Microsoft’s AI research team has collaborated with researchers from the University of Washington to develop a novel smart to-do feature that automatically generates to-do lists from emails.
As simple as that sounds, this technique would help people streamline their workflow and minimize the efforts taken to put all tasks and deadlines together to keep track of them in a more efficient manner.
For instance, say you’ve sent an email to your coworker or guide proposing a deadline for submitting a project. The AI will automatically detect the instance from your email and create a to-do entry corresponding to the context. This means you will be able to take a peek at this auto-generated to-do list and know when a deadline is creeping up on you.
As per the research paper of the project spotted first by The Next Web, the researchers have developed a two-stage framework based on neural text generation and sequence-to-sequence learning. The dataset for the project was created based on the Avocado email corpus.
The process involves identifying the task sentence (query), relevant sentences in the email that supports and gives more details about the query, and timelines. The task sentence is identified using a commitment classifier and the researchers then used an unsupervised extractive algorithm. They are combined using a sequence-to-sequence architecture built with deep neural networks. Take a look at the flowchart below for a better understanding.
Based on initial experiments, the researchers say their model performs at par with human performance while writing To-do lists, which sounds really promising. The researchers will present the paper at the annual conference of the Association for Computational Linguistics (ACL 2020), which will be held online from July 5 to 10 due to the coronavirus pandemic. Let’s wait and see if our email clients will be able to offer an important feature, like automatic to-do list creation, in the coming years.