conflibertR

An R package that brings ConfliBERT, a transformer model purpose-built for conflict and political violence text, into R. Run pretrained inference, benchmark models, fine-tune custom classifiers, and compare architectures, all from your R console.

GitHub

Installation

# Install from GitHub
devtools::install_github("shreyasmeher/conflibertR")

# Set up the Python backend (run once, then restart R)
library(conflibertR)
conflibert_install()

Named Entity Recognition

Extract persons, organizations, locations, weapons, and temporal expressions from conflict text.

ner_results <- conflibert_ner(
  "The United Nations deployed peacekeepers to eastern Congo
   after rebel forces attacked Goma on Tuesday."
)
ner_results

# Vectorized: pass multiple texts at once
ner_batch <- conflibert_ner(c(
  "NATO forces launched airstrikes near Tripoli in March 2024.",
  "President Zelenskyy met with General Syrskyi in Kyiv."
))

Binary Classification

Determine whether text describes a conflict event. Returns a label, confidence score, and per-class probabilities.

classify_results <- conflibert_classify(c(
  "Armed militants stormed a military outpost killing twelve soldiers.",
  "The European Central Bank held interest rates steady at 4.5 percent.",
  "Protesters clashed with riot police outside the parliament building.",
  "A new trade agreement was signed between Japan and Australia."
))
classify_results

Multilabel Event Classification

Score text against four event categories (Armed Assault, Bombing or Explosion, Kidnapping, and Other), each evaluated independently.

event_results <- conflibert_multilabel(c(
  "A car bomb exploded near the government ministry in Baghdad.",
  "Gunmen kidnapped three aid workers travelling through the Sahel region."
))
event_results

Question Answering

Extract answers directly from conflict-related passages.

context <- "On 15 September 2023, ethnic clashes erupted in Manipur, India.
  The Indian Army deployed two battalions to restore order.
  At least 60 people were killed and over 200 injured in the violence."

conflibert_qa(context, "How many people were killed?")
conflibert_qa(context, "Who was deployed to restore order?")
conflibert_qa(context, "Where did the clashes occur?")

Benchmarking

Evaluate the pretrained binary classifier against your own labeled data.

data <- conflibert_example("binary")
bench <- conflibert_benchmark(data$test$text, data$test$label)
bench

Fine-Tuning

Train a custom binary classifier on labeled data. The package ships with example datasets.

data <- conflibert_example("binary")

result <- conflibert_finetune(
  train  = data$train,
  dev    = data$dev,
  test   = data$test,
  model  = "ConfliBERT",
  task   = "binary",
  epochs = 3
)

cat("Accuracy:", result$metrics$accuracy, "\n")
cat("F1 Score:", result$metrics$f1, "\n")

Multiclass classification is also supported:

mc_data <- conflibert_example("multiclass")

mc_result <- conflibert_finetune(
  train  = mc_data$train,
  dev    = mc_data$dev,
  test   = mc_data$test,
  model  = "ConfliBERT",
  task   = "multiclass",
  epochs = 3
)

Model Comparison

Compare ConfliBERT against general-purpose transformers on the same task.

comparison <- conflibert_compare(
  train  = data$train,
  dev    = data$dev,
  test   = data$test,
  models = c("ConfliBERT", "BERT Base Uncased", "DistilBERT Base"),
  task   = "binary",
  epochs = 3
)
comparison

See all available base models: conflibert_models()

Function Reference

Function	Task
conflibert_ner()	Named entity recognition
conflibert_classify()	Binary conflict classification
conflibert_multilabel()	Multilabel event type scoring
conflibert_qa()	Extractive question answering
conflibert_benchmark()	Evaluate pretrained classifier
conflibert_finetune()	Train a custom classifier
conflibert_compare()	Compare multiple architectures
conflibert_example()	Load bundled example datasets
conflibert_models()	List available base models

Citation

@inproceedings{hu2022conflibert,
  title={ConfliBERT: A Pre-trained Language Model for Political Conflict and Violence},
  author={Hu, Yibo and Hosseini, MohammadSaleh and Parolin, Erick Skorupa
          and Osorio, Javier and Khan, Latifur and Brandt, Patrick and D'Orazio, Vito},
  booktitle={Proceedings of NAACL-HLT 2022},
  pages={5469--5482},
  year={2022}
}