Table 1.
Table detailing the construction of searches based on the combination of two inclusion terms and a proximity operator for the inclusion terms.
In one case an exclusion term was used to avoid extraneous literature.
Fig 1.
PRISMA flow diagram detailing screening process of articles included in analysis.
Fig 2.
Venn diagrams: of (a) journal articles retrieved from Web of Science (WoS), Scopus, ScienceDirect, and AGRICOLA; and of (b) journal articles queried per crop.
Fig 3.
Publication counts per crop by decade, 1930–2015.
Fig 4.
Counts of literature on perennial wheat in the top five publications from the periods of 1930–1960, 1961–1990, and 1991–2015.
Fig 5.
Counts of literature on perennial rice in the top five publications from the periods of 1930–1960, 1961–1990, and 1991–2015.
Table 2.
Summary of topics generated for each topic modeling approach.
For complete lists of terms with high posterior probabilities for each topic see S1 Appendix.
Fig 6.
Counts of journal articles assigned by the Gibbs model of the entire library to each of three topics by decade.
The five topic terms with the highest posterior probability of association with each topic are listed with their corresponding topic in the legend.
Fig 7.
Counts of journal articles assigned by the Gibbs model of the wheat collection to each of three topics by decade.
The five topic terms with the highest posterior probability of association with with their corresponding topic in the legend.
Fig 8.
Counts of journal articles assigned by the Gibbs model of the rice collection to each of three topics by decade.
The five topic terms with the highest posterior probability of association with with their corresponding topic in the legend.