Plinydarpa automatically fill in Code machine learning now when you enter text in Google's search box, it automatically conjures up what you need. For example, enter "Wi", the Entry "Wikipedia" Automatically Lenovo, enter "Bra", Google can guess you want to enter "Brad Pitt." While it's possible that Google's automatically filled out search content can be quite different from the user's own thoughts, it does make it easier to guess exactly what you want to search for in the process of redundancy.
now a government-funded team is working on a similar project that allows the world's programming staff to automatically associate computers when they write code. The essence is that computers can guess what they want to knock on before programmers enter code.
this week, Rice University said Darpa, the Pentagon's Advanced Research Projects Agency, has invested 11 million of dollars in programming automatic projects, naming Pliny as the Roman writer's name. "Google's text-search predictions are just an example, and then people will solve more problems in a similar way," says Vivek Sarkar, Dean of the School of Computer Science at Rice University and a major researcher at the Pliny project. ”
to anticipate code lookup vulnerabilities before knocking the code
The
project is attended by institutions and research institutes, including Rice University, University of Texas at Austin, University of Wisconsin, and GrammaTech, which specializes in developer tools. The Pliny Project will index the open source code of the Internet in Shanghai to build a code prediction engine, which in theory Pliny can also locate bugs and security vulnerabilities. If the project is implemented, the High-tech companies that are committed to developing complex software programs will no longer have to worry about recruiting good programmers who don't have enough staff.
Pliny was not the first to try to solve the code-filling project, which Microsoft released Bing Developer Assistant (Bing Developer Assistant) last year. Sarkar says Pliny is a more significant project, "most of the other projects do just a simple textual analysis of the opportunity code structure." ”
Sarkar and his team are trying to create a software that uses different programming languages, not only to analyze text, but also to read the meaning of code expression. Sarkar hope Pliny can automatically fill in the long string code in the future, and check the programmer's language errors and security vulnerabilities.
how to make reasonable and correct predictions?
for the Pliny project, the most difficult part is how to do reasonable and correct prediction code. If you've ever used a paper-clip assistant in Microsoft Office software, Mr. Clippy, or automatically updated your blog assistant, you can see how difficult it is to get your computer to fill out the correct answers automatically. What Google can do is simply rely on the "search hot words" of tens of thousands of netizens to anticipate the search that you might enter. This is not the case for code prediction, which is generally not the best solution for the most widely used code segments.
Sarkar admits that solving the legitimate correctness of automatic completion is the toughest challenge in the Pliny project, but he believes their team is the right choice for the project. Team members have experience in large data analysis in the energy sector and medical research institutions. Sarkar said that Rice University has been hoping to apply machine learning algorithms to the software development field for many years, and DARPA has offered this precious opportunity.
Pliny's Project begins with a study of open source code in the Internet, usually from GitHub and SourceForge hosting services, and other Open-source projects such as Apache Foundation. In addition, Pliny wants the code to automatically fill out the project to develop an enterprise version that indexes the software programs of large companies and businesses.
Pliny also establishes a database system for storing and analyzing code, which provides a number of ways to build optimized code indexing and improves code quality. Programs can automatically recognize "special code" and give priority.
said the results of the Pliny project would look similar to Google's Lenovo search, but it would be more practical.
via Wired
