text

text provides analysis and visualization of human language as an R package employing natural language processing and deep learning with transformer models for psychological and social-science research.

Key Features:

• Modular and end-to-end access: Provides both modular access to transformer-based language models and end-to-end analytical functions for complete workflows.
• Text embedding (textEmbed()): Transforms text into transformer-based word embeddings for downstream analyses.
• Predictive modeling (textTrain(), textPredict()): Trains predictive models using embeddings as input and applies trained models to new data for prediction.
• Semantic comparison (textSimilarity(), textDistance()): Computes semantic similarity and distance scores between texts.
• Projection and visualization (textProjection(), textProjectionPlot()): Projects texts into embedding space and generates visualizations of those projections.
• Centrality analysis (textCentrality(), textCentralityPlot()): Computes and visualizes centrality measures of texts within embedding spaces.

Scientific Applications:

• Psychological language analysis: Analyzes psychological information embedded in language to study human behavior and cognition.
• Social-science research: Supports analysis of language data in social-science investigations.
• Scalable study sizes: Applicable to both small-scale studies and analyses of large datasets.

Methodology:

Uses textEmbed() to generate transformer-based embeddings; uses textTrain() and textPredict() for training and applying predictive models on embeddings; uses textSimilarity() and textDistance() to compute semantic similarity/distance; uses textProjection() and textProjectionPlot() for embedding-space projection and visualization; and uses textCentrality() and textCentralityPlot() for computing and visualizing centrality measures.