Salesforce Metamind have released a paper that details how machines could summarise emails and documents.
The Salesforce Metamind research team have published the paper that could change how emails and documents are handled in the workplace by using intelligent machines to summarise them into bullet points.
Some workers can spend up to 6 hours reading and responding to emails, and this new technology could reduce this time by up to half. Ultimately, this has the potential to increase productivity and save businesses money.
The newly developed algorithm has shown some initial success and was able to sum up an entire New York Times article about Facebook combating Fake News ahead of the UK election into three main bullet points.
- Social network published a series of advertisements in newspapers in Britain on Monday.
- It has removed tens of thousands of fake accounts in Britain.
- It also said it would hire 3,000 more moderators, almost doubling the number of people worldwide who scan for inappropriate or offensive content.
Whilst the software is still not able to summarise as accurately as a human being, this new algorithm is a big step in a new direction. In order for the machine to fully articulate a summary it would require a deep understanding of human language and be much more intelligent than anything produced to date.
It is hoped that the technology will eventually be able to be fully automated and deployed in the workplace.
The machine is ‘taught’ through reinforcement learning, which is similar to how animals learn. By giving the machine positive feedback for working in a certain way, the machine is able to determine the better course of action.
The machine is also taught through supervised learning, which is the process of the machine learning by example. By being ‘fed’ good examples of summaries the machine is able to use its intelligence to determine how to behave.
The startup Metamind was acquired by Salesforce in 2016, and is headed by machine learning alumnus, Richard Socher.
