valid_BCT.Rmd

---
title: 'Validation: BCT'
author: "Paschalis Agapitos"
---

```{r, echo=FALSE}
# if not installed, uncomment and run
# install.packages("stylo")
# install.packages("here")
```
Aim of this notebook is to validate the Bootstrap Consensus Tree method used in the paper *A Stylometric Analysis of Seneca's Disputed Plays: Authorship Verification of Octavia and Hercules Oetaeus*.

```{r}
library(stylo)
library(here)
```

## Setting the working directory

Set the working directory to `validation_PCA_BCT/`. This directory holds the data and the code to validate the PCA and BCT methods.

```{r}
# to set the working directory and ensure compatibility across all of the operating systems (IF THE CODE BELOW DOES NOT RAISE ANY ERROR):
# Session > Restart R > Set Working Directory > Choose Directory... > set to `seneca_stylometry/2_validation/validation_PCA_BCT/results/BCT/MFCs-4grams/`
# we choose this file in order to write the resulting plots automatically there
setwd(file.path("~", "Documents", "projects", "seneca_paper", "seneca_stylometry", "2_validation", "validation_PCA_BCT", "results", "BCT", "MFCs-4grams"))
getwd() # verify directory
```

## Bootstrap Consensus Tree - MFC 4-grams

### Load and prepare corpus for the validation of BCT
Load and prepare the correct corpus for the validation of BCT, following the same preprocessing steps.

```{r}
# load the corpus for the validation of BCT
raw_corpus_bct <- load.corpus(files = "all", 
                              corpus.dir = file.path("..", "..", "..", "..", "validation_corpora", "validation_corpus_BCT", 
                              encoding = "UTF-8")

# tokenize the corpus
tokenized_corpus_bct <- txt.to.words.ext(raw_corpus_bct, 
                                         corpus.lang = "Latin.corr", 
                                         preserve.case = FALSE)
```

#### Remove the pronouns
Remove pronouns from the corpus.

```{r}
# remove pronouns from the tokenized corpus
corpus_no_pronouns_bct <- delete.stop.words(tokenized_corpus_bct, 
                                            stop.words = stylo.pronouns(corpus.lang = "Latin.corr"))

# list of pronouns removed
stylo.pronouns(corpus.lang = "Latin.corr")
```
#### Extracting the features (character 4-grams)
Extract character 4-grams and add them to a frequency table.

```{r}
# Extract character 4-grams
corpus_char_4grams_bct <- txt.to.features(corpus_no_pronouns_bct, 
                                          features = "c", 
                                          ngram.size = 4)

# Create a frequency list of the 4-grams
frequent_features_4grams_bct <- make.frequency.list(corpus_char_4grams_bct, 
                                                    head = 5000)

# Create a table of frequencies for the 4-grams
freqs_4grams_bct <- make.table.of.frequencies(corpus_char_4grams_bct, 
                                              features = frequent_features_4grams_bct, 
                                              relative = TRUE)
```

```{r}
# BCT 4grams - top 100-2000-100 MFC 4 grams - consensus strength 0.5
bct_results_4grams <- stylo(frequencies = freqs_4grams_bct, 
                            distance.measure = "wurzburg", #cosine delta
                            analysis.type = "BCT", 
                            mfw.min = 100, mfw.max = 2000, increment = 100, 
                            custom.graph.title = "Who is the author?", 
                            write.pdf.file = TRUE, 
                            gui = TRUE) # GUI only to doublecheck. In principle no change should be implemented
```

## REPLICATION STEPS ON STYLO'S GUI (BCT validation)

Note: `stylo_config.txt` provides a detailed version of the parameters used throughout this experiment. It can be found in `seneca_stylometry/2_validation/validation_PCA_BCT/results/BCT/MFCs-4grams/`.

-----------------------------------
1) Run the script in the cell below, which will opent the Graphic User Interface (GUI) of *Stylo*
2) On the GUI select the following, if they are not already selected:

* INPUT & LANGUAGE
  - INPUT: `plain text` 
  - LANGUAGE: `Latin (u/v > u)`
* FEATURES
  - FEATURES: `chars`, `ngram size: 4`
  - MFW SETTING: `Minimum: 100`, `Maximum: 2000`, `Increment: 100`, `Start at freq. rank: 1`
  - CULLING: `Delete pronouns: YES`
* STATISTICS
  - STATISTICS: `Consensus Tree`, `Consensus strength : 0.5`
  - DELTA DISTANCE: `Cosine Delta`
* SAMPLING
  - `No sampling`
* OUTPUT
  - GRAPHS: `Onscreen`, `PDF`
  
-----------------------------------