DM Network Analysis Attempt
Jack Grieve
27/10/2020
1 Introduction
This is just a quick attempt to build a directed network-based model of the relationship between modal verbs in double modal (DM) constructions based on Twitter data using Markov models. Basically, the idea is to visualise relationships between pairs of modals based on which modals tend to combine.
2 Packages
library(markovchain)## Package: markovchain
## Version: 0.8.5-2
## Date: 2020-09-07
## BugReport: https://github.com/spedygiorgio/markovchain/issueslibrary(igraph)##
## Attaching package: 'igraph'## The following objects are masked from 'package:stats':
##
## decompose, spectrum## The following object is masked from 'package:base':
##
## unionlibrary(papeR)## Loading required package: car## Loading required package: carData## Loading required package: xtable## Registered S3 method overwritten by 'papeR':
## method from
## Anova.lme car##
## Attaching package: 'papeR'## The following object is masked from 'package:utils':
##
## toLatex3 Data
The data is just a frequency matrix, with indivudal modal verbs as rows and columns, where the frequency in each cell being the frequency of the DM formed by combining the modal in the row in first position with the modal in the column in second position.
DM <- as.data.frame(read.table("DM_FREQ.csv", sep=",", header=TRUE, row.names=1))
DM4 Network Visualisation
So now we want to make a network graph that shows the relationship between 11 modals. And I think we probably want magnitudes reflected here, so like we don’t want to row normalise or make a transition matrix, for example, especially given uneven the distribution is.
DM <- as.matrix(DM)
IG <- graph_from_adjacency_matrix(DM, weighted=TRUE, mode = "directed")
IG## IGRAPH 27cac52 DNW- 11 76 --
## + attr: name (v/c), weight (e/n)
## + edges from 27cac52 (vertex names):
## [1] can ->could can ->may can ->might can ->must
## [5] can ->shall can ->should can ->will can ->would
## [9] could->can could->may could->might could->should
## [13] could->will could->would may ->can may ->could
## [17] may ->might may ->must may ->ought_to may ->shall
## [21] may ->should may ->used_to may ->will may ->would
## [25] might->can might->could might->may might->must
## [29] might->ought_to might->shall might->should might->used_to
## + ... omitted several edgesplot(IG)
Lots of infrequent links here so we can trim and re-plot…
DM.trim <- DM
DM.trim[DM.trim < 20] <- 0
DM.trim## can could may might must ought_to shall should used_to will would
## can 0 135 21 0 0 0 0 20 0 111 0
## could 60 0 0 20 0 0 0 0 0 0 41
## may 273 53 0 0 0 0 0 0 0 22 0
## might 1733 932 0 0 0 32 0 146 0 52 243
## must 164 0 0 0 0 0 0 0 0 0 0
## ought_to 0 0 0 0 0 0 0 0 0 0 0
## shall 0 0 0 0 0 0 0 0 0 0 0
## should 0 33 0 0 0 0 0 0 0 0 177
## used_to 0 144 0 0 0 0 0 0 0 0 20
## will 111 20 26 59 0 0 21 23 0 0 80
## would 0 171 0 29 0 0 0 81 40 52 0IG <- graph_from_adjacency_matrix(DM.trim, weighted=TRUE, mode = "directed")
IG## IGRAPH 35a2ebc DNW- 11 33 --
## + attr: name (v/c), weight (e/n)
## + edges from 35a2ebc (vertex names):
## [1] can ->could can ->may can ->should can ->will
## [5] could ->can could ->might could ->would may ->can
## [9] may ->could may ->will might ->can might ->could
## [13] might ->ought_to might ->should might ->will might ->would
## [17] must ->can should ->could should ->would used_to->could
## [21] used_to->would will ->can will ->could will ->may
## [25] will ->might will ->shall will ->should will ->would
## [29] would ->could would ->might would ->should would ->used_to
## + ... omitted several edgesplot(IG)
If we trim down too much though so that some modals have no links, then the graph kinda breaks down….
DM.trim[DM.trim < 50] <- 0
DM.trim## can could may might must ought_to shall should used_to will would
## can 0 135 0 0 0 0 0 0 0 111 0
## could 60 0 0 0 0 0 0 0 0 0 0
## may 273 53 0 0 0 0 0 0 0 0 0
## might 1733 932 0 0 0 0 0 146 0 52 243
## must 164 0 0 0 0 0 0 0 0 0 0
## ought_to 0 0 0 0 0 0 0 0 0 0 0
## shall 0 0 0 0 0 0 0 0 0 0 0
## should 0 0 0 0 0 0 0 0 0 0 177
## used_to 0 144 0 0 0 0 0 0 0 0 0
## will 111 0 0 59 0 0 0 0 0 0 80
## would 0 171 0 0 0 0 0 81 0 52 0IG <- graph_from_adjacency_matrix(DM.trim, weighted=TRUE, mode = "directed")
IG## IGRAPH e3b1a6a DNW- 11 19 --
## + attr: name (v/c), weight (e/n)
## + edges from e3b1a6a (vertex names):
## [1] can ->could can ->will could ->can may ->can
## [5] may ->could might ->can might ->could might ->should
## [9] might ->will might ->would must ->can should ->would
## [13] used_to->could will ->can will ->might will ->would
## [17] would ->could would ->should would ->willplot(IG)
Not totally sure how to easily remove variables with all zeroes in rows AND columns, so just remove manually and rerun…
DM.trim <- subset(DM.trim, row.names(DM.trim)!=c("shall", "ought_to"), select=-c(shall, ought_to))## Warning in row.names(DM.trim) != c("shall", "ought_to"): longer object length is
## not a multiple of shorter object lengthIG <- graph_from_adjacency_matrix(DM.trim, weighted=TRUE, mode = "directed")
IG## IGRAPH 022044b DNW- 9 19 --
## + attr: name (v/c), weight (e/n)
## + edges from 022044b (vertex names):
## [1] can ->could can ->will could ->can may ->can
## [5] may ->could might ->can might ->could might ->should
## [9] might ->will might ->would must ->can should ->would
## [13] used_to->could will ->can will ->might will ->would
## [17] would ->could would ->should would ->willplot(IG)
Looking pretty good now, but we can clean up…
par(mar=c(.0,0,0,0))
plot.igraph(IG,
vertex.size = 24,
vertex.color="greenyellow",
vertex.shape="circle",
vertex.frame.color="greenyellow",
vertex.frame.lwd=1,
vertex.label.color="black",
vertex.label.family="sans",
vertex.label.cex=.5,
edge.color="black",
edge.width=.8,
edge.arrow.size=.5)
We could also figure out how to scale the arrows and the nodes based on magnitude.
And we could also use chord diagrams. Look cool – but harder to read, especially directions.
library(circlize)## ========================================
## circlize version 0.4.10
## CRAN page: https://cran.r-project.org/package=circlize
## Github page: https://github.com/jokergoo/circlize
## Documentation: https://jokergoo.github.io/circlize_book/book/
##
## If you use it in published research, please cite:
## Gu, Z. circlize implements and enhances circular visualization
## in R. Bioinformatics 2014.
##
## This message can be suppressed by:
## suppressPackageStartupMessages(library(circlize))
## ========================================##
## Attaching package: 'circlize'## The following object is masked from 'package:igraph':
##
## degreechordDiagram(DM )
chordDiagram(DM.trim)