TAN PHING WERN
At Universiti Sains Malaysia, Tan Phing Wern is majoring in Intelligence System and taking Networking as elective course. She is a developer actively involved in web application development, especially on RESTful API, backend server and has experiences working on PHP, Python, Java and C# project. Other than that, majoring in Intelligence System has exposed her to the scope of machine learning, deep learning as well as natural language processing. Recently she finished a project named Natural Language Generation for Data Analytics which mainly working on natural language generation from structured data using rule-based natural language generation.
Matrix No:
137169
Student Email:
Supervisor:
Dr. Gan Keng Hoon
Supervisor Email:
SC056
Natural Language Generation for Data Analytics
Natural Language Generation for Data Analytics generates natural language summary for structured data, in this case a visual representation of data also known as chart and graph. The purpose is to allow user to get the gist of visual representation using words summary. For example, a chart showing up and down trend line is interpreted in friendlier natural language as "The emission increases from 1 ( 1850 ) to 150 ( 1980 ) then decreases from 150 ( 1980 ) to 1 ( 2011 )". In this project, this tool is applied on Expert Search and Analytics visualization components to provide natural language summary for each component in dynamic manner. Rule-based natural language generation is used in generating the text, which means some sentence templates are pre-defined and filled with data during text generation process.
In addition, with similar intention, a Microsoft Office Add-in application and a Chart Description API is also developed. The add-in aimed to generate chart description for simple charts in Word document. This add-in uses Office JavaScript API to interact with content in the Office document where the add-in is running, in this case the content will be the chart object. To generate a chart description, the add-in will extract the data from the selected chart and send a request to the API. After processing, the Chart Description API will respond with generated text summary. Both add-in and Chart Description API are currently hosted on Google Cloud Platform.