Digital Science Invests in Deep Learning Language Platform Writefull
Writing on a computer today, most of us now expect our writing support tools to offer grammar and spell checkers. These features use hard-coded rules to assess if a sentence is correct following the rules of English. The problem with this rule-based approach is that in many cases the rules are not clearly defined. For example, in the use of prepositions, we sit ‘on’ a dining chair, and yet we sit ‘in’ a rocking chair. In other cases, rules do exist, but they fail to address how language is really used.
Today, we are excited to announce our investment in Writefull – a deep learning language platform applied to discipline-specific scientific texts to help improve the clarity of written English. We believe Writefull will help authors to express their work more clearly before they submit articles for publication, especially those who have English as their second language (and quite a few of us native speakers too for that matter). In addition, publishers will have a service to help relieve the administrative burden on editors, maintain house style and to help with quality control.
“When we first met the founders they demonstrated Writefull highlighting a sentence that read “…the tall mountains and high trees”. Although grammatically correct, a native speaker (subconsciously) wouldn’t describe mountains as “tall” and trees as “high” but instead as “…high mountains and tall trees”… Writefull points users to change this. We were impressed.”
Writefull helps authors improve the clarity of their work. It suggests improvements to grammar and spelling and to academic language usage such as sentence structures in scientific writing, discipline-specific vocabulary and appropriate word choice. These suggestions are based on real-world, context-specific usage rather than on a fixed set of grammatical rules.
When we first met the founders of Writefull, Juan and Alberto, they demonstrated the AI highlighting a sentence that read “…the tall mountains and high trees”. Although grammatically correct – as a native speaker, something does not sound quite right about that sentence. The nuance of English is such that we (subconsciously) wouldn’t describe mountains as “tall” and trees as “high”. The recommendation picked up by Writefull’s Deep Learning and N-gram approach pointed the user to change this to “…high mountains and tall trees”. Language is full of such aspects of usage, and these don’t have fixed rules that can be hard-coded in advance.
We were impressed.
In making our investment in Writefull, we now have a solution that steps beyond rules-based approaches to cover style and usage, and applies that technology specifically to scientific writing, training the AI to understand the vocabulary and style within that domain.
The founders, with their machine learning and AI backgrounds, have a feature-packed product roadmap and we here at Digital Science look forward to working with them to help the rest of us express our ideas more clearly.
Read our full press release.