Given the high number of scientific papers that are published every year, it is a challenge to observe the evolution of scientific fields. While, for example, computer science and biology were considered rather unrelated 40 years ago, today, bioinformatics is a well-established field. One way to analyze these questions is to observe the evolution of the term co-occurrence graphs of the abstracts of scientific publications. In a term co-occurrence graph, two terms are connected if they appear together in an abstract of a publication. We weight the edges of this graph with the number of common occurrences. We analyze the evolution of this co-occurrence graph, that we constructed for each year, on the basis of a large collection of scientific articles, with the help of spectral techniques. We present our preliminary observations and discuss our ongoing work.