MLNLP.Rmd

---
title: "IntroNLP"
author: "Aishat Sadiq"
date: "`r Sys.Date()`"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

# Introduction to Natural Language Processing

```{r Data Prep}
#devtools::install_github("ccss-rs/NAMEOFREPO")
set.seed(31415)

## install.packages("tidyverse")
library(tidyverse)
# library(help = "tidyverse")
# ?tidyverse

## install.packages("tidytext")
## install.packages("topicmodels")
## install.packages("superheat")
## install.packages("ggrepel")
## install.packages("spacyr")
## install.packages("tm")
## install.packages("SnowballC")
## install.packages("stm")
## install.packages("textdata")
## install.packages("wordcloud")
library(tidytext)
library(topicmodels)
library(superheat)
library(ggrepel)
library(spacyr) # lem stem
library(tm) # TermDocumentMatrix
library(SnowballC) # Porter stemming 
# library(lda)
library(stm) # Structural Topic Modeling
library(textdata)
library(wordcloud)

```

## Import Dataset

```{r Import Dataset, echo=FALSE}

# find downloaded github folder
getwd()
#setwd(dir)

#Trump Tweets Dataset
# trump_tweets <- read_csv("ML-NLP/trump_tweets.csv")
# View(trump_tweets)

# Panama Papers Dataset
# nodes_entities <- read.csv("/Users/aishatsadiq/Library/Mobile Documents/iCloud~md~obsidian/Documents/PhD/CCSS Data Fellow/ML-NLP/full-oldb.LATEST/nodes-entities.csv")
# View(nodes_entities)

# Mass Mobilization Dataset
mmALL_073120_csv <- read_csv("mmALL_073120_csv.csv") %>%
  filter(year=="2020")
View(mmALL_073120_csv)

```

## Data Preprocessing

```{r Data Preprocessing}

?unnest_tokens  # Help R Doc for needed function
colnames(mmALL_073120_csv) # copy proper column name

# remove trailing and leading spaces 
mmALL_073120_csv$notes <- str_trim(mmALL_073120_csv$notes, side = "both")
head(mmALL_073120_csv$notes)

# Uppercasing w/ Base R toupper(), could also uuse tolower() for lowercase transformation
mmALL_073120_csv[c("notes")] <- sapply(mmALL_073120_csv[c("notes")], function(x) toupper(x))
head(mmALL_073120_csv$notes)

mmALL_073120_csv[c("notes")] <- sapply(mmALL_073120_csv[c("notes")], function(x) tolower(x))
head(mmALL_073120_csv$notes)

# Split raw notes to individual words/tokens, Remove stop words, remove stem w/ porter method, Count Words 

stem_counts <- mmALL_073120_csv %>%
  unnest_tokens(word, notes) %>%
  anti_join(stop_words) %>%
   mutate(stem = wordStem(word)) %>%
  count(id, stem) %>%
    arrange(desc(n))
View(stem_counts)

```

## Analyzing text statistics

```{r}
tf_idf <- mmALL_073120_csv %>%
  unnest_tokens(word, notes) %>%
  anti_join(stop_words) %>%
  mutate(stem = wordStem(word)) %>%
  count(id, stem, region, country) %>%
  arrange(desc(n))%>%
  bind_tf_idf(stem, id, n)
tf_idf 

top_tfidf <- tf_idf  %>%
  arrange(desc(tf_idf))  %>%
  drop_na()
top_tfidf

par(mfrow = c(1, 1))
top_tfidf %>%
    mutate(word = reorder_within(stem, tf_idf, country)) %>%
  filter(region=="Europe") %>%
    ggplot(aes(stem, tf_idf, fill = country)) +
    geom_col(alpha = 0.8, show.legend = FALSE) +
#    facet_wrap(~ country, scales = "free", ncol = 3) +
#    facet_wrap(~ country, scales = "free", ncol = 3) +
#    facet_grid_paginate(~ country, scales = "free", ncol = 3)+
    scale_x_reordered() +
    coord_flip() +
    theme(strip.text=element_text(size=5)) +
    labs(x = NULL, y = "tf-idf",
         title = "Highest tf-idf words in 2020 Mass Mobilization Data")

```

## Topic Modeling Algorithms

Text2vec Package - handle large text datasets with high efficiency Topicmodels Package - LDA, CTM, model evaluation, viz MALLET Package - LDA, Hierarchical LDA TM Package LSAfun Package - Latent Semantic Analysis (LSA) tsne Package - dimensionality reduction and data visualization

Latent Dirichlet Allocation (LDA)

```{r}
# LDA outputs: the topic word distributions and the document-topic memberships (from which topics does a document come from)

mmDTM <- DocumentTermMatrix(stem_counts)
mmDTM

mm_lda <- LDA(mmDTM, k = 10, control = list(seed = 31415))
mm_lda
as.data.frame(terms(mm_lda, 10))

## A LDA_VEM topic model with 4 topics.
topics <- tidy(mm_lda, matrix = "beta")
#topics
memberships <- tidy(mm_lda, matrix = "gamma")
#memberships

top_terms <- 
  topics %>%
  group_by(topic) %>%
  top_n(10, beta) %>%
  ungroup() %>%
  arrange(topic, -beta)
top_terms

```

## Lexicon Approach to Sentiment Analysis

Sentiment Analysis package cleanNLP Package

```{r echo=TRUE}

mm_Sent <- mmALL_073120_csv %>%
  unnest_tokens(word, notes) %>%
  anti_join(stop_words) %>%
  mutate(stem = wordStem(word)) %>%
  inner_join(get_sentiments("nrc"))
View(mm_Sent)


```

## Visualizing Text Data: Networks, Word Clouds, Bar Charts

Wordcloud Package

<https://quanteda.io/articles/pkgdown/examples/plotting.html>

```{r}
library(dplyr)

top_terms %>%
  mutate(term = reorder(term, beta)) %>%
  ggplot(aes(term, beta, fill = factor(topic))) +
  geom_col(show.legend = FALSE) +
  facet_wrap(~ topic, scales = "free") +
  coord_flip()

?wordcloud
wordcloud(mm_Sent$sentiment,colors="purple")

```