paper.md
In evan-l-munson/SAoTD: Sentiment Analysis of Twitter Data

title: "Sentiment Analysis of Twitter Data (saotd)" authors: - affiliation: 1 name: Evan L. Munson orcid: 0000-0002-9958-6800 - affiliation: 1 name: Christopher M. Smith orcid: 0000-0002-8288-270X - affiliation: 1 name: Bradley C. Boehmke orcid: 0000-0002-3611-8516 - affiliation: 1 name: Jason K. Freels orcid: 0000-0002-2415-0340 date: "24 February 2019" output: pdf_document bibliography: paper.bib tags: - text mining - sentiment analysis - natural language processing - latent dirichlet allocation - twitter sentiment analysis affiliations: - index: 1 name: Air Force Institure of Technology

saotd is an R package that provides a programmatic interface to the Twitter API and can be used to acquire tweets based on user-specified #hashtags. The package will clean and tidy the Twitter data, determine the latent topics within the tweets utilizing Latent Dirichlet Allocation (LDA), determine a sentiment score using the Bing lexicon dictionary, and create output visualizations.

The package is available on GitHub and archived on Zenodo. To configure the package a user must follow these steps:

Creating a personal Twitter account if they don't already have one.
Enable a Twitter Developers Account.
Create an application which will provide them with API access tokens required by the saotd package to function.

The package is laid out in five different categories: Acquire, Explore, Topic Analysis, Sentiment Calculation, and Visualizations.

Acquire allows a user to acquire tweets of their choosing by accessing the Twitter API.
Explore provides functions to tidy, explore unigrams, bigrams, tri-grams, in addition to bigram netwrorks and correlation networks.
Topic analysis allows a user to explore the latent topics buried within the tweets.
Sentiment calculation utilizes the Bing lexicon dictionary to score the text [@Hu2004].
Visualizations allow the user to better understand the sentiment of the tweets.

The package utilizes tidy dataframes and therefore depends on the tidyverse package [@Wickham2017] and additionally uses the tidytext package [@Silge2017]. The number of latent topics is determined using the ldatuning package [@Nikita2016] and the latent dirichlet allocation (LDA) topics is determined using the topicmodels package [@Grun2011].

The saotd package has research applications in many disciplines which need to access tweets from the Twitter platform and carry out sentiment analyses. This package was created to quickly determine the sentiment of Twitter and to inform analysts on the opinions contained within tweets.