Sentiment Analysis
SentimentAnalysis performs a sentiment analysis of textual contents in R. This implementation utilizes various existing dictionaries, such as QDAP, Harvard IV or Loughran-McDonald. Furthermore, it can also create customized dictionaries. The latter uses LASSO regularization as a statistical approach to select relevant terms based on an exogenous response variable.
Overview
The most important functions in SentimentAnalysis are:
Compute sentiment scores from contents stored in different formats with
analyzeSentiment()
.If desired, convert the continuous scores to either binary sentiment classes (negative or positive) or tertiary directions (negative, neutral or positive). This conversion can be done with
convertToBinary()
orconvertToDirection()
respectively.Compare the calculated sentiment socres with a baseline (i.e. a gold standard). Here,
compareToResponse()
performs a statistical evaluation, whileplotSentimentResponse()
enables a visual comparison.Generate customized dictionaries with the help of
generateDictionary()
as part of an advanced analysis. However, this prerequisites a response variable (i.e. the baseline).
To see examples of these functions in use, check out the help pages, the demos and the vignette.
Usage
This section shows the basic functionality of how to perform a sentiment analysis. First, install the package from CRAN. Then load the corresponding package SentimentAnalysis.
# install.packages("SentimentAnalysis")
library(SentimentAnalysis)
Quick demonstration
This simple example shows how to perform a sentiment analysis of a single string. The result is a two-level factor with levels "positive" and "negative."
# Analyze a single string to obtain a binary response (positive / negative)
sentiment <- analyzeSentiment("Yeah, this was a great soccer game of the German team!")
convertToBinaryResponse(sentiment)$SentimentGI
#> [1] positive
#> Levels: negative positive
Small example
The following demonstrates some of the functionality provided by SentimentAnalysis. It also shows its visualization and evaluation capabilities.
# Create a vector of strings
documents <- c("Wow, I really like the new light sabers!",
"That book was excellent.",
"R is a fantastic language.",
"The service in this restaurant was miserable.",
"This is neither positive or negative.",
"The waiter forget about my a dessert -- what a poor service!")
# Analyze sentiment
sentiment <- analyzeSentiment(documents)
# Extract dictionary-based sentiment according to the QDAP dictionary
sentiment$SentimentQDAP
#> [1] 0.3333333 0.5000000 0.5000000 -0.3333333 0.0000000 -0.4000000
# View sentiment direction (i.e. positive, neutral and negative)
convertToDirection(sentiment$SentimentQDAP)
#> [1] positive positive positive negative neutral negative
#> Levels: negative neutral positive
response <- c(+1, +1, +1, -1, 0, -1)
compareToResponse(sentiment, response)
#> Warning in cor(sentiment, response): the standard deviation is zero
#> Warning in cor(x, y): the standard deviation is zero
#> Warning in cor(x, y): the standard deviation is zero
#> Warning in cor(sentiment, response): the standard deviation is zero
#> WordCount SentimentGI NegativityGI
#> cor -0.18569534 0.9900115 -0.99748901
#> cor.t.statistic -0.37796447 14.0440465 -28.16913204
#> cor.p.value -0.37796447 14.0440465 -28.16913204
#> lm.t.value -0.37796447 14.0440465 -28.16913204
#> r.squared 0.03448276 0.9801228 0.99498433
#> RMSE 3.82970843 0.4501029 1.18665418
#> MAE 3.33333333 0.4000000 1.10000000
#> Accuracy 0.66666667 1.0000000 0.66666667
#> Precision NaN 1.0000000 NaN
#> Sensitivity 0.00000000 1.0000000 0.00000000
#> Specificity 1.00000000 1.0000000 1.00000000
#> F1 0.00000000 0.5000000 0.00000000
#> BalancedAccuracy 0.50000000 1.0000000 0.50000000
#> avg.sentiment.pos.response 3.25000000 0.3333333 0.08333333
#> avg.sentiment.neg.response 4.00000000 -0.6333333 0.63333333
#> PositivityGI SentimentHE NegativityHE
#> cor 0.9429542 0.4152274 -0.083045480
#> cor.t.statistic 5.6647055 0.9128709 -0.166666667
#> cor.p.value 5.6647055 0.9128709 -0.166666667
#> lm.t.value 5.6647055 0.9128709 -0.166666667
#> r.squared 0.8891626 0.1724138 0.006896552
#> RMSE 0.7136240 0.8416254 0.922958207
#> MAE 0.6666667 0.7500000 0.888888889
#> Accuracy 0.6666667 0.6666667 0.666666667
#> Precision NaN NaN NaN
#> Sensitivity 0.0000000 0.0000000 0.000000000
#> Specificity 1.0000000 1.0000000 1.000000000
#> F1 0.0000000 0.0000000 0.000000000
#> BalancedAccuracy 0.5000000 0.5000000 0.500000000
#> avg.sentiment.pos.response 0.4166667 0.1250000 0.083333333
#> avg.sentiment.neg.response 0.0000000 0.0000000 0.000000000
#> PositivityHE SentimentLM NegativityLM
#> cor 0.3315938 0.7370455 -0.40804713
#> cor.t.statistic 0.7029595 2.1811142 -0.89389841
#> cor.p.value 0.7029595 2.1811142 -0.89389841
#> lm.t.value 0.7029595 2.1811142 -0.89389841
#> r.squared 0.1099545 0.5432361 0.16650246
#> RMSE 0.8525561 0.7234178 0.96186547
#> MAE 0.8055556 0.6333333 0.92222222
#> Accuracy 0.6666667 0.8333333 0.66666667
#> Precision NaN 1.0000000 NaN
#> Sensitivity 0.0000000 0.5000000 0.00000000
#> Specificity 1.0000000 1.0000000 1.00000000
#> F1 0.0000000 0.3333333 0.00000000
#> BalancedAccuracy 0.5000000 0.7500000 0.50000000
#> avg.sentiment.pos.response 0.2083333 0.2500000 0.08333333
#> avg.sentiment.neg.response 0.0000000 -0.1000000 0.10000000
#> PositivityLM RatioUncertaintyLM SentimentQDAP
#> cor 0.6305283 NA 0.9865356
#> cor.t.statistic 1.6247248 NA 12.0642877
#> cor.p.value 1.6247248 NA 12.0642877
#> lm.t.value 1.6247248 NA 12.0642877
#> r.squared 0.3975659 NA 0.9732526
#> RMSE 0.7757911 0.9128709 0.5398902
#> MAE 0.7222222 0.8333333 0.4888889
#> Accuracy 0.6666667 0.6666667 1.0000000
#> Precision NaN NaN 1.0000000
#> Sensitivity 0.0000000 0.0000000 1.0000000
#> Specificity 1.0000000 1.0000000 1.0000000
#> F1 0.0000000 0.0000000 0.5000000
#> BalancedAccuracy 0.5000000 0.5000000 1.0000000
#> avg.sentiment.pos.response 0.3333333 0.0000000 0.3333333
#> avg.sentiment.neg.response 0.0000000 0.0000000 -0.3666667
#> NegativityQDAP PositivityQDAP
#> cor -0.94433955 0.9429542
#> cor.t.statistic -5.74114834 5.6647055
#> cor.p.value -5.74114834 5.6647055
#> lm.t.value -5.74114834 5.6647055
#> r.squared 0.89177719 0.8891626
#> RMSE 1.06840137 0.7136240
#> MAE 1.01111111 0.6666667
#> Accuracy 0.66666667 0.6666667
#> Precision NaN NaN
#> Sensitivity 0.00000000 0.0000000
#> Specificity 1.00000000 1.0000000
#> F1 0.00000000 0.0000000
#> BalancedAccuracy 0.50000000 0.5000000
#> avg.sentiment.pos.response 0.08333333 0.4166667
#> avg.sentiment.neg.response 0.36666667 0.0000000
# Optional visualization: plotSentimentResponse(sentiment$SentimentQDAP, response)
Dictionary generation
Research in finance and social sciences nowadays utilizes content analysis to understand human decisions in the face of textual materials. While content analysis has received great traction lately, the available tools are not yet living up to the needs of researchers. This package implements a novel approach named "**dictionary generation" to study tone, sentiment and reception of textual materials.
The approach utilizes LASSO regularization to extract words from documents that statistically feature a positive and negative polarity. This immediately reveals manifold implications for practitioners, finance research and social sciences: researchers can use R to extract text components that are relevant for readers and test their hypothesis based on these.
- Proellochs, Feuerriegel and Neumann (2018): Statistical inferences for polarity identification in natural language, PLOS ONE 13(12):e0209323. DOI: 10.1371/journal.pone.0209323
License
SentimentAnalysis is released under the MIT License
Copyright (c) 2021 Stefan Feuerriegel & Nicolas Pröllochs