Multiple Correspondance Analysis of annotated data about Multi Criteria Assessment Methods used in the agri-food research

The annotation of scientific papers has been done in three steps described in detail in a related data paper (Gésan-Guiziou et al. 2019a). Firstly, the set of papers published by INRAE researchers from 2007 to 2017 have been extracted from the Web of Science (WOS) using a set of key-words. These search queries were performed mid-2017. The resulting corpus of papers (4920 papers) has been manually typed MCA (954 papers) or non-MCA (3966 papers) by domain experts using a set of positive and negative criteria defining the notion of MCA articles. Scientific articles dealing with MCA in the AgriFood sector have been associated with annotations. Finally, a group of INRAE experts in the field of MCA and/or application domains classified those articles according to eight criteria: Type of study, Purposes, Audience, Assessed dimensions, Assessed system/object, Spatial scale, Time scale, Actors’ contribution. Annotation files are available on a public repository using DOIs listed in related data paper (Gésan-Guiziou et al. 2019a). Each of the eight criteria having between 5 and 10 modalities, a total of 58 binary values, representing the present or absent modalities, characterized each article. In other terms, the data set consisted of a cloud of 954 points in a 58 dimensional space. We performed a Multiple Correspondance Analysis on this dataset, in order to analyse the structure of this cloud, and to extract clusters of homogeneous subsets of publications. We used the R package FactomineR, to perform this study.