gpwrajeg_2021_paper-r-notebook.Rmd

---
title: R Notebook for *A corpus-based study of constructional equivalence for the
  Indonesian translation of ROB and STEAL based on the __OpenSubtitle__ English-Indonesian Parallel Corpus*
author: '[Gede Primahadi Wijaya Rajeg](https://udayananetworking.unud.ac.id/lecturer/880-gede-primahadi-wijaya-rajeg) <a itemprop="sameAs" content="https://orcid.org/0000-0002-2047-8621" href="https://orcid.org/0000-0002-2047-8621" target="orcid.widget" rel="noopener noreferrer" style="vertical-align:top;"><img src="https://orcid.org/sites/default/files/images/orcid_16x16.png" style="width:1em;margin-right:.5em;" alt="ORCID iD icon"></a>'
output:
  word_document:
    toc: yes
  html_notebook:
    fig_caption: yes
    fig_width: 6
    fig_retina: 2
    toc: yes
    toc_float: yes
---

```{r setup, message = FALSE, warning = FALSE, include = FALSE}
# global option chunck
knitr::opts_chunk$set(fig.width = 7, 
                      fig.asp = 0.618,
                      fig.retina = 2,
                      dpi = 300,
                      dev = "png",
                      echo = TRUE,
                      message = FALSE,
                      warning = FALSE)
library(tidyverse)
library(readxl)
library(RColorBrewer)
library(vcd)
```


The R codes in this notebook use the following packages: tidyverse, readxl, RColorBrewer, and vcd.

If you use, or are inspired by, the codes, the data, and analyses from this study, you can cite this R Notebook and the paper as well:

- citation for the R Notebook and related materials in this paper:

  > Rajeg, G. P. W. (2021a). Constructional equivalence in the Indonesian translations of ROB and STEAL (Version 0.0.1). https://doi.org/10.17605/OSF.IO/PZC8Y

- citation for the paper (to be updated when published):

```{r load-main-ROB-and-STEAL-data, message = FALSE, warning = FALSE, include = FALSE}
rob <- readr::read_tsv("ROB_sample_conc_main.tsv")
steal <- readr::read_tsv("STEAL_sample_conc_main.tsv")
```

### code preparation for data of ROB

```{r profiled-participants-rob}
rob_eng_profiled_participants <- rob %>% 
  # verbal construction of ROB
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", negate = TRUE)) %>%
  count(ROLE_ENG_THIEF) %>% 
  rename(is_mentioned = ROLE_ENG_THIEF) %>% 
  mutate(role = "thief") %>% 
  bind_rows(rob %>% 
              # verbal construction of ROB
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", 
                                                 negate = TRUE)) %>%
              mutate(ROLE_ENG_TARGET = as.logical(ROLE_ENG_TARGET)) %>% 
              count(ROLE_ENG_TARGET) %>% 
              rename(is_mentioned = ROLE_ENG_TARGET) %>% 
              mutate(role = "target")) %>% 
  bind_rows(rob %>% 
              # verbal construction of ROB
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_GOODS = as.logical(ROLE_ENG_GOODS)) %>% 
              count(ROLE_ENG_GOODS) %>% 
              rename(is_mentioned = ROLE_ENG_GOODS) %>% 
              mutate(role = "goods")) %>% 
  mutate(role = factor(role, levels = c("thief", "target", "goods"))) %>% 
  group_by(role) %>% 
  arrange(role, desc(is_mentioned)) %>% 
  mutate(perc_mention = round(n/sum(n) * 100, 2)) %>% 
  ungroup()

# binomial test of the profiled participant roles for ROB
rob_eng_profiled_participants_binom <- rob_eng_profiled_participants %>% 
  pivot_wider(-perc_mention, 
              names_from = is_mentioned, 
              values_from = n, 
              values_fill = 0L) %>% 
  mutate(n = `TRUE`+`FALSE`, 
         pbin = map_dbl(pmap(list(x = `TRUE`, n = n), binom.test), "p.value"), 
         pholm = p.adjust(pbin, method = "holm"),
         signif = "ns", 
         signif = if_else(pholm < 0.05, "*", signif), 
         signif = if_else(pholm < 0.01, "**", signif), 
         signif = if_else(pholm < 0.001, "***", signif),
         lemma = "ROB")
```

```{r profiled-participants-rob-core-argument}
rob_eng_profiled_participants_coreargs <- rob %>% 
  # verbal construction of ROB
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", negate = TRUE)) %>% 
  count(ROLE_ENG_THIEF, SYN_ENG_THIEF) %>% 
  rename(is_mentioned = ROLE_ENG_THIEF,
         syntax = SYN_ENG_THIEF) %>% 
  mutate(role = "thief") %>% 
  bind_rows(rob %>% 
              # verbal construction of ROB
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_TARGET = as.logical(ROLE_ENG_TARGET)) %>% 
              count(ROLE_ENG_TARGET, SYN_ENG_TARGET) %>% 
              rename(is_mentioned = ROLE_ENG_TARGET,
                     syntax = SYN_ENG_TARGET) %>% 
              mutate(role = "target")) %>% 
  bind_rows(rob %>% 
              # verbal construction of ROB
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_GOODS = as.logical(ROLE_ENG_GOODS)) %>% 
              count(ROLE_ENG_GOODS, SYN_ENG_GOODS) %>% 
              rename(is_mentioned = ROLE_ENG_GOODS,
                     syntax = SYN_ENG_GOODS) %>% 
              mutate(role = "goods")) %>% 
  mutate(role = factor(role, levels = c("thief", "target", "goods")))
rob_eng_profiled_participants_coreargs <- rob_eng_profiled_participants_coreargs %>% 
  mutate(core_args = TRUE,
         core_args = if_else(is.na(syntax), NA, core_args),
         core_args = if_else(str_detect(syntax, "passive_AGENT|possessive_determiner|PP_complement"), 
                             FALSE, core_args)) %>% 
  group_by(role, core_args) %>% 
  mutate(n_coreargs_by_role = sum(n)) %>% 
  group_by(role) %>% 
  arrange(role, desc(n_coreargs_by_role))

rob_eng_profiled_participants_coreargs_distinct <- rob_eng_profiled_participants_coreargs %>% 
  select(role, core_args, n_coreargs_by_role) %>% 
  distinct() %>% 
  filter(!is.na(core_args)) %>% 
  arrange(role, desc(core_args)) %>% 
  mutate(perc_coreargs_by_role = round(n_coreargs_by_role/sum(n_coreargs_by_role) * 100, digits = 1))

# binomial test of the core profiled participant roles for ROB
rob_eng_profiled_participants_coreargs_binom <- rob_eng_profiled_participants_coreargs_distinct %>% 
  pivot_wider(-perc_coreargs_by_role, names_from = core_args, values_from = n_coreargs_by_role, values_fill = 0L) %>% 
  mutate(n = `TRUE`+`FALSE`, 
         pbin = map_dbl(pmap(list(x = `TRUE`, n = n), binom.test), "p.value"), 
         pholm = p.adjust(pbin, method = "holm"),
         signif = "ns", 
         signif = if_else(pholm < 0.05, "*", signif), 
         signif = if_else(pholm < 0.01, "**", signif), 
         signif = if_else(pholm < 0.001, "***", signif),
         lemma = "ROB",
         perc_coreargs = round(`TRUE`/n * 100, 1),
         perc_non_coreargs = round(`FALSE`/n * 100, 1))
```

### code preparation for data of STEAL

```{r profiled-participants-steal}
steal_eng_profiled_participants <- steal %>% 
  # verbal construction of STEAL
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", negate = TRUE)) %>%
  mutate(ROLE_ENG_THIEF = as.logical(ROLE_ENG_THIEF)) %>% 
  count(ROLE_ENG_THIEF) %>% 
  rename(is_mentioned = ROLE_ENG_THIEF) %>% 
  mutate(role = "thief") %>% 
  bind_rows(steal %>% 
              # verbal construction of STEAL
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_TARGET = as.logical(ROLE_ENG_TARGET)) %>% 
              count(ROLE_ENG_TARGET) %>% 
              rename(is_mentioned = ROLE_ENG_TARGET) %>% 
              mutate(role = "target")) %>% 
  bind_rows(steal %>% 
              # verbal construction of STEAL
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", 
                                                 negate = TRUE)) %>% 
              filter(!is.na(ROLE_ENG_GOODS)) %>% 
              mutate(ROLE_ENG_GOODS = as.logical(ROLE_ENG_GOODS)) %>% 
              count(ROLE_ENG_GOODS) %>% 
              rename(is_mentioned = ROLE_ENG_GOODS) %>% 
              mutate(role = "goods")) %>% 
  mutate(role = factor(role, levels = c("thief", "target", "goods"))) %>% 
  group_by(role) %>% 
  arrange(role, desc(is_mentioned)) %>% 
  mutate(perc_mention = round(n/sum(n) * 100, 2)) %>% 
  ungroup()

# binomial test of the profiled participant roles for STEAL
steal_eng_profiled_participants_binom <- steal_eng_profiled_participants %>% 
  pivot_wider(-perc_mention, names_from = is_mentioned, values_from = n, values_fill = 0L) %>% 
  mutate(n = `TRUE`+`FALSE`, 
         pbin = map_dbl(pmap(list(x = `TRUE`, n = n), binom.test), "p.value"), 
         pholm = p.adjust(pbin, method = "holm"),
         signif = "ns", 
         signif = if_else(pholm < 0.05, "*", signif), 
         signif = if_else(pholm < 0.01, "**", signif), 
         signif = if_else(pholm < 0.001, "***", signif),
         lemma = "STEAL")
```

```{r profiled-participants-steal-core-argument}
steal_eng_profiled_participants_coreargs <- steal %>% 
  # verbal construction of STEAL
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", negate = TRUE)) %>%
  count(ROLE_ENG_THIEF, SYN_ENG_THIEF) %>% 
  rename(is_mentioned = ROLE_ENG_THIEF,
         syntax = SYN_ENG_THIEF) %>% 
  mutate(role = "thief",
         is_mentioned = as.logical(is_mentioned)) %>% 
  bind_rows(steal %>% 
              # verbal construction of STEAL
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_TARGET = as.logical(ROLE_ENG_TARGET)) %>% 
              count(ROLE_ENG_TARGET, SYN_ENG_TARGET) %>% 
              rename(is_mentioned = ROLE_ENG_TARGET,
                     syntax = SYN_ENG_TARGET) %>% 
              mutate(role = "target")) %>% 
  bind_rows(steal %>% 
              # verbal construction of STEAL
              filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", 
                                                 negate = TRUE)) %>% 
              mutate(ROLE_ENG_GOODS = as.logical(ROLE_ENG_GOODS)) %>% 
              count(ROLE_ENG_GOODS, SYN_ENG_GOODS) %>% 
              rename(is_mentioned = ROLE_ENG_GOODS,
                     syntax = SYN_ENG_GOODS) %>% 
              mutate(role = "goods")) %>% 
  mutate(role = factor(role, levels = c("thief", "target", "goods")))
steal_eng_profiled_participants_coreargs <- steal_eng_profiled_participants_coreargs %>% 
  mutate(core_args = TRUE,
         core_args = if_else(is.na(syntax), NA, core_args),
         core_args = if_else(str_detect(syntax, "passive_AGENT|possessive|(?i)PP_complement"), 
                             FALSE, core_args)) %>% 
  group_by(role, core_args) %>% 
  mutate(n_coreargs_by_role = sum(n)) %>% 
  group_by(role) %>% 
  arrange(role, desc(n_coreargs_by_role))

steal_eng_profiled_participants_coreargs_distinct <- steal_eng_profiled_participants_coreargs %>% 
  select(role, core_args, n_coreargs_by_role) %>% 
  distinct() %>% 
  filter(!is.na(core_args)) %>% 
  arrange(role, desc(core_args)) %>% 
  mutate(perc_coreargs_by_role = round(n_coreargs_by_role/sum(n_coreargs_by_role) * 100, digits = 1))


# binomial test of the core profiled participant roles for STEAL
steal_eng_profiled_participants_coreargs_binom <- steal_eng_profiled_participants_coreargs_distinct %>% 
  pivot_wider(-perc_coreargs_by_role, names_from = core_args, values_from = n_coreargs_by_role, values_fill = 0L) %>% 
  mutate(n = `TRUE`+`FALSE`, 
         pbin = map_dbl(pmap(list(x = `TRUE`, n = n), binom.test), "p.value"),
         pholm = p.adjust(pbin, method = "holm"),
         signif = "ns", 
         signif = if_else(pholm < 0.05, "*", signif), 
         signif = if_else(pholm < 0.01, "**", signif), 
         signif = if_else(pholm < 0.001, "***", signif),
         lemma = "STEAL",
         perc_coreargs = round(`TRUE`/n * 100, 1),
         perc_non_coreargs = round(`FALSE`/n * 100, 1))
```


### Analysis for the lexical equivalence for ROB and STEAL

```{r lexical-equivalence-database, include = FALSE, message = FALSE}
rob_equivalence_df <- read_tsv("https://raw.githubusercontent.com/gederajeg/rob-steal-parallel-corpora/main/SENASTEK/SENASTEK_data_rob1.txt")
steal_equivalence_df <- read_tsv("https://raw.githubusercontent.com/gederajeg/rob-steal-parallel-corpora/main/SENASTEK/SENASTEK_data_steal1.txt")
```


This study is based on data in Rajeg et al. [-@rajeg_pemanfaatan_2021]. [Table \@ref(tab:lex-equivalence-rob)](#lex-equivalence-rob) shows the lexical equivalence in Indonesian for word forms of the lemma ROB and [Table \@ref(tab:lex-equivalence-steal)](#lex-equivalence-steal) shows the same data for the lemma STEAL.

```{r lexical-equivalence-table}
rob_lex_equiv <- rename(summarise(arrange(group_by(count(rob_equivalence_df, NODE, TRANSLATION_NODE), NODE), desc(n), desc(TRANSLATION_NODE)), `Indonesian translation` = paste(paste("*", TRANSLATION_NODE, "* (", n, ")", sep = ""), collapse = "; ")), `English form` = NODE)

steal_lex_equiv <- rename(summarise(arrange(group_by(count(steal_equivalence_df, NODE, TRANSLATION_NODE), NODE), desc(n), desc(TRANSLATION_NODE)), `Indonesian translation` = paste(paste("*", TRANSLATION_NODE, "* (", n, ")", sep = ""), collapse = "; ")), `English form` = NODE)
```

```{r lex-equivalence-rob}
knitr::kable(rob_lex_equiv, caption = "Indonesian lexical equivalence for word forms of the lemma ROB")
```


```{r lex-equivalence-steal}
knitr::kable(steal_lex_equiv, caption = "Indonesian lexical equivalence for word forms of the lemma STEAL")
```


### Analysis for profiled participant roles in English only

```{r profiled-participants-ENG-all-chisq}
eng_profiled_participants <- bind_rows(rob_eng_profiled_participants_binom,
                                       steal_eng_profiled_participants_binom) %>% 
  mutate(perc_mention = round((`TRUE`/n) * 100, 2))

eng_profiled_participants_matrix <- eng_profiled_participants %>% 
  select(role, lemma, `TRUE`) %>% 
  pivot_wider(names_from = lemma, values_from = `TRUE`) %>% 
  data.frame(row.names = 1) %>% 
  as.matrix()
names(attr(eng_profiled_participants_matrix, "dimnames")) <- c("Participant roles", "Lemmas")

eng_profiled_participants_chisq <- chisq.test(eng_profiled_participants_matrix)
eng_profiled_participants_effsizes <- vcd::assocstats(eng_profiled_participants_matrix)
eng_profiled_participants_cramer <- round(eng_profiled_participants_effsizes$cramer, 3)
```

```{r profiled-participants-ENG-all-chisq-core-args}
eng_profiled_participants_coreargs <- bind_rows(rob_eng_profiled_participants_coreargs_binom,
                                                steal_eng_profiled_participants_coreargs_binom)

eng_profiled_participants_matrix_coreargs <- eng_profiled_participants_coreargs %>% 
  select(role, lemma, `TRUE`) %>% 
  pivot_wider(names_from = lemma, values_from = `TRUE`) %>% 
  data.frame(row.names = 1) %>% 
  as.matrix()
names(attr(eng_profiled_participants_matrix_coreargs, "dimnames")) <- c("Participant roles", "Lemmas")

eng_profiled_participants_chisq_coreargs <- chisq.test(eng_profiled_participants_matrix_coreargs)
eng_profiled_participants_effsizes_coreargs <- vcd::assocstats(eng_profiled_participants_matrix_coreargs)
eng_profiled_participants_cramer_coreargs <- round(eng_profiled_participants_effsizes_coreargs$cramer, 2)
```


```{r plot-profiled-participants-ENG-all-idn-lang-OLD, fig.cap = "Distribusi peran partisipan verba ROB dan STEAL", fig.asp = 0.718, eval = FALSE, echo = FALSE}
p1_idn <- eng_profiled_participants %>% 
  ggplot(aes(x = role, y = perc_mention, fill = lemma)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste("N=", `TRUE`, sep = "")),
            position = position_dodge(.9),
            vjust = -.5) +
  geom_text(aes(label = paste(perc_mention, "%", sep = "")),
            position = position_dodge(.9),
            vjust = 1.5,
            col = c(rep("black", 3), rep("white", 3))) +
  # group_by(role) %>% 
  # mutate(perc = round(`TRUE`/sum(`TRUE`) * 100, 2)) %>% 
  # ggplot(aes(x = role, y = perc, fill = lemma)) +
  # geom_col() +
  # geom_text(aes(label = `TRUE`),
  #           position = position_stack(.5),
  #           col = c(rep("white", 3), rep("black", 3))) +
  theme_bw() +
  labs(y = "%\n(of explicit mention)",
       x = "Participant roles",
       fill = "Lemma") +
  theme(axis.text = element_text(size = 13),
        axis.title = element_text(size = 15),
        legend.text = element_text(size = 13),
        legend.title = element_text(size = 15)) +
  scale_fill_manual(values = hcl.colors(palette = "RdYlBu", n = 5)[c(3, 5)]) +
  scale_x_discrete(labels = c("THIEF", "TARGET", "GOODS"))

p1_idn
# ggsave("figs/plot-profiled-participants-eng.png", plot = p1_idn, height = 4.5, width = 7, dpi = 300)
```

```{r plot-profiled-participants-ENG-all-idn-lang-OLD-combined-core-non-core, fig.cap = "Distribusi peran partisipan verba ROB dan STEAL", fig.asp = 0.718, eval = FALSE, echo = FALSE}
p1_idn <- eng_profiled_participants %>% 
  mutate(role = toupper(role),
         role = factor(role, levels = c("THIEF", "TARGET", "GOODS"))) %>% 
  ggplot(aes(x = lemma, y = perc_mention, fill = role)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste("N=", `TRUE`, sep = "")),
            position = position_dodge(.9),
            vjust = -.5) +
  geom_text(aes(label = paste(perc_mention, "%", sep = "")),
            position = position_dodge(.9),
            vjust = 1.5,
            col = "white") +
  theme_light() +
  labs(y = "%\n(of explicit mention)",
       x = "Lemma",
       fill = "Participant roles") +
  theme(axis.text = element_text(size = 13),
        axis.title = element_text(size = 15),
        legend.text = element_text(size = 13),
        legend.title = element_text(size = 15))

p1_idn
# ggsave("figs/plot-profiled-participants-eng.png", plot = p1_idn, height = 4.5, width = 7, dpi = 300)
```

```{r plot-profiled-participants-ENG-all-core-args, fig.cap = "Distribusi peran partisipan verba ROB dan STEAL", fig.asp = 0.718}
p1_idn <- eng_profiled_participants_coreargs %>% 
  mutate(role = toupper(role),
         role = factor(role, levels = c("THIEF", "TARGET", "GOODS"))) %>% 
  ggplot(aes(x = lemma, y = perc_coreargs, fill = role)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste("N=", `TRUE`, sep = "")),
            position = position_dodge(.9),
            vjust = if_else(eng_profiled_participants_coreargs$`TRUE` > 1, -.5, -2)) +
  geom_text(aes(label = paste(perc_coreargs, "%", sep = "")),
            position = position_dodge(.9),
            vjust = if_else(eng_profiled_participants_coreargs$`TRUE` > 1, 1.5, -.5),
            col = if_else(eng_profiled_participants_coreargs$`TRUE` > 1, "white", "gray50")) +
  theme_light() +
  labs(y = "%\n(in core syntactic arguments)",
       x = "Lemma",
       fill = "Participant roles") +
  theme(axis.text = element_text(size = 13),
        axis.title = element_text(size = 15),
        legend.text = element_text(size = 13),
        legend.title = element_text(size = 15))

p1_idn
# ggsave("figs/plot-profiled-participants-eng.png", plot = p1_idn, height = 4.5, width = 7, dpi = 300)
```


From the `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "thief") %>% pull(n)` cases of the encoding of THIEF of ROB, `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "thief") %>% pull(perc_coreargs)`% (N=`r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "thief") %>% .[[2]]`) is mapped onto the core function of grammatical subject, which can be in the active, main clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[1,3]]`), relative clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[2,3]]`), and subordinate clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[3,3]]`). The remaining one token shows the mention of the THIEF in a prepositional phrase of passive agent (@rob_thief_passive_agent).

(@rob_thief_passive_agent) ... and Hunt getting robbed right next door and by the same guy who rescued us...

As for STEAL, all occurrences of the THIEF role are profiled as the grammatical subject in the active, main clause (N=`r steal_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[1,3]]`), relative clause (N=`r steal_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[2,3]]`), and subordinate clause (N=`r steal_eng_profiled_participants_coreargs %>% filter(role == "thief", is_mentioned, core_args) %>% .[[3,3]]`).



From the `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "target") %>% pull(n)` cases of the encoding of TARGET of ROB, `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "target") %>% pull(perc_coreargs)`% (N=`r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "target") %>% .[[2]]`) is mapped onto the core functions of (i) direct object of ROB in the active, main clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "target", is_mentioned, core_args, syntax == "direct_object") %>% pull(n)`) and (ii) passive subject of ROB in the passive, main clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "target", is_mentioned, core_args, syntax == "passive_subject") %>% pull(n)`). The remaining `r rob_eng_profiled_participants_coreargs %>% filter(role == "target", is_mentioned, !core_args) %>% pull(n) %>% sum` tokens map the TARGET onto a preposition phrase (@rob_target_pp_complement) or possessive determiner in a noun phrase (@rob_target_pp_complement).

(@rob_target_pp_complement) ...that three elderly gentleman robbed 2.3 million dollar from the Williamsburg Savings Bank.
(@rob_target_possessive) We rob her advantage.

From the `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "goods") %>% pull(n)` cases of the encoding of GOODS of ROB, only `r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "goods") %>% pull(perc_coreargs)`% (N=`r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "goods") %>% .[[2]]`) is mapped onto the core functions of the direct object of ROB in the active, main clause (N=`r rob_eng_profiled_participants_coreargs %>% filter(role == "goods", is_mentioned, core_args, syntax == "direct_object") %>% pull(n)`) (see examples @rob_target_pp_complement and @rob_target_possessive). The predominant cases (`r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "goods") %>% pull(perc_non_coreargs)`%; N=`r rob_eng_profiled_participants_coreargs_binom %>% filter(role == "goods") %>% .[[3]]`) map the GOODS of ROB onto an oblique preposition phrase, as suggested in the literature (Goldberg 1995).



```{r plot-profiled-participants-ENG-all-assocplot-notebook, fig.asp = 0.818, fig.cap = "Association plot for the profiled participants of ROB and STEAL", eval = FALSE, echo = FALSE}
# png("figs/plot-assocplot-profiled-roles-english.png", width = 7, height = 6, units = "in", res = 600)
vcd::assoc(eng_profiled_participants_matrix, shade = TRUE, legend = legend_resbased(fontsize = 10), gp_labels = gpar(fontsize = 9), labeling_args = list(rot_labels = 0, set_varnames = c(Lemmas = "Lemma", `Participant roles` = "Participants"), set_labels = list(`Participant roles` = c("THIEF", "TARGET", "GOODS"))))
# dev.off()
```

```{r plot-profiled-participants-ENG-all-assocplot-notebook-coreargs, fig.asp = 0.818, fig.cap = "Association plot for the profiled participants of ROB and STEAL"}
# png("figs/plot-assocplot-profiled-roles-english.png", width = 7, height = 6, units = "in", res = 600)
vcd::assoc(eng_profiled_participants_matrix_coreargs, shade = TRUE, legend = legend_resbased(fontsize = 10), gp_labels = gpar(fontsize = 9), labeling_args = list(rot_labels = 0, set_varnames = c(Lemmas = "Lemma", `Participant roles` = "Participants"), set_labels = list(`Participant roles` = c("THIEF", "TARGET", "GOODS"))))
# dev.off()
```



```{r data-preparation}
# only take the transitive verbal and gerundial usages of ROB and STEAL
steal1 <- steal %>% 
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL|NP|_TRANS)$", negate = TRUE)) %>% # verbal construction of ROB
  mutate(across(matches("^ROLE_(ENG|IDN)_"), as.logical))

rob1 <- rob %>% 
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "(INTR|NOMINAL)$", negate = TRUE)) %>% # verbal construction of ROB
  mutate(across(matches("^ROLE_(ENG|IDN)_"), as.logical),
         SYN_IDN_TARGET = replace(SYN_IDN_TARGET, ROLE_IDN_TARGET == FALSE & SYN_IDN_TARGET == "direct_object", NA))
```



```{r cross-tabulation-and-signif-test-THIEF}
rob_thief_profile_eng_idn <- rob1 %>% 
  mutate(core_args = if_else(SYN_ENG_THIEF == "passive_AGENT", "FALSE", "TRUE"),
         core_args = if_else(is.na(SYN_ENG_THIEF), "NA", core_args)) %>% 
  filter(core_args != "FALSE") %>% # select only the core argument of THIEF in English
  count(ROLE_ENG_THIEF, ROLE_IDN_THIEF) %>% 
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "THIEF_not_mentioned", "THIEF_mentioned"))) %>% 
  rename(English = ROLE_ENG_THIEF,
         Indonesian = ROLE_IDN_THIEF) %>% 
  arrange(desc(n))
rob_thief_profile_eng_idn_matrix <- rob_thief_profile_eng_idn %>% 
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>% 
  column_to_rownames("English") %>% 
  as.matrix()
names(attr(rob_thief_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
rob_thief_profile_eng_idn_fye <- fisher.test(rob_thief_profile_eng_idn_matrix)
rob_thief_profile_eng_idn_effsize <- vcd::assocstats(rob_thief_profile_eng_idn_matrix)
rob_thief_profile_eng_idn_phicoeff <- round(rob_thief_profile_eng_idn_effsize$phi, 2)

steal_thief_profile_eng_idn <- steal1 %>% 
  count(ROLE_ENG_THIEF, ROLE_IDN_THIEF) %>% 
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "THIEF_not_mentioned", "THIEF_mentioned"))) %>% 
  rename(English = ROLE_ENG_THIEF,
         Indonesian = ROLE_IDN_THIEF) %>% 
  arrange(desc(n))
steal_thief_profile_eng_idn_matrix <- steal_thief_profile_eng_idn %>% 
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>% 
  column_to_rownames("English") %>% 
  as.matrix()
names(attr(steal_thief_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
steal_thief_profile_eng_idn_fye <- fisher.test(steal_thief_profile_eng_idn_matrix)
steal_thief_profile_eng_idn_effsize <- vcd::assocstats(steal_thief_profile_eng_idn_matrix)
steal_thief_profile_eng_idn_phicoeff <- round(steal_thief_profile_eng_idn_effsize$phi, 2)
```



```{r cross-tabulation-and-signif-test-GOODS}
rob_goods_profile_eng_idn <- rob1 %>% 
  mutate(core_args = if_else(SYN_ENG_GOODS == "PP_complement", "FALSE", "TRUE"),
         core_args = if_else(is.na(SYN_ENG_GOODS), "NA", core_args)) %>% 
  filter(core_args != "FALSE") %>% # select only the core argument of GOODS in English
  count(ROLE_ENG_GOODS, ROLE_IDN_GOODS) %>% 
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "GOODS_not_mentioned", "GOODS_mentioned"))) %>% 
  rename(English = ROLE_ENG_GOODS,
         Indonesian = ROLE_IDN_GOODS) %>% 
  # arrange(desc(n))
  arrange(English)
rob_goods_profile_eng_idn_matrix <- rob_goods_profile_eng_idn %>% 
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>% 
  column_to_rownames("English") %>% 
  as.matrix()
names(attr(rob_goods_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
rob_goods_profile_eng_idn_matrix <- rob_goods_profile_eng_idn_matrix[ , c(2, 1)]
rob_goods_profile_eng_idn_fye <- fisher.test(rob_goods_profile_eng_idn_matrix)
rob_goods_profile_eng_idn_effsize <- vcd::assocstats(rob_goods_profile_eng_idn_matrix)
rob_goods_profile_eng_idn_phicoeff <- round(rob_goods_profile_eng_idn_effsize$phi, 2)
rob_goods_semantics_eng <- rob1 %>% 
  count(SEM_ENG_GOODS, sort = TRUE) %>% 
  filter(if_all(where(is.character), ~!is.na(.))) %>% 
  mutate(perc = round(n/sum(n) * 100, 1))
rob_goods_semantics_idn <- rob1 %>% 
  count(SEM_IDN_GOODS, sort = TRUE) %>% 
  filter(if_all(where(is.character), ~!is.na(.))) %>% 
  mutate(perc = round(n/sum(n) * 100, 1))

steal_goods_profile_eng_idn <- steal1 %>% 
  count(ROLE_ENG_GOODS, ROLE_IDN_GOODS) %>% 
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "GOODS_not_mentioned", "GOODS_mentioned"))) %>% 
  rename(English = ROLE_ENG_GOODS,
         Indonesian = ROLE_IDN_GOODS) %>% 
  # arrange(desc(n))
  arrange(English)
steal_goods_profile_eng_idn_matrix <- steal_goods_profile_eng_idn %>% 
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>% 
  column_to_rownames("English") %>%  
  as.matrix()
names(attr(steal_goods_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
steal_goods_profile_eng_idn_matrix <- steal_goods_profile_eng_idn_matrix[, c(2,1)]
steal_goods_profile_eng_idn_fye <- fisher.test(steal_goods_profile_eng_idn_matrix)
steal_goods_profile_eng_idn_effsize <- vcd::assocstats(steal_goods_profile_eng_idn_matrix)
steal_goods_profile_eng_idn_phicoeff <- round(steal_goods_profile_eng_idn_effsize$phi, 2)
steal_goods_semantics_eng <- steal1 %>% 
  count(SEM_ENG_GOODS, sort = TRUE) %>% 
  filter(if_all(where(is.character), ~!is.na(.))) %>% 
  mutate(perc = round(n/sum(n) * 100, 1))
steal_goods_semantics_eng_no_pronoun <- steal1 %>% 
  count(SEM_ENG_GOODS, sort = TRUE) %>% 
  filter(if_all(where(is.character), ~!is.na(.)), SEM_ENG_GOODS != "PRONOUN") %>% 
  mutate(perc = round(n/sum(n) * 100, 1))
steal_goods_semantics_idn <- steal1 %>% 
  count(SEM_IDN_GOODS, sort = TRUE) %>% 
  filter(if_all(where(is.character), ~!is.na(.))) %>% 
  mutate(perc = round(n/sum(n) * 100, 1))
```



```{r cross-tabulation-and-signif-test-TARGETS}
rob_target_profile_eng_idn <- rob1 %>%
  filter(SYN_ENG_TARGET %in% c("direct_object", "passive_subject")) %>% 
  mutate(ROLE_IDN_TARGET = replace(ROLE_IDN_TARGET, SYN_IDN_TARGET %in% c("nominal_modifier", "possessive_determiner", "pp_complement"), FALSE)) %>% 
  count(ROLE_ENG_TARGET, ROLE_IDN_TARGET) %>%
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "TARGET_not_mentioned", "TARGET_mentioned"))) %>%
  rename(English = ROLE_ENG_TARGET,
         Indonesian = ROLE_IDN_TARGET) %>% 
  arrange(desc(n))
rob_target_profile_eng_idn_matrix <- rob_target_profile_eng_idn %>%
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>%
  column_to_rownames("English") %>% 
  data.frame() %>%
  as.matrix()
names(attr(rob_target_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
rob_target_animacy <- rob1 %>% 
  count(ANIMACY_ENG_TARGET, name = "n_rob") %>% 
  mutate(perc_rob = round(n_rob/sum(n_rob) * 100, 2)) %>% 
  arrange(desc(n_rob))
rob_target_type <- rob1 %>% 
  count(TYPE_ENG_TARGET, name = "n_rob") %>% 
  mutate(perc_rob = round(n_rob/sum(n_rob) * 100, 2)) %>% 
  arrange(desc(n_rob))
rob_target_semantics <- rob1 %>% 
  count(SEM_ENG_TARGET, name = "n_rob") %>% 
  mutate(perc_rob = round(n_rob/sum(n_rob) * 100, 2)) %>% 
  arrange(desc(n_rob))

# ROB's TARGET is always mentioned in English, but there are few cases where it is not mentioned in the translation. So there is only one variable of interest, namely (not-)mention of the Target in Indonesian, with two outcomes: mention vs. not mention of TARGET. This is the motivation to run the Exact Binomial Test
rob_target_profile_eng_idn_binom <- binom.test(rob_target_profile_eng_idn_matrix)

steal_target_profile_eng_idn <- steal1 %>% 
  filter(!SYN_ENG_TARGET %in% c("possessive_determiner", "possessive_of-phrase", "pp_complement")) %>% 
  count(ROLE_ENG_TARGET, ROLE_IDN_TARGET) %>% 
  mutate(across(where(is.logical), function(x) ifelse(x == FALSE, "TARGET_not_mentioned", "TARGET_mentioned"))) %>% 
  rename(English = ROLE_ENG_TARGET,
         Indonesian = ROLE_IDN_TARGET) %>% 
  arrange(desc(n))
steal_target_profile_eng_idn_matrix <- steal_target_profile_eng_idn %>% 
  pivot_wider(names_from = Indonesian, values_from = n, values_fill = 0L) %>% 
  column_to_rownames("English") %>%  
  as.matrix()
names(attr(steal_target_profile_eng_idn_matrix, "dimnames")) <- c("English", "Indonesian")
steal_target_profile_eng_idn_fye <- fisher.test(steal_target_profile_eng_idn_matrix)
steal_target_profile_eng_idn_effsize <- vcd::assocstats(steal_target_profile_eng_idn_matrix)
steal_target_profile_eng_idn_phicoeff <- steal_target_profile_eng_idn_effsize$phi
steal_target_animacy <- steal1 %>% 
  filter(!is.na(ANIMACY_ENG_TARGET)) %>% 
  count(ANIMACY_ENG_TARGET, name = "n_steal") %>% 
  mutate(perc_steal = round(n_steal/sum(n_steal) * 100, 2)) %>% 
  arrange(desc(n_steal))
steal_target_type <- steal1 %>% 
  filter(!is.na(TYPE_ENG_TARGET)) %>% 
  count(TYPE_ENG_TARGET, name = "n_steal") %>% 
  mutate(perc_steal = round(n_steal/sum(n_steal) * 100, 2)) %>% 
  arrange(desc(n_steal))
steal_target_semantics <- steal1 %>% 
  filter(!is.na(SEM_ENG_TARGET)) %>% 
  count(SEM_ENG_TARGET, name = "n_steal") %>% 
  mutate(perc_steal = round(n_steal/sum(n_steal) * 100, 2)) %>% 
  arrange(desc(n_steal))


rob_steal_target_semantics <- steal_target_semantics %>% 
  full_join(rob_target_semantics, by = "SEM_ENG_TARGET")
rob_steal_target_semantics1 <- rob_steal_target_semantics %>% 
  filter(if_all(where(is.numeric), ~!is.na(.))) %>% 
  mutate(perc_steal = round(n_steal/sum(n_steal) * 100, 1),
         perc_rob = round(n_rob/sum(n_rob) * 100, 1))
rob_steal_target_semantics1_mtx <- rob_steal_target_semantics1 %>% 
  select(matches("^SEM|^n_")) %>% 
  column_to_rownames("SEM_ENG_TARGET") %>% 
  as.matrix()
rob_steal_target_semantics_chisq <- chisq.test(rob_steal_target_semantics1_mtx, correct = FALSE)
steal_target_pbin <- rob_steal_target_semantics1 %>% pull(n_steal) %>% binom.test() %>% .$p.value
rob_target_pbin <- rob_steal_target_semantics1 %>% pull(n_rob) %>% binom.test() %>% .$p.value
```


For the semantic types of the TARGET, HUMAN is highly significantly more often as the filler of the TARGET of STEAL (N=`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(n_steal)`; `r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(perc_steal)`%) compared to the LOCATION/PLACE (N=`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(n_steal)`; `r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(perc_steal)`%) (*p*~Binomial~ ~two-tailed~ `r if (steal_target_pbin > 0.01 & steal_target_pbin < 0.05) " < 0.05" else if (steal_target_pbin > 0.001 & steal_target_pbin < 0.01) " < 0.01" else if (steal_target_pbin < 0.001) " < 0.001"`). In contrast, for ROB, the distribution of the semantic types of its TARGET as HUMAN (N=`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(n_rob)`; `r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(perc_rob)`%) and LOCATION/PLACE (N=`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(n_rob)`; `r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(perc_rob)`%) is not significantly different (*p*~Binomial~ ~two-tailed~ `r if (rob_target_pbin > 0.05) paste("=", round(rob_target_pbin, 3), sep = "") else if (rob_target_pbin > 0.01 & rob_target_pbin < 0.05) " < 0.05" else if (rob_target_pbin > 0.001 & rob_target_pbin < 0.01) " < 0.01" else if (rob_target_pbin < 0.001) " < 0.001"`). Overall, the proportion of HUMAN TARGET is greater for STEAL (`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(perc_steal)`%) than for ROB (`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "HUMAN") %>% pull(perc_rob)`%), while the proportion of LOCATION/PLACE TARGET is greater for ROB (`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(perc_rob)`%) than for STEAL (`r rob_steal_target_semantics1 %>% filter(SEM_ENG_TARGET == "PLACE") %>% pull(perc_steal)`%); these asymmetries are statistically significant (*X*^2^=`r round(rob_steal_target_semantics_chisq$statistic, 2)`; *df*=`r rob_steal_target_semantics_chisq$parameter`; *p*~two~~-~~tailed~ `r if (rob_steal_target_semantics_chisq$p.value > 0.05) paste("=", round(rob_steal_target_semantics_chisq$p.value, 3), sep = "") else if (rob_steal_target_semantics_chisq$p.value > 0.01 & rob_steal_target_semantics_chisq$p.value < 0.05) " < 0.05" else if (rob_steal_target_semantics_chisq$p.value > 0.001 & rob_steal_target_semantics_chisq$p.value < 0.01) " < 0.01" else if (rob_steal_target_semantics_chisq$p.value < 0.001) " < 0.001"`) with moderate effect size ($\phi$ =`r round(assocstats(rob_steal_target_semantics1_mtx)$phi, 3)`).


### constructional equivalence for the (de-)profiling of THIEF

```{r plot-profiled-participants-THIEF-ENG-IDN-ROB-STEAL-barplot-idn-lang, fig.cap = "Kesepadanan konstruksional ke(tidak)munculan peran PENCURI lema ROB dan STEAL", fig.asp=.75}
rs_thief_profile_eng_idn <- rob_thief_profile_eng_idn %>% 
  mutate(Lemma = "ROB",
         English = factor(English, levels = unique(English)),
         Indonesian = factor(Indonesian, levels = unique(Indonesian))) %>%
  bind_rows(steal_thief_profile_eng_idn %>% 
              mutate(Lemma = "STEAL",
                     English = factor(English, levels = unique(English)),
                     Indonesian = factor(Indonesian, levels = unique(Indonesian)))) %>% 
  mutate(Lemma = factor(Lemma, levels = c("ROB", "STEAL"))) %>% 
  group_by(Lemma, English) %>% 
  mutate(perc = n / sum(n) * 100,
         perc = round(perc, digits = 1))


p2_thief_eng_idn <- rs_thief_profile_eng_idn %>% 
  # ggplot(aes(x = English, y = perc, fill = Indonesian)) +
  # geom_col() + 
  # geom_text(aes(label = ifelse(rs_thief_profile_eng_idn$perc > 5, n, "")),
  #           position = position_stack(vjust = 0.5)) +
  ggplot(aes(x = English, y = n, fill = Indonesian)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste(perc, "% (N=", n, ")", sep = "")),
            position = position_dodge(.9), vjust = -.5, size = 3) +
  facet_wrap(~Lemma, scales = "free_x") +
  theme_bw() +
  scale_x_discrete(labels = c("THIEF_profiled_Eng", "THIEF_not-profiled_Eng")) +
  scale_fill_manual(values = brewer.pal(11, "RdYlGn")[c(7, 1)],
                    labels = c("THIEF_profiled_Idn", "THIEF_not-profiled_Idn"),
                    name = "Indonesian translation") +
  labs(y = "Token frequency",
       # y = "Persentase",
       caption = "Numbers inside the brackets are the observed token frequencies.",
       x = "English source text") +
  theme(axis.text = element_text(size = 9),
        axis.title = element_text(size = 15),
        legend.title = element_text(size = 15),
        legend.text = element_text(size = 9.75),
        legend.position = "top",
        plot.caption = element_text(size = 9))

p2_thief_eng_idn

# ggsave("figs/plot-profiled-participants-THIEF-all-verbs.png", plot = p2_thief_eng_idn, height = 5, width = 7.5, dpi = 300)
```

For ROB, `r rs_thief_profile_eng_idn %>% filter(Lemma == "ROB", English == "THIEF_mentioned", perc == max(perc)) %>% pull(perc)`% from the total `r rs_thief_profile_eng_idn %>% filter(Lemma == "ROB", English == "THIEF_mentioned") %>% pull(n) %>% sum()` cases where the THIEF is profiled in English,,,

THIEF not profiled in English is in `r rs_thief_profile_eng_idn %>% filter(Lemma == "ROB", English == "THIEF_not_mentioned", perc == max(perc)) %>% pull(perc)`% of all `r rs_thief_profile_eng_idn %>% filter(Lemma == "ROB", English == "THIEF_not_mentioned") %>% pull(n) %>% sum()`

This is a highly significant correlation with robust effect (Phi-coefficient = `r rob_thief_profile_eng_idn_phicoeff`).

Odds ratio for the THIEF to be profiled in the English and Indonesian translation is `r format(round(rob_thief_profile_eng_idn_fye$estimate, 1), big.mark = ",")`.



### constructional equivalence for the (de-)profiling of GOODS

```{r plot-profiled-participants-GOODS-ENG-IDN-ROB-STEAL-barplot-idn-lang, fig.cap = "Kesepadanan konstruksional ke(tidak)munculan peran BARANG lema ROB dan STEAL", fig.asp=.75}

rs_goods_profile_eng_idn <- rob_goods_profile_eng_idn %>% 
  mutate(Lemma = "ROB",
         English = factor(English, levels = unique(English)),
         Indonesian = factor(Indonesian, levels = unique(Indonesian))) %>%
  bind_rows(steal_goods_profile_eng_idn %>% 
              mutate(Lemma = "STEAL",
                     English = factor(English, levels = unique(English)),
                     Indonesian = factor(Indonesian, levels = unique(Indonesian)))) %>% 
  mutate(Lemma = factor(Lemma, levels = c("ROB", "STEAL"))) %>% 
  group_by(Lemma, English) %>% 
  mutate(perc = n / sum(n) * 100,
         perc = round(perc, digits = 1))

p2_goods_eng_idn <- rs_goods_profile_eng_idn %>% 
  # ggplot(aes(x = English, y = perc, fill = Indonesian)) +
  # geom_col() + 
  # geom_text(aes(label = ifelse(rs_goods_profile_eng_idn$perc > 10, n, "")),
  #           position = position_stack(vjust = 0.5)) +
  ggplot(aes(x = English, y = n, fill = Indonesian)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste(perc, "% (N=", n, ")", sep = "")),
            position = position_dodge(.9), vjust = -.5, size = 3) +
  facet_wrap(~Lemma, scales = "free_x") +
  theme_bw() +
  scale_x_discrete(labels = c("GOODS_profiled_Eng", "GOODS_not-profiled_Eng")) +
  scale_fill_manual(values = brewer.pal(11, "RdYlGn")[c(7, 1)],
                    labels = c("GOODS_profiled_Idn", "GOODS_not-profiled_Idn"),
                    name = "Indonesian translation") +
  labs(y = "Token frequency",
       # y = "Persentase",
       caption = "Numbers inside the brackets are the observed token frequencies.",
       x = "English source text") +
  theme(axis.text = element_text(size = 9),
        axis.title = element_text(size = 15),
        legend.title = element_text(size = 15),
        legend.text = element_text(size = 9.75),
        legend.position = "top",
        plot.caption = element_text(size = 9))
p2_goods_eng_idn

# ggsave("figs/plot-profiled-participants-GOODS-all-verbs.png", plot = p2_goods_eng_idn, height = 5, width = 7.5, dpi = 300)
```

From the total `r rs_goods_profile_eng_idn %>% filter(English == "GOODS_not_mentioned", Lemma == "ROB") %>% pull(n) %>% sum()` casen when the GOODS is not profiled by ROB, the Indonesian translations preserve this deprofiling in `r rs_goods_profile_eng_idn %>% filter(English == "GOODS_not_mentioned", Lemma == "ROB", perc == max(perc)) %>% pull(perc)`% of the cases.


### constructional equivalence for the (de-)profiling of TARGET

```{r plot-profiled-participants-TARGET-ENG-IDN-ROB-STEAL-barplot-idn-lang, fig.cap = "Kesepadanan konstruksional ke(tidak)munculan peran TARGET lema ROB dan STEAL", fig.asp=.75}
rs_target_profile_eng_idn <- rob_target_profile_eng_idn %>% 
  mutate(Lemma = "ROB",
         English = factor(English, levels = unique(English)),
         Indonesian = factor(Indonesian, levels = unique(Indonesian))) %>%
  bind_rows(steal_target_profile_eng_idn %>% 
              mutate(Lemma = "STEAL",
                     English = factor(English, levels = unique(English)),
                     Indonesian = factor(Indonesian, levels = unique(Indonesian)))) %>% 
  mutate(Lemma = factor(Lemma, levels = c("ROB", "STEAL"))) %>% 
  group_by(Lemma, English) %>% 
  mutate(perc = n / sum(n) * 100,
         perc = round(perc, digits = 1))


p2_target_eng_idn <- rs_target_profile_eng_idn %>% 
  # ggplot(aes(x = English, y = perc, fill = Indonesian)) +
  # geom_col() + 
  # geom_text(aes(label = ifelse(rs_target_profile_eng_idn$perc > 10, n, "")),
  #           position = position_stack(vjust = 0.5)) +
  ggplot(aes(x = English, y = n, fill = Indonesian)) +
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = paste(perc, "% (N=", n, ")", sep = "")),
            position = position_dodge(.9), vjust = -.5, size = 3) +
  facet_wrap(~Lemma, scales = "free_x") +
  theme_bw() +
  scale_x_discrete(labels = c("TARGET_profiled_Eng", "TARGET_not-profiled_Eng")) +
  scale_fill_manual(values = brewer.pal(11, "RdYlGn")[c(7, 1)],
                    labels = c("TARGET_profiled_Idn", "TARGET_not-profiled_Idn"),
                    name = "Indonesian translation") +
  labs(y = "Token frequency",
       # y = "Persentase",
       caption = "Numbers inside the brackets are the observed token frequencies.",
       x = "English source text") +
  theme(axis.text = element_text(size = 9.5),
        axis.title = element_text(size = 15),
        legend.title = element_text(size = 15),
        legend.text = element_text(size = 9.5),
        legend.position = "top",
        plot.caption = element_text(size = 9))
p2_target_eng_idn

# ggsave("figs/plot-profiled-participants-TARGET-all-verbs.png", plot = p2_target_eng_idn, height = 5, width = 7.5, dpi = 300)
```




### Construction Type for R&S and its translation equivalence

```{r count-of-construction-patterns-for-each-lang-and-each-verb}
## count of constructional patterns for ROB======
### English data
rob_cxn_eng <- rob1 %>% 
  count(CXN_ENG, sort = TRUE, name = "N_CXN_ENG") %>% 
  filter(!is.na(CXN_ENG), str_detect(CXN_ENG, "-PP$", negate = TRUE))

### Indonesian translation data
rob_cxn_idn <- rob1 %>% 
  count(CXN_IDN, sort = TRUE, name = "N_CXN_IDN") %>% 
  filter(!is.na(CXN_IDN))

### x-tab for English and Indonesian data
rob_cxn <- rob1 %>% 
  count(CXN_ENG, CXN_IDN, sort = TRUE, name = "a") %>% 
  filter(CXN_ENG != "TARGET-PP") %>% 
  mutate(CXN_ENG_lab = factor(CXN_ENG, levels = unique(CXN_ENG)),
         CXN_IDN_col = factor(CXN_IDN, levels = unique(CXN_IDN)))

## count of constructional patterns for STEAL======
### English data
steal_cxn_eng <- steal1 %>% 
  count(CXN_ENG, sort = TRUE, name = "N_CXN_ENG") %>% 
  filter(!is.na(CXN_ENG))

### Indonesian translation data
steal_cxn_idn <- steal1 %>% 
  count(CXN_IDN, sort = TRUE, name = "N_CXN_IDN") %>% 
  filter(!is.na(CXN_IDN))

### x-tab for English and Indonesian data
steal_cxn <- steal1 %>% 
  count(CXN_ENG, CXN_IDN, sort = TRUE, name = "a") %>% 
  mutate(CXN_ENG_lab = factor(CXN_ENG, levels = unique(CXN_ENG)),
         CXN_IDN_col = factor(CXN_IDN, levels = unique(CXN_IDN)))

brew_paired <- brewer.pal(brewer.pal.info[rownames(brewer.pal.info) == "Paired", 1], "Paired")
cxn_idn <- unique(c(rob_cxn$CXN_IDN, steal_cxn$CXN_IDN))
brew_paired_selected <- brew_paired[c(2, 1, 6, 7, 3, 5)]
names(brew_paired_selected) <- cxn_idn
cxn_idn_colours <- brew_paired_selected

```


```{r define-statistical-functions}
fye_compute <- function(a, b, c, d, alt) {
  return(fisher.test(matrix(c(a, b, c, d), nrow = 2), alternative = alt)$p.value)
}

oddsratio <- function(a, b, c, d, alt) {
  return(fisher.test(matrix(c(a, b, c, d), nrow = 2), alternative = alt)$estimate)
}

phicoefficient <- function(a, b, c, d) {
  return(round(vcd::assocstats(matrix(c(a, b, c, d), nrow = 2))$phi, digits = 2))
}
```



```{r cxn-plot-ROB, fig.cap = "Distribusi padanan bahasa Indonesia konstruksi verba ROB", fig.asp = 0.718, fig.width = 7.85}
p1_rob_cxn_idn <- rob_cxn %>% 
  ggplot(aes(x = CXN_ENG_lab, y = a, fill = CXN_IDN_col)) + 
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = a), 
            position = position_dodge(.9), 
            vjust = -.5,
            size = 4) +
  labs(x = "Construction types (English)",
       y = "Token frequency",
       fill = "Construction types\n(Indonesian)") +
  theme_light() +
  theme(axis.text = element_text(size = 9.75),
        axis.title = element_text(size = 14),
        legend.text = element_text(size = 9.75),
        legend.title = element_text(size = 14),
        plot.title = element_text(size = 15)) +
  scale_fill_manual(values = cxn_idn_colours[names(cxn_idn_colours) %in% unique(rob_cxn$CXN_IDN)])

p1_rob_cxn_idn

# ggsave("figs/plot-cxn-equivalence-ROB.png", plot = p1_rob_cxn_idn, height = 6, width = 8, dpi = 300)
  
```

```{r cxn-fye-ROB}
## FYE computation=======
rob_cxn_fye <- rob_cxn %>% 
  select(-CXN_ENG_lab, -CXN_IDN_col) %>% 
  complete(CXN_ENG, CXN_IDN, fill = list(a = 0L)) %>% 
  left_join(rob_cxn_eng, by = "CXN_ENG") %>% 
  left_join(rob_cxn_idn, by = "CXN_IDN") %>% 
  mutate(perc = (a/N_CXN_ENG) * 100,
         SUM = sum(a),
         b = N_CXN_ENG - a,
         c = N_CXN_IDN - a,
         d = SUM - (a + b + c),
         a_exp = (N_CXN_ENG * N_CXN_IDN)/SUM,
         assoc = if_else(a > a_exp, "attraction", "repulsion"),
         alt = if_else(a > a_exp, "greater", "less"),
         pfye = pmap_dbl(list(a, b, c, d, alt), fye_compute),
         collstr = if_else(alt == "greater", -log10(pfye), log10(pfye)),
         pholm = p.adjust(pfye, method = "holm"),
         signif_fye = if_else(pfye < 0.05, TRUE, FALSE),
         signif_holm = if_else(pholm < 0.05, TRUE, FALSE),
         signif_holm_dec = if_else(pholm < 0.05 & pholm > 0.01, "*", "ns"),
         signif_holm_dec = if_else(pholm < 0.01 & pholm > 0.001, "**", signif_holm_dec),
         signif_holm_dec = if_else(pholm < 0.001, "***", signif_holm_dec),
         odds_ratio = pmap_dbl(list(a, b, c, d, alt), oddsratio),
         phi = pmap_dbl(list(a, b, c, d), phicoefficient))
rob_cxn_fye_short <- rob_cxn_fye %>% 
  select(-N_CXN_ENG, -N_CXN_IDN, -SUM, -b, -c, -d, -a_exp, -alt, -signif_fye, -signif_holm) %>% 
  arrange(desc(collstr))
rob_cxn_fye_significant <- rob_cxn_fye_short %>% 
  filter(signif_holm_dec != "ns", pholm < 0.001)
```

p-value (Holm's adjustment) for T-obj_T-obj pattern = `r rob_cxn_fye_significant %>% filter(CXN_ENG == "TARGET-OBJ", CXN_IDN == "TARGET-OBJ") %>% pull(pholm) %>% format(digits = 4)`. The effect is strong (Phi-coefficient = `r rob_cxn_fye_significant %>% filter(CXN_ENG == "TARGET-OBJ", CXN_IDN == "TARGET-OBJ") %>% pull(phi)`).


p-value (Holm's adjustment) for T-subj-pass_T-subj-pass pattern = `r rob_cxn_fye_significant %>% filter(CXN_ENG == "TARGET-SUBJ-PASS", CXN_IDN == "TARGET-SUBJ-PASS") %>% pull(pholm) %>% format(digits = 4)`. The effect is also strong (Phi-coefficient = `r rob_cxn_fye_significant %>% filter(CXN_ENG == "TARGET-SUBJ-PASS", CXN_IDN == "TARGET-SUBJ-PASS") %>% pull(phi)`).

```{r cxn-plot-STEAL, fig.asp = 0.718, fig.cap = "Distribusi padanan bahasa Indonesian konstruksi verba STEAL"}
steal_colours <- cxn_idn_colours[names(cxn_idn_colours) %in% unique(steal_cxn$CXN_IDN)]
steal_colours <- steal_colours[order(names(steal_colours))]
p1_steal_cxn_idn <- steal_cxn %>% 
  ggplot(aes(x = CXN_ENG_lab, y = a, fill = CXN_IDN_col)) + 
  geom_col(position = position_dodge(.9)) +
  geom_text(aes(label = a), 
            position = position_dodge(.9), 
            vjust = -.5,
            size = 4) +
  labs(x = "Construction types (English)",
       y = "Token frequency",
       fill = "Construction types\n(Indonesian)") +
  theme_light() +
  theme(axis.text = element_text(size = 9.75),
        axis.title = element_text(size = 14),
        legend.text = element_text(size = 9.75),
        legend.title = element_text(size = 14),
        plot.title = element_text(size = 15)) +
  scale_fill_manual(values = steal_colours)

p1_steal_cxn_idn

# ggsave("figs/plot-cxn-equivalence-STEAL.png", plot = p1_steal_cxn_idn, height = 6, width = 8, dpi = 300)
  
```


```{r cxn-fye-STEAL}
## FYE computation=======
steal_cxn_fye <- steal_cxn %>% 
  select(-CXN_ENG_lab, -CXN_IDN_col) %>% 
  complete(CXN_ENG, CXN_IDN, fill = list(a = 0L)) %>% 
  left_join(steal_cxn_eng, by = "CXN_ENG") %>% 
  left_join(steal_cxn_idn, by = "CXN_IDN") %>% 
  mutate(perc = (a/N_CXN_ENG) * 100,
         SUM = sum(a),
         b = N_CXN_ENG - a,
         c = N_CXN_IDN - a,
         d = SUM - (a + b + c),
         a_exp = (N_CXN_ENG * N_CXN_IDN)/SUM,
         assoc = if_else(a > a_exp, "attraction", "repulsion"),
         alt = if_else(a > a_exp, "greater", "less"),
         pfye = pmap_dbl(list(a, b, c, d, alt), fye_compute),
         collstr = if_else(alt == "greater", -log10(pfye), log10(pfye)),
         pholm = p.adjust(pfye, method = "holm"),
         signif_fye = if_else(pfye < 0.05, TRUE, FALSE),
         signif_holm = if_else(pholm < 0.05, TRUE, FALSE),
         signif_holm_dec = if_else(pholm < 0.05 & pholm > 0.01, "*", "ns"),
         signif_holm_dec = if_else(pholm < 0.01 & pholm > 0.001, "**", signif_holm_dec),
         signif_holm_dec = if_else(pholm < 0.001, "***", signif_holm_dec),
         odds_ratio = pmap_dbl(list(a, b, c, d, alt), oddsratio),
         phi = pmap_dbl(list(a, b, c, d), phicoefficient))
steal_cxn_fye_short <- steal_cxn_fye %>% 
  select(-N_CXN_ENG, -N_CXN_IDN, -SUM, -b, -c, -d, -a_exp, -alt, -signif_fye, -signif_holm) %>% 
  arrange(desc(collstr))
steal_cxn_fye_significant <- steal_cxn_fye_short %>% 
  filter(signif_holm_dec != "ns", pholm < 0.001)
```

p-value (Holm's adjustment) for G-obj_G-obj pattern = `r steal_cxn_fye_significant %>% filter(CXN_ENG == "GOODS-OBJ", CXN_IDN == "GOODS-OBJ") %>% pull(pholm) %>% format(digits = 4)`. The effect is strong (Phi-coefficient = `r steal_cxn_fye_significant %>% filter(CXN_ENG == "GOODS-OBJ", CXN_IDN == "GOODS-OBJ") %>% pull(phi)`).


p-value (Holm's adjustment) for G-subj-pass_G-subj-pass pattern = `r steal_cxn_fye_significant %>% filter(CXN_ENG == "GOODS-SUBJ-PASS", CXN_IDN == "GOODS-SUBJ-PASS") %>% pull(pholm) %>% format(digits = 4)`. The effect is also strong (Phi-coefficient = `r steal_cxn_fye_significant %>% filter(CXN_ENG == "GOODS-SUBJ-PASS", CXN_IDN == "GOODS-SUBJ-PASS") %>% pull(phi)`).