Learn R Programming

Topics

topics is an R-package enabling Differential Language Analysis using topics. It can produce LDA topics, but also support the text-package (www.r-text.org) in analyzing BERTtopics.

When using this package, please cite:

Ackermann L., Zhuojun G. & Kjell O.N.E. (2024). An R-package for visualizing text in topics. https://github.com/theharmonylab/topics. DOI:zenodo.org/records/11165378.

Installation

The topics package uses JAVA, which is another programming language. Please start by downloading and installing it from www.java.com/en/download/. Then open R and run:

install.packages("devtools")
devtools::install_github("theharmonylab/topics")

# if you run in to any installation problem, try installing rJava first.

# Before open the library, consider setting this option (can increase 5000);  without it the code may ran out of memory
options(java.parameters = "-Xmx5000m")

Table of Contents

  1. Overview
  2. Installation
  3. Usage

Overview

The pipeline is composed of the following steps:

1. Data Preprocessing The data preprocessing converts the data into a document term matrix (DTM) and removes stopwords, punctuation, etc. which is the data format needed for the LDA model.

2. Model Training The model training step trains the LDA model on the DTM with a number of iterations and predefined amount of topics.

3. Model Inference The model inference step uses the trained LDA model to infer the topic term distribution of the documents.

4. Statistical Analysis The analysis includes the methods like linear regression, binary regression, ridge regression or correlation to analyze the relationship between the topics and the prediction variable. It is possible to control for a number of variables and to adjust the p-value for multiple comparisons.

5. Visualization The visualization step creates wordclouds of the significant topics found by the statistical analysis.

Usage

In an example where the topics are used to predict the PHQ-9 score, the pipeline can be run as follows:

1. Data Preprocessing To preprocess the data, run the following command:

data <- read.csv("data.csv")
dtm <- topicsDtm(data = data$text)

# Check the results from the dtm and refine stopwords and removal rates if necessary
topicsDtmEval(dtm)

2. Model Training To train the LDA model, run the following command:

model <- topicsModel(dtm = dtm,
                  num_topics = 20,
                  num_iterations = 1000)
                  

3. Model Inference To infer the topic term distribution of the documents, run the following command:

preds <- topicsPreds(model = model,
                    data = data$text)

4. Statistical Analysis To analyze the relationship between the topics and the prediction variable, run the following command:

test <- topicsTest(model = model,
                  data = data,
                  preds = preds,
                  pred_var = "phq9",
                  control_vars = c("age",..),
                  test_method = "linear_regression")

5. Visualization To visualize the significant topics as wordclouds, run the following command:

topicsPlots(model = model,
            test = test)

Copy Link

Version

Install

install.packages('topics')

Version

0.21.0

License

GPL-3

Maintainer

Last Published

November 23rd, 2024

Functions in topics (0.21.0)

dep_wor_data

Example data about mental health descirptions .
topicsDtmEval

Summarize and Visualize your Document Term Matrix
topicsDtm

Document Term Matrix
topicsGrams

N-grams
topicsPreds

Predict topic distributions
topicsTest

Statistically test topics
topicsModel

Topic modelling
topicsPlot

Plot word clouds