Analyzing the Opinions in a Big Text Document.
"Opitools"
An R-package for analyzing Opinions in Big Digital Text Document (DTD)
Description
The opitools is an R package for exploring a digital text document (DTD) as well as performing the impact analysis of the opinions expressed by the DTD. The package is particularly suited for opinion-based DTD, such as individual-level posts (e.g. commentaries on Twitter or Facebook) and comments (as in an online product reviews). First, an exploratory function word_imp can be used to identify words relating to different themes (or subjects) that are referenced in a DTD. The opi_impact function can then be utilized to investigate whether an identified/specified theme (or subject) impacts the overall opinion expressed by the document. The potentials of opitools for application across a wide range of domains is described briefly here (see the vignette for details).
Installation from CRAN
From an R console, type:
#install.packages("opitools")
library(opitools)
To install the developmental version of the package, type: remotes::install_github("MAnalytics/opitools"). Please, report any installation problems in the issues.
Example usage
Below is an example usage of how opitools can be employed to identify themes (or subjects) from a DTD and then deployed to perform opinion impact analysis.
Importing the dataset
The policing_dtd - a DTD comprising Twitter posts about police/policing in a neighbourhood, will be used in this demonstration.
> data(policing_dtd)
Identify themes (subjects)
Utilize word_imp function to highlights terms or words in accordance to their importance in the DTD. Through visual inspection, collate related terms or words that denote specific theme or subject from the generated wordcloud. Several words relating to the COVId-19 pandemic, including 'infect', 'pandemic', and 'lockdown' can be identified as been important in the document. These words are provided as covid_theme in the package. The function can be ran as follows (see the documentation for details):
> p1a <- word_imp(textdoc = policing_dtd, metric= "tf",
words_to_filter=c("police","policing"))
Impact analysis
The impact analysis can be conducted as follows:
#Running the analysis
results <- opi_impact(textdoc = policing_dtd, theme_keys=covid_theme, metric = 1,
fun = NULL, nsim = 99, alternative="two.sided",
pplot = TRUE, quiet=FALSE)
print(results)
$test
[1] "Test of significance (Randomization testing)"
$criterion
[1] "two.sided"
$exp_summary
Min. 1st Qu. Median Mean 3rd Qu. Max.
-8.240 -5.880 -5.880 -5.028 -3.530 -1.180
$p_table
|observed_score |S_beat |nsim |pvalue |signif |
|:--------------|:------|:----|:-------|:------|
|-5.88 |56 |99 |0.52 |' |
$p_key
[1] "0.99'" "0.05*" "0.025**" "0.01***"
$p_formula
[1] "(S_beat + 1)/(nsim + 1)"
#......
The research question of the analysis above can be stated as follows:
RQ1: "Does COVID-19 pandemic influence public opinion on neighourhood policing?"
The output shows an overall negative opinion (-5.88) of the public on the neighbourhood policing, and that the pandemic has not had a significant impacts (pvalue = 0.52) on the opinion expressed by the public. (More detailed explanation can be found in the study adepeju, M. and Jimoh, F. (2021)).
Other applications
Table 1 summarizes the analysis using different example datasets provided in the package. The output from the law enforcement application (as in above) is entered in the first row of the table. Other research questions investigated are as follow:
RQ2a: "Do the refreshment outlets/items impact customers’ opinion of the Piccadilly train services?"
RQ2b: "Do the signages influence customers’ opinion of the Piccadilly train services?"
RQ3: "How does the democratic candidate (Hillary Clinton) affects viewers’ opinion of the presidential debate?"
Table 1. Impact analysis results
| RQs | Primary data | Theme_keys | Score function | Observed Score (S) | P-value |
|:-----: | :----------: | :---------------: | :-------------:| :-----------------:| :---------:|
| RQ1 | policing_dtd | covid_theme | '1' | -5.88 | 0.52 |
| RQ2a | reviews_dtd | refreshment_theme | '1' | 67.92 | 0.01 |
| RQ2b | reviews_dtd | signage_theme | '1' | 67.92 | 0.1 |
| RQ3 | debate_dtd | direct input | '1' | -0.33 | 0.93 |
In each example, the same opinion score function is employed (metric = 1, i.e. the polarity score = (P - N)/(P + N)*100, where P and N represent positive and negative sentiments, respectively). See the documentation for details. Employing a threshold of p=0.05, any p-values less or equal to the threshold (e.g. RQ2a) represent a significant impact of the specified theme (i.e. refreshment_theme) on the overall opinion score computed based on the reviews_dtd.
References
- Adepeju, M. and Jimoh, F. (2021). An Analytical Framework for Measuring Inequality in the Public Opinions on Policing – Assessing the impacts of COVID-19 Pandemic using Twitter Data. https://doi.org/10.31235/osf.io/c32qh.