Initialization

Load libraries and data, and set initial parameters

Load libraries:

op <- options(warn = (-1)) # suppress warnings 
library(readtext)
library(stringr)
options(op) # reset the default value

Directory and file names:

path =  file.path("/Users", "bmcgillivray", "Documents", "OneDrive", "The Alan Turing Institute", "Martina Astrid Rodda - MAR dphil project", fsep = "/")
path_wl = paste(path, "TAL paper", "ancient lexicography", sep = "/")
word_list_file_name = 'word lists for philomen.txt'

Initial parameters

Set parameters for initial exploration:

window = 1
freq_threshold = 1
lexicon = "SCHMIDT"

Word list

Read word list:

word_list_file = readtext(file = paste(path_wl, word_list_file_name, sep = "/"), encoding = "UTF-8")
word_list_text = word_list_file$text

Process word list:

word_list_text = gsub("^.+?asterisk\n", "", word_list_text)
word_list = strsplit(word_list_text, "\n")[[1]]
#word_list = strsplit(word_list, "\t")
#word_list = as.list(word_list, recursive = F)
word_list = data.frame(word_list)
word_list = word_list[!word_list$word_list == "",] # remove empty rows
word_list = gsub("*", "", word_list, fixed = T) # remove *
word_list = gsub("\t$", "", word_list, perl = T) # remove *
word_list = data.frame(word_list)
word_list = str_split(word_list$word_list, "\t", 3) # split by tab
word_list = data.frame(matrix(unlist(word_list), ncol = 3, byrow = T))
colnames(word_list) = c("ID", "lemma", "cat")
word_list2 = str_split(word_list$cat, "/", 2) # split by /
word_list2 = data.frame(matrix(unlist(word_list2), ncol = 2, byrow = T))
colnames(word_list2) = c("frequency_class", "polysemy_class")
word_list2$frequent = ifelse(word_list2$frequency_class == "F", T, F)
word_list2$polysemous = ifelse(word_list2$polysemy_class == "P", T, F)
word_list = data.frame(word_list[,c("lemma")], word_list2[,c("frequent", "polysemous")])
colnames(word_list)[1] = "lemma"
word_list
##        lemma frequent polysemous
## 1       ὄμμα     TRUE       TRUE
## 2      ἔργον     TRUE       TRUE
## 3      νόμος     TRUE       TRUE
## 4       γυνή     TRUE       TRUE
## 5     τραῦμα     TRUE       TRUE
## 6      μνήμη     TRUE       TRUE
## 7   θεραπεία     TRUE       TRUE
## 8       ἀνήρ     TRUE       TRUE
## 9     ἄνεμος     TRUE      FALSE
## 10     ἔλεος     TRUE      FALSE
## 11  ἐπιθυμία     TRUE      FALSE
## 12     ἐλπίς     TRUE      FALSE
## 13    ἡσυχία     TRUE      FALSE
## 14      εὐχή     TRUE      FALSE
## 15     ἔχθρα     TRUE      FALSE
## 16   κίνησις     TRUE      FALSE
## 17        ἴς    FALSE       TRUE
## 18     τάγμα    FALSE       TRUE
## 19      κάρα    FALSE       TRUE
## 20    ὀξύτης    FALSE       TRUE
## 21      ῥόος    FALSE       TRUE
## 22     ἄδεια    FALSE       TRUE
## 23    ῥάβδος    FALSE       TRUE
## 24    κλῆσις    FALSE       TRUE
## 25     ὀδύνη    FALSE      FALSE
## 26       ἠχή    FALSE      FALSE
## 27  βραδυτής    FALSE      FALSE
## 28     ὕπνος    FALSE      FALSE
## 29  ψυχρότης    FALSE      FALSE
## 30 οἰκειότης    FALSE      FALSE
## 31     μῖσος    FALSE      FALSE
## 32     κρήνη    FALSE      FALSE

Evaluation files

Function for creating name of evaluation file from parameters:

path_ev_fun = function(path_var, window_var, freq_threshold_var, lexicon_var) {
  print(paste("Creating name of evaluation file for", path_var, window_var, freq_threshold_var, lexicon_var, sep = " "))
  path_ev = paste(path_var, "Evaluation", "output", paste("semantic-space-w", window_var, "_t", freq_threshold_var, sep = ""), paste("Lexicon_", lexicon_var, sep = ""), sep = "/")
  evaluation_file_name = paste("summary_overlap_Lexicon_", lexicon_var, "_semantic-space_w", window_var, "_t", freq_threshold_var, "_neighbours.txt", sep = "")
  print(paste("Path_ev:", path_ev, sep = " "))
  print(paste("Evaluation_file_name:", evaluation_file_name, sep = " "))
  outlist <- list("path_ev" = path_ev, "evaluation_file_name" = evaluation_file_name)
return(outlist) 
}

Function for reading evaluation file:

read_ev_file_fun = function(path_var, window_var, freq_threshold_var, lexicon_var){
  print(paste("Reading evaluation file for ", path_var, window_var, freq_threshold_var, lexicon_var, sep = " "))
  ev_list = path_ev_fun(path_var, window_var, freq_threshold_var, lexicon_var)
  path_ev = ev_list$path_ev
  evaluation_file_name = ev_list$evaluation_file_name 
  evaluation_file = readtext(file = paste(path_ev, evaluation_file_name, sep = "/"), encoding = "UTF-8")
  evaluation_text = evaluation_file$text
  print(paste("Length of evaluation text:", length(evaluation_text[[1]])), sep = " ")
  return (evaluation_text)
}

Function for processing evaluation text:

process_evaluation_text_fun = function(path_var, window_var, freq_threshold_var, lexicon_var){
  print(paste("Processing evaluation text for ", path_var, window_var, freq_threshold_var, lexicon_var, sep = " "))
  evaluation_text = read_ev_file_fun(path_var, window_var, freq_threshold_var, lexicon_var)
  evaluation_text = gsub("Mean of precision.+$", "", evaluation_text)
  evaluation = strsplit(evaluation_text, "\n")[[1]]
  evaluation = data.frame(evaluation)
  evaluation = str_split(evaluation$evaluation, "\t", 6) # split by tab
  cnames = evaluation[[1]]
  evaluation = data.frame(matrix(unlist(evaluation), ncol = 6, byrow = T))
  colnames(evaluation) = cnames
  evaluation = evaluation[2:nrow(evaluation),]
  evaluation$precision = as.numeric(as.character(evaluation$precision))
  evaluation$recall = as.numeric(as.character(evaluation$recall)) 
  print(paste("Dimension of evaluation data frame:", dim(evaluation), sep = " "))
  return (evaluation)
}

Create evaluation dataset:

evaluation = process_evaluation_text_fun(path, window, freq_threshold, lexicon)
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t1/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w1_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 1029"
## [2] "Dimension of evaluation data frame: 6"

Combine datasets

Combine word list with evaluation dataset:

data = merge(word_list, evaluation, by = c("lemma"))

Distributions

hist(data$precision)

hist(data$recall)

shapiro.test(data$precision)
## 
##  Shapiro-Wilk normality test
## 
## data:  data$precision
## W = 0.63833, p-value = 1.172e-07
shapiro.test(data$recall)
## 
##  Shapiro-Wilk normality test
## 
## data:  data$recall
## W = 0.62816, p-value = 8.599e-08

The distributions are not normal. However, the t-test assumes that the means of the different samples are normally distributed; it does not assume that the population is normally distributed. By the central limit theorem, means of samples from a population with finite variance approach a normal distribution regardless of the distribution of the population. Rules of thumb say that the sample means are basically normally distributed as long as the sample size is at least 20 or 30. For a t-test to be valid on a sample of smaller size, the population distribution would have to be approximately normal. The t-test is invalid for small samples from non-normal distributions, but it is valid for large samples from non-normal distributions. In our case we have 32 data points.

Relationship between frequency and evaluation metrics

par(mfrow = c(1,2))
boxplot(precision ~ frequent, data = data, main = "Precision by frequency")
boxplot(recall ~ frequent, data = data, main = "Recall by frequency")

par(mfrow = c(1,1))

Frequency does not seem to make a difference to evaluation metrics.

Is there a difference between frequent and non-frequent lemmas in terms of their precision?

tapply(data$precision, data$frequent, mean)
##  FALSE   TRUE 
## 0.0375 0.0375
tapply(data$precision, data$frequent, sd)
##      FALSE       TRUE 
## 0.06191392 0.06191392
boxplot(data$precision ~ data$frequent)

t.test(precision ~ frequent, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  precision by frequent
## t = 0, df = 30, p-value = 1
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.04470509  0.04470509
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##              0.0375              0.0375

No.

Is there a difference between frequent and non-frequent lemmas in terms of their recall?

tapply(data$recall, data$frequent, mean)
##      FALSE       TRUE 
## 0.05797697 0.04195350
tapply(data$recall, data$frequent, sd)
##     FALSE      TRUE 
## 0.1051991 0.0758737
boxplot(data$recall ~ data$frequent)

t.test(recall ~ frequent, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  recall by frequent
## t = 0.49415, df = 27.282, p-value = 0.6252
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.05047807  0.08252501
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.05797697          0.04195350

No.

The high p-value indicates that we cannot reject the null hypothesis that the two groups have the same mean.

Relationship between polysemy and evaluation metrics

par(mfrow = c(1,2))
boxplot(precision ~ polysemous, data = data, main = "Precision by polysemy")
boxplot(recall ~ polysemous, data = data, main = "Recall by polysemy")

par(mfrow = c(1,1))

Polysemy, rather than frequency, seems to make a difference to evaluation metrics, with polysemous lemmas performing worse than monosemous ones.

Is there a difference between polysemous and monosemous lemmas in terms of their precision?

tapply(data$precision, data$polysemous, mean)
##   FALSE    TRUE 
## 0.04375 0.03125
tapply(data$precision, data$polysemous, sd)
##      FALSE       TRUE 
## 0.06291529 0.06020797
boxplot(data$precision ~ data$polysemous)

t.test(precision ~ polysemous, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  precision by polysemous
## t = 0.57417, df = 29.942, p-value = 0.5701
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.03196507  0.05696507
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##             0.04375             0.03125

No. The high p-value indicates that we cannot reject the null hypothesis that the two groups have the same mean.

Is there a difference between polysemous and monosemous lemmas in terms of their recall?

tapply(data$recall, data$polysemous, mean)
##      FALSE       TRUE 
## 0.06329853 0.03663194
tapply(data$recall, data$polysemous, sd)
##      FALSE       TRUE 
## 0.10637037 0.07257194
boxplot(data$recall ~ data$polysemous)

t.test(recall ~ polysemous, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  recall by polysemous
## t = 0.82836, df = 26.477, p-value = 0.4149
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.03944728  0.09278045
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.06329853          0.03663194

No. The high p-value indicates that we cannot reject the null hypothesis that the two groups have the same mean.

Regression models

ev.pr.lm <- lm(precision ~ frequent + polysemous, data = data)
summary(ev.pr.lm)
## 
## Call:
## lm(formula = precision ~ frequent + polysemous, data = data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.04375 -0.04375 -0.03125  0.05625  0.16875 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)  
## (Intercept)     4.375e-02  1.918e-02   2.281    0.030 *
## frequentTRUE   -1.472e-18  2.214e-02   0.000    1.000  
## polysemousTRUE -1.250e-02  2.214e-02  -0.565    0.577  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.06263 on 29 degrees of freedom
## Multiple R-squared:  0.01087,    Adjusted R-squared:  -0.05735 
## F-statistic: 0.1593 on 2 and 29 DF,  p-value: 0.8534

Model diagnostics:

par(mfrow = c(2,2))
plot(ev.pr.lm)
## hat values (leverages) are all = 0.09375
##  and there are no factor predictors; no plot no. 5

par(mfrow = c(1,1))

Not a good model.

ev.re.lm <- lm(recall ~ frequent + polysemous, data = data)
summary(ev.re.lm)
## 
## Call:
## lm(formula = recall ~ frequent + polysemous, data = data)
## 
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -0.071310 -0.055287 -0.044644  0.008989  0.262023 
## 
## Coefficients:
##                Estimate Std. Error t value Pr(>|t|)  
## (Intercept)     0.07131    0.02824   2.525   0.0173 *
## frequentTRUE   -0.01602    0.03261  -0.491   0.6268  
## polysemousTRUE -0.02667    0.03261  -0.818   0.4201  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.09223 on 29 degrees of freedom
## Multiple R-squared:  0.03043,    Adjusted R-squared:  -0.03643 
## F-statistic: 0.4552 on 2 and 29 DF,  p-value: 0.6388

Model diagnostics:

par(mfrow = c(2,2))
plot(ev.re.lm)
## hat values (leverages) are all = 0.09375
##  and there are no factor predictors; no plot no. 5

par(mfrow = c(1,1))

Not a good model.

All parameter combinations

Loop over all parameter combinations:

windows = c(1,5,10)
#windows = c(5,10)
freq_thresholds = c(1,20,50,100)
#freq_thresholds = c(20,50,100)
lexicons = c("SCHMIDT", "AGWN", "POLLUX")
#lexicons = c("POLLUX")

Create datasets

evaluations = data.frame(lemma=(character()),
                 precision=double(), 
                 recall=double(), 
                 window=integer(),
                 freq_threshold=integer(),
                 lexicon=factor(),
                 stringsAsFactors=FALSE) 
for (window in windows) {
  print(paste("Window:", window, sep = ""))
  for (f in freq_thresholds) {
    print(paste("Freq threshold:", f, sep = ""))
    for (lexicon in lexicons){
      print(paste("Lexicon:", lexicon, sep = ""))
      if ( !(window == 1 && f ==    1 && lexicon == "AGWN") && !(window == 5 && f ==    1   && lexicon ==   "AGWN") && !(window == 10   && f ==1 && lexicon ==  "AGWN")){
        print(paste("Continue with", "window:", window, "Freq threshold:", f, "Lexicon:", lexicon, sep = " "))
      evaluation = process_evaluation_text_fun(path, window, f, lexicon)
      evaluation = evaluation[,c("lemma", "precision", "recall")]
      evaluation$window = window
      evaluation$freq_threshold = f
      evaluation$lexicon = lexicon
      evaluations = rbind(evaluations, evaluation)}
    }
  }
}
## [1] "Window:1"
## [1] "Freq threshold:1"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 1 Freq threshold: 1 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t1/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w1_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 1029"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:AGWN"
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 1 Freq threshold: 1 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 1 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t1/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w1_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 309"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:20"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 1 Freq threshold: 20 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t20/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w1_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 701"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 1 Freq threshold: 20 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t20/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w1_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 6864"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 1 Freq threshold: 20 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 20 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t20/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w1_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 236"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:50"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 1 Freq threshold: 50 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t50/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w1_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 531"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 1 Freq threshold: 50 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t50/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w1_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 4666"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 1 Freq threshold: 50 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 50 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t50/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w1_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 177"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:100"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 1 Freq threshold: 100 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t100/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w1_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 423"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 1 Freq threshold: 100 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t100/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w1_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 3329"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 1 Freq threshold: 100 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 1 100 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w1_t100/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w1_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 146"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Window:5"
## [1] "Freq threshold:1"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 5 Freq threshold: 1 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t1/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w5_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 1046"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:AGWN"
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 5 Freq threshold: 1 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 1 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t1/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w5_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 313"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:20"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 5 Freq threshold: 20 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t20/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w5_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 701"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 5 Freq threshold: 20 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t20/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w5_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 6865"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 5 Freq threshold: 20 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 20 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t20/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w5_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 236"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:50"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 5 Freq threshold: 50 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t50/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w5_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 531"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 5 Freq threshold: 50 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t50/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w5_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 4666"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 5 Freq threshold: 50 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 50 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t50/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w5_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 177"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:100"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 5 Freq threshold: 100 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t100/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w5_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 423"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 5 Freq threshold: 100 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t100/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w5_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 3329"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 5 Freq threshold: 100 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 5 100 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w5_t100/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w5_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 146"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Window:10"
## [1] "Freq threshold:1"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 10 Freq threshold: 1 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t1/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w10_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 1046"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:AGWN"
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 10 Freq threshold: 1 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 1 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t1/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w10_t1_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 313"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:20"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 10 Freq threshold: 20 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t20/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w10_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 701"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 10 Freq threshold: 20 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t20/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w10_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 6865"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 10 Freq threshold: 20 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 20 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t20/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w10_t20_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 236"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:50"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 10 Freq threshold: 50 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t50/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w10_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 531"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 10 Freq threshold: 50 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t50/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w10_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 4666"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 10 Freq threshold: 50 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 50 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t50/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w10_t50_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 177"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Freq threshold:100"
## [1] "Lexicon:SCHMIDT"
## [1] "Continue with window: 10 Freq threshold: 100 Lexicon: SCHMIDT"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 SCHMIDT"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 SCHMIDT"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 SCHMIDT"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t100/Lexicon_SCHMIDT"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_SCHMIDT_semantic-space_w10_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 423"
## [2] "Dimension of evaluation data frame: 6"  
## [1] "Lexicon:AGWN"
## [1] "Continue with window: 10 Freq threshold: 100 Lexicon: AGWN"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 AGWN"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 AGWN"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 AGWN"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t100/Lexicon_AGWN"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_AGWN_semantic-space_w10_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 3329"
## [2] "Dimension of evaluation data frame: 6"   
## [1] "Lexicon:POLLUX"
## [1] "Continue with window: 10 Freq threshold: 100 Lexicon: POLLUX"
## [1] "Processing evaluation text for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 POLLUX"
## [1] "Reading evaluation file for  /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 POLLUX"
## [1] "Creating name of evaluation file for /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project 10 100 POLLUX"
## [1] "Path_ev: /Users/bmcgillivray/Documents/OneDrive/The Alan Turing Institute/Martina Astrid Rodda - MAR dphil project/Evaluation/output/semantic-space-w10_t100/Lexicon_POLLUX"
## [1] "Evaluation_file_name: summary_overlap_Lexicon_POLLUX_semantic-space_w10_t100_neighbours.txt"
## [1] "Length of evaluation text: 1"
## [1] "Dimension of evaluation data frame: 146"
## [2] "Dimension of evaluation data frame: 6"
evaluations$lexicon = as.factor(evaluations$lexicon)

Prepare final dataset:

data = merge(word_list, evaluations, by = c("lemma"))
data$lexicon = as.factor(data$lexicon)
dim(data)
## [1] 690   8
summary(data)
##      lemma      frequent       polysemous        precision      
##  ἔχθρα  : 33   Mode :logical   Mode :logical   Min.   :0.00000  
##  ἡσυχία : 33   FALSE:330       FALSE:360       1st Qu.:0.00000  
##  μῖσος  : 33   TRUE :360       TRUE :330       Median :0.00000  
##  ἄδεια  : 21                                   Mean   :0.05565  
##  ἄνεμος : 21                                   3rd Qu.:0.10000  
##  ἀνήρ   : 21                                   Max.   :0.40000  
##  (Other):528                                                    
##      recall            window       freq_threshold      lexicon   
##  Min.   :0.00000   Min.   : 1.000   Min.   :  1.00   AGWN   :279  
##  1st Qu.:0.00000   1st Qu.: 1.000   1st Qu.: 20.00   POLLUX : 36  
##  Median :0.00000   Median : 5.000   Median : 50.00   SCHMIDT:375  
##  Mean   :0.05318   Mean   : 5.333   Mean   : 47.07                
##  3rd Qu.:0.08333   3rd Qu.:10.000   3rd Qu.:100.00                
##  Max.   :0.50000   Max.   :10.000   Max.   :100.00                
##

Distributions

hist(data$precision)

hist(data$recall)

shapiro.test(data$precision)
## 
##  Shapiro-Wilk normality test
## 
## data:  data$precision
## W = 0.68646, p-value < 2.2e-16
shapiro.test(data$recall)
## 
##  Shapiro-Wilk normality test
## 
## data:  data$recall
## W = 0.6561, p-value < 2.2e-16

The distributions are not normal. However, The t-test assumes that the means of the different samples are normally distributed; it does not assume that the population is normally distributed. By the central limit theorem, means of samples from a population with finite variance approach a normal distribution regardless of the distribution of the population. Rules of thumb say that the sample means are basically normally distributed as long as the sample size is at least 20 or 30. For a t-test to be valid on a sample of smaller size, the population distribution would have to be approximately normal. The t-test is invalid for small samples from non-normal distributions, but it is valid for large samples from non-normal distributions. In our case we have 690 data points.

Relationship between frequency and evaluation metrics

par(mfrow = c(1,2))
boxplot(precision ~ frequent, data = data, main = "Precision by frequency")
boxplot(recall ~ frequent, data = data, main = "Recall by frequency")

par(mfrow = c(1,1))

Frequency does not seem to make a difference to evaluation metrics.

Is there a statistically significant difference between frequent and non-frequent lemmas in terms of their precision?

tapply(data$precision, data$frequent, mean)
##      FALSE       TRUE 
## 0.06181818 0.05000000
tapply(data$precision, data$frequent, sd)
##      FALSE       TRUE 
## 0.09515479 0.06881407
boxplot(data$precision ~ data$frequent)

t.test(precision ~ frequent, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  precision by frequent
## t = 1.855, df = 594.79, p-value = 0.0641
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.0006944971  0.0243308607
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.06181818          0.05000000

No.

Is there a statistically significant difference between frequent and non-frequent lemmas in terms of their recall?

tapply(data$recall, data$frequent, mean)
##      FALSE       TRUE 
## 0.05868311 0.04813696
tapply(data$recall, data$frequent, sd)
##      FALSE       TRUE 
## 0.10258151 0.07662795
boxplot(data$recall ~ data$frequent)

t.test(recall ~ frequent, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  recall by frequent
## t = 1.5191, df = 606.28, p-value = 0.1293
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.003088139  0.024180439
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.05868311          0.04813696

No.

The high p-value indicates that we cannot reject the null hypothesis that the two groups have the same mean.

Relationship between polysemy and evaluation metrics

par(mfrow = c(1,2))
boxplot(precision ~ polysemous, data = data, main = "Precision by polysemy")
boxplot(recall ~ polysemous, data = data, main = "Recall by polysemy")

par(mfrow = c(1,1))

Polysemy, rather than frequency, seems to make a difference to evaluation metrics, with polysemous lemmas performing worse than monosemous ones.

Is there a statistically significant difference between polysemous and monosemous lemmas in terms of their precision?

tapply(data$precision, data$polysemous, mean)
##      FALSE       TRUE 
## 0.06916667 0.04090909
tapply(data$precision, data$polysemous, sd)
##      FALSE       TRUE 
## 0.08652612 0.07554719
boxplot(data$precision ~ data$polysemous)

t.test(precision ~ polysemous, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  precision by polysemous
## t = 4.5785, df = 686.39, p-value = 5.561e-06
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.01613962 0.04037554
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.06916667          0.04090909

Yes. The low p-value indicates that we can reject the null hypothesis that the two groups have the same mean.

Is there a statistically significant difference between polysemous and monosemous lemmas in terms of their recall?

tapply(data$recall, data$polysemous, mean)
##      FALSE       TRUE 
## 0.06662527 0.03851405
tapply(data$recall, data$polysemous, sd)
##      FALSE       TRUE 
## 0.09638540 0.08023477
boxplot(data$recall ~ data$polysemous)

t.test(recall ~ polysemous, data = data, paired = F)
## 
##  Welch Two Sample t-test
## 
## data:  recall by polysemous
## t = 4.176, df = 681.79, p-value = 3.351e-05
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.01489414 0.04132829
## sample estimates:
## mean in group FALSE  mean in group TRUE 
##          0.06662527          0.03851405

Yes. The low p-value indicates that we can reject the null hypothesis that the two groups have the same mean.

Regression models

ev.pr.lm <- lm(precision ~ frequent + polysemous, data = data)
summary(ev.pr.lm)
## 
## Call:
## lm(formula = precision ~ frequent + polysemous, data = data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.07584 -0.06333 -0.03477  0.03667  0.35272 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     0.075839   0.005410  14.017  < 2e-16 ***
## frequentTRUE   -0.012511   0.006196  -2.019   0.0439 *  
## polysemousTRUE -0.028561   0.006196  -4.609 4.81e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.08128 on 687 degrees of freedom
## Multiple R-squared:  0.03496,    Adjusted R-squared:  0.03215 
## F-statistic: 12.44 on 2 and 687 DF,  p-value: 4.914e-06

Model diagnostics:

par(mfrow = c(2,2))
plot(ev.pr.lm)

par(mfrow = c(1,1))

Not a good model.

ev.re.lm <- lm(recall ~ frequent + polysemous, data = data)
summary(ev.re.lm)
## 
## Call:
## lm(formula = recall ~ frequent + polysemous, data = data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.07262 -0.06138 -0.03300  0.02256  0.46700 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     0.072617   0.005919  12.269  < 2e-16 ***
## frequentTRUE   -0.011234   0.006778  -1.657   0.0979 .  
## polysemousTRUE -0.028384   0.006778  -4.187 3.19e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.08892 on 687 degrees of freedom
## Multiple R-squared:  0.02823,    Adjusted R-squared:  0.0254 
## F-statistic: 9.978 on 2 and 687 DF,  p-value: 5.349e-05
par(mfrow = c(2,2))
plot(ev.re.lm)

par(mfrow = c(1,1))

Not a good model.

ev.pr.lm2 <- lm(precision ~ frequent + polysemous + lexicon + window + freq_threshold, data = data)
summary(ev.pr.lm2)
## 
## Call:
## lm(formula = precision ~ frequent + polysemous + lexicon + window + 
##     freq_threshold, data = data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.16197 -0.05045 -0.03461  0.04837  0.35301 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     6.927e-02  8.919e-03   7.767 2.95e-14 ***
## frequentTRUE   -1.558e-02  5.953e-03  -2.616  0.00909 ** 
## polysemousTRUE -1.756e-02  6.091e-03  -2.883  0.00406 ** 
## lexiconPOLLUX   1.102e-01  1.420e-02   7.763 3.03e-14 ***
## lexiconSCHMIDT -6.969e-04  6.267e-03  -0.111  0.91150    
## window         -3.849e-04  8.052e-04  -0.478  0.63281    
## freq_threshold -8.705e-06  8.439e-05  -0.103  0.91787    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.07787 on 683 degrees of freedom
## Multiple R-squared:  0.1193, Adjusted R-squared:  0.1115 
## F-statistic: 15.42 on 6 and 683 DF,  p-value: < 2.2e-16
par(mfrow = c(2,2))
plot(ev.pr.lm2)

par(mfrow = c(1,1))

Not a good model.

Stepwise regression:

ev.pr.m0 <- lm(precision ~ 1, data = data)
ev.pr.lm.step <- step(ev.pr.m0, scope = ~ frequent + polysemous + lexicon + window + recall + freq_threshold)#, direction = "forward")
## Start:  AIC=-3440.07
## precision ~ 1
## 
##                  Df Sum of Sq    RSS     AIC
## + recall          1   2.49475 2.2082 -3959.7
## + lexicon         2   0.46813 4.2348 -3508.4
## + polysemous      1   0.13748 4.5655 -3458.5
## + frequent        1   0.02405 4.6789 -3441.6
## <none>                        4.7030 -3440.1
## + window          1   0.00139 4.7016 -3438.3
## + freq_threshold  1   0.00135 4.7016 -3438.3
## 
## Step:  AIC=-3959.71
## precision ~ recall
## 
##                  Df Sum of Sq    RSS     AIC
## + lexicon         2   0.41593 1.7923 -4099.7
## + polysemous      1   0.01585 2.1924 -3962.7
## <none>                        2.2082 -3959.7
## + frequent        1   0.00393 2.2043 -3958.9
## + freq_threshold  1   0.00122 2.2070 -3958.1
## + window          1   0.00050 2.2077 -3957.9
## - recall          1   2.49475 4.7030 -3440.1
## 
## Step:  AIC=-4099.71
## precision ~ recall + lexicon
## 
##                  Df Sum of Sq    RSS     AIC
## + frequent        1   0.01010 1.7822 -4101.6
## <none>                        1.7923 -4099.7
## + window          1   0.00055 1.7917 -4097.9
## + freq_threshold  1   0.00014 1.7921 -4097.8
## + polysemous      1   0.00001 1.7923 -4097.7
## - lexicon         2   0.41593 2.2082 -3959.7
## - recall          1   2.44255 4.2348 -3508.4
## 
## Step:  AIC=-4101.61
## precision ~ recall + lexicon + frequent
## 
##                  Df Sum of Sq    RSS     AIC
## <none>                        1.7822 -4101.6
## + window          1   0.00053 1.7816 -4099.8
## - frequent        1   0.01010 1.7923 -4099.7
## + freq_threshold  1   0.00006 1.7821 -4099.6
## + polysemous      1   0.00000 1.7822 -4099.6
## - lexicon         2   0.42210 2.2043 -3958.9
## - recall          1   2.41178 4.1940 -3513.1
summary(ev.pr.lm.step)
## 
## Call:
## lm(formula = precision ~ recall + lexicon + frequent, data = data)
## 
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -0.145057 -0.026309 -0.007445  0.004614  0.230953 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     0.034002   0.003739   9.094  < 2e-16 ***
## recall          0.675224   0.022177  30.447  < 2e-16 ***
## lexiconPOLLUX   0.080277   0.009142   8.781  < 2e-16 ***
## lexiconSCHMIDT -0.026557   0.004122  -6.442 2.21e-10 ***
## frequentTRUE   -0.007693   0.003905  -1.970   0.0492 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.05101 on 685 degrees of freedom
## Multiple R-squared:  0.6211, Adjusted R-squared:  0.6188 
## F-statistic: 280.7 on 4 and 685 DF,  p-value: < 2.2e-16

The higher the recall, the higher the precision, as expected. More interestingly, Schmidt’s lexicon is associated to lower precision compared to the baseline, AGWN.

Model diagnostics:

par(mfrow = c(2,2))
plot(ev.pr.lm.step)

par(mfrow = c(1,1))

Not a great model.

ev.re.m0 <- lm(recall ~ 1, data = data)
ev.re.lm.step <- step(ev.re.m0, 
                      scope = ~ frequent + polysemous + lexicon + window + precision)
## Start:  AIC=-3320.94
## recall ~ 1
## 
##              Df Sum of Sq    RSS     AIC
## + precision   1   2.96486 2.6243 -3840.6
## + lexicon     2   0.27679 5.3124 -3352.0
## + polysemous  1   0.13606 5.4531 -3335.9
## + frequent    1   0.01915 5.5700 -3321.3
## <none>                    5.5892 -3320.9
## + window      1   0.00798 5.5812 -3319.9
## 
## Step:  AIC=-3840.59
## recall ~ precision
## 
##              Df Sum of Sq    RSS     AIC
## + lexicon     2   0.37600 2.2483 -3943.3
## <none>                    2.6243 -3840.6
## + polysemous  1   0.00571 2.6186 -3840.1
## + window      1   0.00357 2.6208 -3839.5
## + frequent    1   0.00023 2.6241 -3838.6
## - precision   1   2.96486 5.5892 -3320.9
## 
## Step:  AIC=-3943.29
## recall ~ precision + lexicon
## 
##              Df Sum of Sq    RSS     AIC
## + polysemous  1   0.02066 2.2277 -3947.7
## <none>                    2.2483 -3943.3
## + window      1   0.00332 2.2450 -3942.3
## + frequent    1   0.00048 2.2479 -3941.4
## - lexicon     2   0.37600 2.6243 -3840.6
## - precision   1   3.06407 5.3124 -3352.0
## 
## Step:  AIC=-3947.66
## recall ~ precision + lexicon + polysemous
## 
##              Df Sum of Sq    RSS     AIC
## <none>                    2.2277 -3947.7
## + window      1   0.00336 2.2243 -3946.7
## + frequent    1   0.00036 2.2273 -3945.8
## - polysemous  1   0.02066 2.2483 -3943.3
## - lexicon     2   0.39095 2.6186 -3840.1
## - precision   1   2.97423 5.2019 -3364.5
                      #direction = "forward"
summary(ev.re.lm.step)
## 
## Call:
## lm(formula = recall ~ precision + lexicon + polysemous, data = data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.15445 -0.03198 -0.01482  0.01828  0.31073 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)    -0.006976   0.004405  -1.584   0.1137    
## precision       0.843005   0.027876  30.242  < 2e-16 ***
## lexiconPOLLUX  -0.048343   0.010777  -4.486 8.52e-06 ***
## lexiconSCHMIDT  0.038958   0.004509   8.641  < 2e-16 ***
## polysemousTRUE -0.011308   0.004486  -2.521   0.0119 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.05703 on 685 degrees of freedom
## Multiple R-squared:  0.6014, Adjusted R-squared:  0.5991 
## F-statistic: 258.4 on 4 and 685 DF,  p-value: < 2.2e-16
png(paste(path, "Evaluation", "plots", "model_diagnostics.png", sep = "/"))
par(mfrow = c(2,2))
plot(ev.re.lm.step)
par(mfrow = c(1,1))
dev.off()
## quartz_off_screen 
##                 2
hist(resid(ev.re.lm.step))

plot(ev.re.lm.step, which = 1)

The histogram of standardised residuals shows that the assumption of normality is valid. The fitted values and residuals plot shows that the assumption of homoscedasticity is valid, as there is no pattern in the scatterplot and the width of the scatterplot as predicted values increase is roughly the same.

The higher the precision, the higher the recall, as expected. More interestingly, Schmidt’s lexicon is associated to higher recall compared to the baseline, AGWN.

Check https://data.library.virginia.edu/diagnostic-plots/.