A corporal and LDA analysis of abstracts of academic conference papers
Sébastien Louvigné, Jie Shi, Yoshihiro Kato, Neil Rubens, Maomi Ueno · 2013
The academic Abstract genre is one of the most important parts of a research paper because it is viewed at first by readers. When well written, it enables them to quickly and accurately identify the main ideas and contributions of the work presented. Hence writing an abstract is the most difficult part. Poorly written abstracts can affect negatively the significance and understanding of the paper. In order to investigate the rhetorical moves (or communicative stages) creating the meaning of abstracts, this study analyzes a corpus of 1,500 international conference papers (ICALT) using Corpus Linguistic tools and LDA (Latent Dirichlet Allocation) algorithm. This study shows that abstracts are written focusing on research purpose and methods rather than describing the background and commenting results. LDA algorithm provides a list of topics within the corpus of abstracts based on research areas. Future works on LDA algorithm should strengthen its semantic approach based on the rhetorical moves.