NEWS

---

TextAnalysisR 0.1.5.9000

• A codebook can be built from the texts instead of supplied.
• Sentiment analysis reports which words moved the score.

TextAnalysisR 0.1.5 (2026-09-13)

• Coded output tells a declined unit from a failed call.
• Confidence outside 0 to 1 is dropped, not clamped.
• Inter-coder agreement reports the unit count each statistic used.
• Coded assignments export as CSV or Excel.
• Added estimate_topic_effects() for STM prevalence effects.
• Fixed neural-classifier sentiment scoring.
• Prevented the hosted app from crashing on large datasets.

TextAnalysisR 0.1.4 (2026-07-27)

• New math_mode argument in prep_texts() keeps numbers, math operators, and symbols, and strips only sentence-end punctuation.
• First CRAN release.

TextAnalysisR 0.1.3

• Updated semantic analysis and topic modeling functions.
• Refreshed vignettes and pkgdown reference index.

TextAnalysisR 0.1.0

• Renamed fit_embedding_topics() to fit_embedding_model(). The old name is deprecated.
• Multi-format file import (PDF, DOCX, XLSX, CSV, TXT).
• Hybrid topic modeling (STM + BERTopic).
• Semantic similarity and document clustering.
• Lexical diversity metrics (TTR, MTLD).
• Log-odds ratio analysis and lexical dispersion plots.
• Neural sentiment analysis via transformers.
• spaCy support for POS tagging, NER, lemmatization, and dependency parsing.
• Sentence-transformer document embeddings.
• LLM integration (OpenAI, Gemini) for topic labeling and RAG.

TextAnalysisR 0.0.2

• Documentation improvements.

TextAnalysisR 0.0.1

• First development release.
• Text preprocessing, STM topic modeling, basic Shiny app.