papers.qmd

---
title: "Systematic literature review"
bibliography: references.bib
title-block-banner: true
subtitle: "A papers' analysis"
author:
  - name: Olivier Caron
    email: olivier.caron@dauphine.psl.eu
    affiliations: 
      name: "Paris Dauphine - PSL"
      city: Paris
      state: France
  - name: Christophe Benavent
    email: christophe.benavent@dauphine.psl.eu
    affiliations: 
      name: "Paris Dauphine - PSL"
      city: Paris
      state: France
date : "last-modified"
toc: true
number-sections: true
number-depth: 5
format:
  html:
    theme:
      light: yeti
      dark: darkly
    code-fold: true
    code-summary: "Display code"
    code-tools: true #enables to display/hide all blocks of code
    code-copy: true #enables to copy code
    grid:
      body-width: 1800px
      margin-width: 100px
    toc: true
    toc-location: left
  pdf: default
execute:
  echo: true
  warning: false
  message: false
editor: visual
fig-align: "center"
highlight-style: breeze
css: styles.css
reference-location: margin
---

## Libraries and loading data

```{r libraries and load data}
library(cowplot)
library(tidyverse)
library(ggstatsplot)

#read.csv transforms ":" in "." so it was complicated to replicate code
#every column name is now lowered, the dots are replaced with _
#every underscore at the start or end of the column name is deleted

nlp_papers <- read.csv("nlp_papers_utf8.csv", fileEncoding = "UTF-8")
nlp_full <- read.csv("nlp_full_data_utf8.csv", fileEncoding = "UTF-8")
new_colnames <- gsub("\\.", "_", colnames(nlp_papers)) %>%
  trimws(., whitespace = "_") %>%
  gsub("_+", "_", .) %>%
  tolower()


colnames(nlp_papers) <- new_colnames
```

## A look at the number of marketing publications with NLP

The number of papers in marketing reviews using NLP methods has increased a lot in recent years:\

```{r graph number of publications}
#| fig-width: 50
#| fig-height: 20
#| column: screen-inset-shaded
nlp_papers <- nlp_papers %>%
  mutate(year = substr(prism_coverdate,1,4))


#get rid of conference papers
nlp_papers_journal_only <- nlp_papers %>%
  filter(!grepl("conference", subtypedescription, ignore.case = TRUE)) %>%
  filter(year < 2023)


  
t0 <-as.data.frame(prop.table(table(nlp_papers_journal_only$prism_publicationname)))
g01<-ggplot(t0,aes(x=reorder(Var1, Freq), y=Freq))+geom_bar(stat="identity")+
  coord_flip()+
  labs( title="Number of articles per review", y="Proportion", x= NULL)

t1<-as.data.frame(table(nlp_papers_journal_only$year))

g02<-ggplot(t1, aes(x=Var1, y=Freq, group=1))+
  geom_smooth(color="Grey70", linewidth=2)+
  geom_line(stat="identity", size=1.1) + 
  labs( title="Number of publications per year", y="", x=NULL)
plotgrid <- plot_grid(g01,
                      g02,
                      label_size = 10,
                      ncol=2,
                      rel_widths =  c(2,1))
ggsave(filename="images/evolution_publications_nlp_marketing.png",
       width = 80, 
       height = 40, 
       units = "cm")
plotgrid


```

## A look at the number of citations

It seems that research endeavors within the nascent domain of marketing using NLP methodologies exhibit a resemblance to research conducted on a global scale regarding the number of citations. Indeed, it showcases a pronounced inequity, where numerous papers find themselves receiving minimal or no citations whatsoever.

The subsequent graphs are done with the ggstatsplot R package [@ggstatsplot].

```{r}
#| layout-nrow: 4
#| echo: fenced

set.seed(42)

gghistostats(
  data       = nlp_papers,
  x          = citedby_count,
  title      = "Distribution of citations",
  test.value = 12,
  binwidth   = 1,
  xlab = "Number of citations"
)


nlp_papers$log_citedby = log(nlp_papers$citedby_count+1)
ggscatterstats(
  data  = nlp_papers,
  x     = author_count,
  y     = log_citedby,
  xlab  = "Number of authors",
  ylab  = "Ln(Number of citations+1)",
  title = "Distribution of citations by number of authors",
  label.var = dc_creator,
  label.expression = citedby_count > 200,
  point.label.args = list(alpha = 0.7, size = 4, color = "grey50"),
  xfill = "#CC79A7", ## fill for marginals on the x-axis
  yfill = "#009E73", ## fill for marginals on the y-axis
)

ggstatsplot::ggbetweenstats(data = nlp_papers,
                            x = subtypedescription,
                            y = log_citedby,
                            xlab = "",
                            ylab = "Number of citations (ln scale)",
                            plottype = "box",
                            type = "p",
                            conf.level = 0.95)

nlp_article_per_country <- nlp_full %>%
  filter(!is.na(afid)) %>%
  group_by(entry_number) %>%
  reframe(unique_countries = unique(affiliation_country),
          nbcitations      = unique(citedby_count),
          publication_id   = unique(eid),
          title            = unique(dc.title),
          journal          = unique(prism.publicationName),
          first_author     = unique(dc.creator),
          date_publication = unique(prism.coverDate),
          nb_authors       = unique(author_count),
          keyword          = unique(authkeywords))


#dataframe with the sum of citations per country per article
nlp_article25 <- nlp_article_per_country %>%
  group_by(unique_countries) %>%
  summarise(sumcitations = sum(nbcitations)) %>%
  arrange(desc(sumcitations)) %>%
  head(25)
  

  ggdotplotstats(
  data       = nlp_article25,
  y          = unique_countries,
  x          = sumcitations,
  test.value = 25,
  type       = "robust",
  title      = "Distribution of academic citations among the top 25 countries",
  xlab       = "Number of citations"
)

#ggsave("images/distribution_citations_authors.svg", width=15, height = 12)
```

## The landscape of research

```{r}

```