Distributional semantic analysis in a technical corpus

ABSTRACT

This paper presents the methodology of a distributional analysis for the exploration of semantic similarity in a small technical corpus. The aim is to show the semantic issues at stake in a series of statistical experiments. First-order co-occurrences of a technical node are clustered based on shared-order second and third-order co-occurrences and on the respective association strength. The results of the statistical clustering techniques show semantic similarities and dissimilarities between first-order co-occurrences by means of proximities and distances on 2D-plots and 3D-plots. By fine-tuning the statistical analysis and by using enriched distributional data, we aim at achieving more revealing semantic interpretations.

KEYWORDS:

• Distributional semantics
• multidimensional scaling (MDS)
• semantic similarity
• second-order and third-order cooccurrences

Disclosure statement

No potential conflict of interest was reported by the author.