How shtosti/accelerated-drift works
This is a research pipeline for measuring stylistic changes in arXiv and medRxiv preprints around LLM adoption. Researchers invoke configured collection, preprocessing, analysis, and visualization commands; the result is statistical tables and publication figures, with topic-level and cross-topic comparisons as additional analysis branches. The graph follows documented stages and source-index references. Some internal implementation and external API wiring was not sampled, so uncertain links are omitted.
Main entry points: analyze.py Corpus analysis · cli.py Command interface · collect.py Record collection · visualize.py Figure generation
Components 26
Runtime control
- Command interface
src/not_an_llm/cli.py - Run configuration
src/not_an_llm/config.py
Corpus preparation
- Record collection
src/not_an_llm/pipelines/collect.py - arXiv client
src/not_an_llm/clients/arxiv.py - medRxiv client
src/not_an_llm/clients/medarxiv.py - bioRxiv client
src/not_an_llm/clients/biorxiv.py - Semantic Scholar client
src/not_an_llm/clients/semantic_scholar.py - Paper records
data/raw - Text preprocessing
src/not_an_llm/pipelines/preprocess.py - Text normalization
src/not_an_llm/preprocessing/text.py - Prepared corpus
data/processed
Statistical analysis
- Corpus analysis
src/not_an_llm/pipelines/analyze.py - Feature extraction
src/not_an_llm/analysis/feature_extractor.py - Readability metrics
src/not_an_llm/analysis/readability.py - Feature families
src/not_an_llm/analysis/feature_groups.py - Trend aggregation
src/not_an_llm/analysis/trends.py - Interrupted time series
src/not_an_llm/analysis/interrupted_time_series.py - Topic modeling
src/not_an_llm/analysis/topic_modeling/model.py - Topic reporting
src/not_an_llm/analysis/topic_modeling/pipeline.py - Topic comparison
src/not_an_llm/analysis/topic_modeling/comparison.py
Reports and outputs
- Analysis tables
data/analysis - Figure generation
src/not_an_llm/pipelines/visualize.py - Rendered figures
data/visuals
Other components
- Researcher
- Preprint APIs
- Embedding models
Connections 38
- Researcher → Command interface: runs commands
README.md - Command interface → Run configuration: loads settings
src/not_an_llm/cli.py - Command interface → Record collection: dispatches collection
src/not_an_llm/cli.py - Command interface → Text preprocessing: dispatches preprocessing
src/not_an_llm/cli.py - Command interface → Corpus analysis: dispatches analysis
src/not_an_llm/cli.py - Command interface → Figure generation: dispatches visualization
src/not_an_llm/cli.py - Command interface → Topic comparison: dispatches comparison
src/not_an_llm/cli.py - Record collection → Run configuration: reads settings
src/not_an_llm/pipelines/collect.py - Record collection → arXiv client: queries
src/not_an_llm/pipelines/collect.py - Record collection → medRxiv client: queries
src/not_an_llm/pipelines/collect.py - Record collection → bioRxiv client: queries
src/not_an_llm/pipelines/collect.py - Record collection → Semantic Scholar client: queries
src/not_an_llm/pipelines/collect.py - arXiv client → Preprint APIs: requests records
- medRxiv client → Preprint APIs: requests records
- Record collection → Paper records: writes records
README.md - Text preprocessing → Run configuration: reads settings
src/not_an_llm/pipelines/preprocess.py - Text preprocessing → Text normalization: normalizes text
src/not_an_llm/pipelines/preprocess.py - Text preprocessing → Paper records: reads records
README.md - Text preprocessing → Prepared corpus: writes corpus
README.md - Corpus analysis → Run configuration: reads settings
src/not_an_llm/pipelines/analyze.py - Corpus analysis → Prepared corpus: reads corpus
README.md - Corpus analysis → Feature extraction: extracts features
src/not_an_llm/pipelines/analyze.py - Corpus analysis → Readability metrics: computes metrics
src/not_an_llm/pipelines/analyze.py - Corpus analysis → Feature families: uses feature families
src/not_an_llm/pipelines/analyze.py - Corpus analysis → Trend aggregation: aggregates trends
src/not_an_llm/pipelines/analyze.py - Corpus analysis → Interrupted time series: fits models
README.md - Corpus analysis → Topic reporting: runs topic analysis
src/not_an_llm/pipelines/analyze.py - Topic reporting → Topic modeling: assigns topics
src/not_an_llm/analysis/topic_modeling/pipeline.py - Topic modeling → Embedding models: encodes documents
README.md - Corpus analysis → Analysis tables: writes statistics
README.md - Topic reporting → Analysis tables: writes topic reports
src/not_an_llm/analysis/topic_modeling/pipeline.py - Topic comparison → Interrupted time series: computes ITS
src/not_an_llm/analysis/topic_modeling/comparison.py - Topic comparison → Analysis tables: reads comparison inputs
- Figure generation → Run configuration: reads settings
src/not_an_llm/pipelines/visualize.py - Figure generation → Analysis tables: reads tables
README.md - Figure generation → Trend aggregation: plots trends
src/not_an_llm/pipelines/visualize.py - Figure generation → Interrupted time series: plots ITS results
src/not_an_llm/pipelines/visualize.py - Figure generation → Rendered figures: writes figures
README.md