Author Identification: Using Text Mining, Feature Engineering & Network Embedding

Authorship analysis is a challenging area that has been developed through centuries and with research done widely scattered across multiple disciples of mainly computational linguistics, text mining, data mining, stylometry and machine learning. Conventional techniques from the past relied heavily on stylometry and text-based content analysis of document text for authorship analysis. More recent developments use network embedding training focus heavily on document attributes to build a network and predict the author. We propose a system that incorporated the strengths of both text mining and network embedding methods and utilizes both the document text and document attributes fully when available. In this paper, we describe the system overview and implementation in detail and discuss how by supporting a more multi-faceted information based network embedding, it can be possible to get improved results. Finally we discuss our results and suggest some future improvements in terms of results, speed and performance for our system to handle larger corpus