and eliminate common and repeated errors, so that their teacher can focus on those issues which require ‘human’ support, such as discourse organisation, argumentation or nuance” [E6]. It was reported from two Turkish Universities who have been using the tool that “Teachers found it useful for drafting, focusing on language outside of class and practicing language. It allowed them to focus on different tasks and provide more salient feedback.” Students were using the tool 30+ times to redraft homework. This allowed them to correct many issues before handing in the assignment, so that teacher feedback was reduced, and the “demotivating” effect of many minor corrections was reduced. [E9].

In December 2019, ELiT was fully acquired by Cambridge Assessment and Cambridge University Press [E10] [text removed for publication] [E11]. The press release described ELiT as ‘world-class experts in using AI to support English learning…they have already worked with Cambridge English on ground-breaking projects and resources for learners’ [E10]. ALTA continues to support the R&D base of the company [E12].