Artificial language comprehension may be improved by reading Harry Potter, if the results of a new study are developed. This study could, one day, advance the field of artificial intelligence.
Maluuba is a Canadian company centered on teaching machines how to learn, and researchers there now utilize the works of J.K. Rowling.
Harry Potter and the Philosopher's Stone was one of the hundreds of books studied by machines designed to study artificial intelligence. An algorithm was applied to the text, providing the machine with the ability to answer questions about the story.
Using this formula, the text was examined at several levels, including paragraphs and sentences, in addition to individual words. This could allow computers to not only record, but actually comprehend stories presented to their artificial brains. Reading comprehension tests, in the form of multiple-choice questions, were presented to the artificial intelligence (AI) systems, which answered the inquiries with 70 percent accuracy. Such systems may be adapted to take over mundane reading tasks, which many people gloss over or ignore entirely.
"We're interested in use cases like user manuals, patient records, or customer service documents. In those areas, you really don't have a glut of data," Mohamed Musbah, Maluuba vice president of product, said.
Comprehending text is one of the more challenging subjects for researchers in the field of artificial intelligence. This is because knowledge of the outside world is needed to understand many of the nuances of the written word. Without this knowledge, computers are unable to distinguish meanings between sets of words that are spelled the same, despite having radically different meanings.
Some of the largest technology companies in the world, including Facebook and Google, are currently working on AI systems capable of comprehending written language. This requirement for real-world knowledge to understand written text is currently being satiated through the use of vast amounts of annotated words.
Microsoft recently produced a new chatbot named Tay, designed to interact with the public on Twitter and other social media. However, users quickly tricked the system into producing responses with sexual and racist dialog. The company later took Tay offline while they tweak the system.
A technique known as deep learning was employed by researchers, aimed at teaching machines how to comprehend stories. Typically, machines utilizing the learning approach are randomly configured prior to training. In this study, investigators carefully prepared the network for learning, allowing the network to quickly learn new material. This approach resulted in a 15 percent increase in the reading comprehension of the new system.
Although true reading comprehension by an AI system is still in development, a tale of a boy wizard could bring that possibility one step closer.
