AI-Driven Language Processing Could Shorten Our Emails

Like many people in the early weeks of a new year, you may well be seeking ways to improve efficiencies in your company for the coming 12 months. Time wasted by office workers trawling through emails and documents could be one area for you to examine.

Here’s why: A recent report by McKinsey management consultants estimated that employees typically spend 28% of their working week on reading and answering emails, which works out to about 2¹/₂ hours per day per staffer.

Other studies suggest that there’s an additional  “interruption effect”Opens a new window that lasts as long as 23 minutes from the time your staffers hit the send button until they return to their original tasks.

The distractions of that deluge of emails and documents aside, the environmental impact of fossil-fuel energy used in data centers and the internet to ferry emails to their digital destinations is becoming a major concern.

Mike Berners-Lee, a professor at Britain’s Lancaster University environment center (and brother of World Wide Web originator Sir Tim Berners-Lee) noted the “huge and growing” carbon footprint of ITOpens a new window .

A study commissioned by the British energy company OVO reported by The Guardian estimated that Britons send more than 64 million unnecessary emails every day. And if every adult sent one fewer “thank you” note a day, the UK would save more than 16,000 tons of carbon a year — the equivalent of 81,500 London-to-Madrid flights.

Even as the number of collaborative software tools and efficiency-friendly apps increases, the need for your employees to read emails and documents is unlikely to disappear completely. But what could help save time dealing with office emails is an emerging artificial intelligence technology called “automatic summarization.” Simply put, the tech can shorten emails and documents automatically without dumbing them down.

The AI challenge

Distilling chunky paragraphs into shorter sentences doesn’t come easy for AI because it generally relies on a semantic understanding of text, a human specialty, and is beyond the reach of many natural language processing models. But progress is being made.

Google Brain, in collaboration with London’s Imperial College, has devised the Pegasus systemOpens a new window , or as the researchers call it, “pre-training with extracted gap-sentences for abstractive summarization sequence-to-sequence.”

Leveraging Google’s Transformers neural architecture with pre-training objectives specifically for abstractive text summarization, the researchers analyzed 12 data sets spanning news, science, short stories, instructions, emails, patents and legislative bills, with what they claimed were state-of-the-art results.

They reported that abstractive summarizations aim to generate accurate and concise summaries from documents, in contrast to extractive summarization, which merely copies fragments from documents. The abstracts may generate “novel words” as they go along, the researchers wrote, but they also capture principal information in concise sentences that are “linguistically fluent.”

The team used a technique called “gap sentence generation” that masked key sentences from documents and then used AI to work out what was missing from the wider body of text.

Into the fray

Microsoft, which is also researching text summarization, released a paper in 2018 about its two-step approach to summarize textOpens a new window . Its researchers used an AI model that processed an input sequence and predicted the next characters of a target sequence with a neural network that learned from graph representations of annotated natural language.

Another Silicon Valley company, Primer,Opens a new window builds machines that can read, write and analyze large data sets. It helps the national intelligence community and corporate clients with large-scale data processing. Customers feed documents into its machines and its software selects key information, which can then be searched by topic and other categories.

In December, IBM showed off its “Project Debater” software at the University of Cambridge’s debating forum. The system can extract and categorize arguments from text or audioOpens a new window , summarize them and present them via synthesized speech. It was “trained” on more than 30,000 statements which human reviewers rated for their persuasiveness.

The same month, a team at Ben-Gurion University in Beersheba, Israel, unveiled a multilingual sentence extractor, a system that the Ben-Gurion team said can translate and summarize text in nine languages.

Just where is the technology going? A recent report from the Research and Markets organizationOpens a new window  forecast that the global Natural Language Processing market, which includes machine translation, information extraction and automatic summarization, will more than double from a value of $10.2 billion in 2019 to $26.4 billion by 2024.

The rewards will motivate tech companies and universities around the globe to develop the technology. Engineers and researchers will undoubtedly continue to make remarkable advances in AI-driven language processing.