tanziro/bangla-crime-investigation-patterns-v2

Bangla Crime Investigation Patterns V2

This dataset is an anonymized, structured extraction from 1,078 public Bangla crime-investigation video transcripts. It was built to support machine-learning and LLM research on crime-pattern information extraction, case summarization, event sequencing, entity/relationship extraction, and Bengali investigative-report analysis.

The public release does not include raw full transcripts. It contains short redacted evidence snippets linked to extracted facts, plus bilingual Bangla/English fields where possible.

Dataset Structure

The dataset is published as six related table configs:

Each table is available as JSONL and Parquet. Hugging Face Dataset Viewer configs use the split Parquet files.

Source And Processing

Source material came from public Bangla investigation-video transcripts from:

• —Channel 24 Search Light
• —Independent TV / Taalash
• —Jamuna TV Investigation 360
• —Somoy TV investigation programming
• —Star News Investigation

The transcripts were processed locally into a privacy-preserving structured dataset. V2 applies stricter context review than the earlier rules-heavy v1, reducing noisy substring matches and weak primary crime labels while preserving useful heterogeneous facts.

Privacy And Safety

• —Raw full transcripts are excluded from the public dataset.
• —Direct identifiers are redacted or replaced with stable anonymous role IDs.
• —Exact GPS coordinates are not inferred.
• —Evidence spans are short, redacted snippets only.
• —The dataset is intended for research and analysis, not for identifying real people or making legal determinations.
• —Private local audit artifacts such as review_packets/ are intentionally excluded from the Hugging Face upload.

Intended Uses

Appropriate uses include:

• —Bangla/English crime-pattern information extraction research
• —LLM-readable structured case analysis
• —Event, entity, and relationship extraction benchmarking
• —Bengali ASR-noise-tolerant downstream NLP experiments
• —Dataset-card and schema examples for privacy-preserving media transcript processing

Out-of-scope uses include:

• —Identifying private individuals
• —Predictive policing or automated law-enforcement decisions
• —Generating operational criminal instructions
• —Treating extracted records as verified legal facts

Validation

The final local validation passed:

• —cases: 1,078
• —observations: 13,133
• —entities: 7,717
• —events: 3,234
• —relationships: 2,411
• —evidence_spans: 19,521
• —Privacy leak scan: {}
• —Transcript count matched case count: 1,078

Citation

If you use this dataset, cite it as:

@dataset{bangla_crime_investigation_patterns_v2,
  title = {Bangla Crime Investigation Patterns V2},
  author = {Tanzir Rahman},
  year = {2026},
  publisher = {Hugging Face},
  note = {Anonymized structured dataset derived from public Bangla investigation transcripts}
}