For robots, it is difficult to get information from a statement like a human. Therefore, before training the robot, there is usually a process of screening the training materials, so that it is easier for the robot to find modular information in a large amount of data after being fed into the data.
This type of training is more suitable for computer technology and only takes time to accumulate, but it is a problem for people's commonly used language.
This is related to how people express themselves. Chris Callison-Burch, assistant professor of computer and information science at the University of Pennsylvania, said: "One of the many difficulties in natural language is that you can express the same information in many different ways, and catching all these changes is to build a complex One of the challenges in the model process."
The straightforward point is that the language used by humans is more complicated and cannot be directly used as training materials for robots.
But three MIT researchers, including Regina Barzilay, have recently come up with a new perspective. They want the robots to look for information online and to find more information to supplement information. For research, the most direct benefit of doing so is that it can reduce the labor cost of the previous period.
To put it simply, they used a technique of artificial intelligence - reinforcement learning, to test two application scenarios. One is to find information in the news of the US shootings, including the location of the shooting, the number of casualties; the other is to collect similarities in food fraud, including the type of food, the type and location of food fraud.
Only 300 news items are used for each test. However, in the test, every time an MIT-trained robot analyzes an article, it will find 9 or 10 news articles on the Internet as relevant data for analysis to supplement the lack of the original information. The difficulty is that the robot judges that the two news describes the same thing and extracts data from multiple sources.
The paper cites a case of murder. The following is a snippet from three news stories:
Police officers announced on Monday that a couple and four children died in their home in South Dakota. Their home had a fire. They apparently died from murder... An official statement said that Scott Westerhuis The cause of death was "a shotgun shot suspected of suicide."
According to the official, six members of a South Dakota family were found dead in a home burned by fire. They died of shootings, and one of them was suspected of suicide.
AG Jackle said that all the evidence supports his reasoning based on the preliminary findings of September: Scott Westerhuis shot his wife and children with a shotgun, lit his house, and committed suicide with a shotgun.
In the end, the robots they train need to find news like the following two articles online, and dig out two pieces of information: the person who robs is Scott Westerhuis, and the death toll is six.
In both tests, the MIT-developed system was able to search based on vocabulary related to the target information. For example, the murderer of the shooting case would normally be arrested with “policeâ€, “idenTIfiedâ€, “arrested†) "When words are related, they can judge specific people.
Researchers at the Massachusetts Institute of Technology say they use this new technology for data collation, which is 10% more accurate than previous screening methods. However, for the average person, the interesting thing about this new technology is that it allows the robot to work much closer to how humans think: after getting a message, supplement it with more information.
Custom LFP Battery Pack Advantages Compared to other battery options, custom lithium iron phosphate battery packs provide superior thermal and chemical stability. This allows for improved safety characteristics over lithium-ion batteries that utilize less-safe cathode materials.
Hot-swappable Internal LFP Battery Pack Average Nominal Voltage: 25.6 VDC Nominal Energy: 266 Wh providing 12 minutes run time for 1kW load Weight: 8 lbs; 43% improvement over lead acid battery packs Recommended Charge: Up to 8 A Discharge Peak Rate: 80 A Operating Temperature: -20 °C to 60 °C Storage Temperature: -40 °C to 70 °C
For a safe and reliable power solution, GLIDA Lithium Ion packs are solid choice for your application needs.
With extensive Quality Control certifications and procedures, GLIDA gives it's customer the confidence in every product offering.
12V Lfp Battery,Lfp Battery Pack 48V,Lfp For Li-Ion Battery,Lfp Lithium Ion Battery Pack
Shenzhen Glida Electronics Co., Ltd. , https://www.szglida.com