Pattern System: An Rule Based Intent Recognition Platform

In my junior year, having acquired a solid technical background in CS major through coursework and side projects, I decided to improve my research capability by joining a research lab in Microsoft Research Asia(MSRA). After knowing about my research interest, my mentor Zijia Lin at MSRA decided to let me take over a project in his group, focusing on the research of NLP area.

I was assigned an demanding work: building the front and back ends of the Pattern System. The Pattern System aims to solve the intent recognition in natural language processing(NLP). This project is mainly to provide users with a rule-based online platform, users can write Pattern on the platform to match the Intent in Utterance, and this intent recognition can be applied to the chat robot. Here, I implemented the HTML CSS Jquery architecture in C# MVC4 frame for the front end, and Microsoft SQL server for the back end. Before I undertook this website, it had a prototype with approximately ten thousand lines’ codes. Following the guidance of my mentor, I studied the codes within a week, and independently learned all the architecture needed. Despite the workload, I finished the coding with great efficiency and effectiveness. Eventually total codes of 27,000 lines of websites were implemented and then introduced in Microsoft’s internal operations and used by Microsoft employees.

Although this project’s codes was not open-source, I was allowed to make a PPT to display the basic function and interface of Pattern System. Click here to view the PPT.

Comments