Recently, the “2019 Machine Reading Comprehension Competition” jointly organized by the Chinese Computer Society, the Chinese Information Society of China and Baidu Company ended in perfection. After 2 months of intense competition in the world’s 2502 teams, deep thinking about artificial intelligence ranked first in both evaluation indicators.
Recently, the “2019 Machine Reading Comprehension Competition” jointly organized by China Computer Society, China Information Society and Baidu Company ended in perfection. After 2 months of intense competition in the world’s 2502 teams, deep thinking about artificial intelligence ranked first in both evaluation indicators.
It is worth mentioning that in the “Machine Reading Comprehension Competition” in 2018, the deep thinking of artificial intelligence also stood out among more than 1,600 domestic and foreign teams, ranking third and winning the global award. This time, the grand prize of the competition, on the one hand, proves the deep thinking of the frontier strength in the field of global machine reading comprehension; on the other hand, while the competition greatly promotes the progress of reading comprehension technology, it will also deeply consider this research and development-oriented, Entrepreneurial companies that continue to lay out the scenes are more likely to be present in the world.
What is the machine comprehension?
The so-called machine reading comprehension, the concept is basically similar to the reading comprehension made by everyone in the student era. It is also given a piece of material and questions, giving the correct answer, but the protagonist is changed from human to AI model.
Although machine reading comprehension seems to only let AI come to an exam, it is the biggest challenge in natural language processing technology after speech judgment and semantic understanding: let the agent understand the full text context.
In a nutshell, speech recognition helps the machine “listen”, and image recognition helps the machine “see”, but how the machine understands and understands is a problem solved by semantic understanding.
Comparative to spell checking and automatic translation, semantic understanding is not just computation and recording, but active analysis and understanding, so reading comprehension has always been considered the landmark tip of natural language processing (NLP).
One of the keys to natural language processing is semantic understanding, because machines can’t do human understanding of natural language, such as the beginning of the article. Therefore, semantic understanding has always been regarded as “the jewel in the crown of artificial intelligence”, which combines linguistics, computer science, and artificial intelligence. Its purpose is to “make machines understand natural language.”.
Semantic understanding has also been an important direction for research and capital attention. According to the statistics of Tencent Research Institute at the end of 2018, among the top artificial intelligence companies in China, the top three areas of financing are computer vision and image, natural language processing, and automatic driving/assisted driving, while ranking second in nature. Language processing, financing 12.2 billion yuan, accounting for 19%.
“Like people go to watch TV, people look at the picture at the same time, listen to the sound, look at the subtitles to understand, and the work done by the human brain is multimodal semantic understanding.” Deep thinking about artificial intelligence CEO and AI Algorithm scientist Yang Zhiming mentioned in the previous “2019WISE Super Evolution”.
In fact, some important events and indicators of machine reading comprehension are being constantly refreshed and broken. In the most authoritative DuReader dataset in Chinese, the two indicators of ROUGE-L and BLEU-4, human performance is 57.4 and 56.1, and the BMAnet model developed by the deep thinking has achieved 63.13 and 59.34 performance on these two indicators.
Machine reading comprehension is an important task of machine understanding and an important part of semantic understanding. The breakthrough of data indicators in machine reading comprehension often represents a breakthrough in machine understanding or semantic understanding.
Thinking about the technical breakthroughs that will be achieved in real-life applications?
Dream more than “winning” – deep thinking about implementing multiple application scenarios
At present, the deep thinking mainly lies in the application scenarios of smart car network digital cockpit, car smart marketing, mobile smart mobile terminal, smart home, smart medical health and so on. Cooperation customers include Chery, Huawei, Jiuyang, Xiaomi, SAIC, Keda Xunfei, Zhejiang University First Hospital, and Fudan University Affiliated Tumor Hospital.
The main application scenarios include:
Smart Car Network Digital Cockpit Scenario: Number of Smart Car NetworkingIn the cockpit scenario, the deep thinking is based on the multimodal deep semantic understanding and human-machine dialogue engine (iDeepWise.ai), providing multi-modal information (speech, text, visual) perception and understanding inside and outside the digital cockpit of the smart car network, v2x Information Awareness and Understanding, Terminal Multimodal Deep Semantic Understanding AI DPU chip edge computing and other three core capabilities of the “multimodal semantic understanding and human-car interaction brain” (iDeepWise.ai.car) AI SaaS service.
iDeepWise.ai has multiple rounds of contextual understanding. The machine has contextual “memory” and personalized interaction and understanding, and ultimately establishes information between cars and homes, cars and people, and people and home. Link, one-stop AI service that realizes multi-scenario fusion.
Automotive Smart Marketing Scenario: Deeply think about providing multi-modal information for car sales scenarios for deep semantic understanding and human-machine dialogue iDeepWise.ai.sales AI SaaS service to help auto customers Multi-dimensional tracking and analysis of potential users, increase order conversion rate, and create big data entry for large-scale car buyers.
Mobile smart mobile terminal interaction scenario: Deeply think about the AI SaaS service of iDeepWise.ai.mobile, which provides intelligent human-machine dialogue for travel, health consultation, smart office, leisure and entertainment. Especially in the field of travel, it provides one-stop AI intelligent travel life service for 200 million intelligent terminal users, including automatic completion of booking ticket train tickets and automatic completion of hotel reservations through man-machine dialogue.
Smart Home Human-Computer Interaction: Deeply think about the semantic understanding and multi-modal dialogue of intelligent multi-modal information such as smart speakers and smart home appliances. iDeepWise.ai.Home AI Saas service to realize family Convenient operation, interactive audio-visual entertainment, nutrition and healthy diet recommendations for home-based equipment.
Smart Healthcare: Deep thinking about products based on multimodal deep semantic understanding and human-machine dialogue engine (iDeepWise.ai), providing visual semantic understanding of pathological cells. ideepwise.AI CIAS and people AI Saas service for dialogue-based health consultation, in which the artificial intelligence-assisted reading system (ideepwise.AI CIAS) helps doctors to screen and diagnose cervical cells, reduce the workload and work intensity of pathologists, and improve the efficiency of cervical cancer screening. And diagnostic accuracy, can achieve large-scale screening services for cervical cancer; human-machine dialogue-type health consultation AI services can allow people to achieve authoritative consultation on healthy nutrition such as maternal and child, chronic diseases through human-machine dialogue.
In fact, the most fundamental reason why artificial intelligence has become the current cusp and hotspot is that the application scene of artificial intelligence has begun to land – this makes the commercial imagination of artificial intelligence unprecedentedly upgraded and enlarged. As a leader in the field of artificial intelligence multi-modal deep semantic understanding, deep thinking through the continuous multi-scenario landing, it also proves that the future of artificial intelligence is worth our expectation.