config.yaml

project:
  name: "traffic-incident-summarization"
  random_seed: 42

paths:
  raw_csv: "data/raw/US_Accidents_March23.csv"
  cleaned_csv: "data/interim/cleaned_incidents.csv"
  experiment_sample_csv: "data/interim/experiment_sample.csv"
  eval_candidates_csv: "data/interim/eval_candidates.csv"
  eval_set_csv: "data/processed/eval_set_with_refs.csv"
  model_outputs_csv: "data/processed/model_outputs.csv"
  aggregate_metrics_csv: "data/processed/aggregate_metrics.csv"
  gcc_dir: "data/raw/gcc"
  gcc_manifest_csv: "data/raw/gcc/source_manifest.csv"
  gcc_combined_structured_csv: "data/interim/gcc_structured_combined.csv"
  gcc_narratives_csv: "data/interim/gcc_narratives.csv"
  combined_corpus_csv: "data/interim/combined_incident_corpus.csv"

dataset:
  kaggle_handle: "sobhanmoosavi/us-accidents"
  expected_filename: "US_Accidents_March23.csv"
  auto_download: true

gcc:
  enabled: true
  include_in_combined_corpus: true
  narrative_rows_per_source: 250
  default_country: "United Arab Emirates"
  source_priority:
    - "dubai_pulse_incidents"
    - "uae_federal_traffic_stats"
    - "abu_dhabi_open_data"
  sources:
    dubai_pulse_incidents:
      label: "Dubai Pulse Traffic Incidents"
      official_url: "https://www.dubaipulse.gov.ae/data/dp-traffic/dp_traffic_incidents-open"
      access_type: "open-data-portal"
      local_sample_csv: "data/raw/gcc/dubai_pulse_incidents_sample.csv"
      status: "bundled_sample"
      notes: "Real-time Dubai Police incident feed surfaced through Dubai Pulse. The bundled CSV is a normalized local sample schema for immediate demo use."
    uae_federal_traffic_stats:
      label: "UAE Federal Traffic Statistics"
      official_url: "https://uaestat.fcsc.gov.ae/vis?df%5Bag%5D=FCSA&df%5Bds%5D=FCSC-RDS&df%5Bid%5D=DF_TRA_TYPE&df%5Bvs%5D=3.0.0"
      access_type: "web-statistics-portal"
      local_sample_csv: "data/raw/gcc/uae_federal_traffic_stats_sample.csv"
      status: "bundled_sample"
      notes: "Federal accident indicators by emirate and accident type. The bundled CSV is a narrative-ready extracted sample schema."
    abu_dhabi_open_data:
      label: "Abu Dhabi Open Data"
      official_url: "https://data.abudhabi/opendata/dataset"
      access_type: "open-data-catalog"
      local_sample_csv: "data/raw/gcc/abu_dhabi_incidents_sample.csv"
      status: "bundled_sample"
      notes: "Abu Dhabi open data catalog entry point. The bundled CSV is a normalized road incident sample for regional coverage."

data:
  text_column: "Description"
  min_chars: 50
  max_chars: 1500
  experiment_sample_size: 3000
  eval_candidate_size: 300
  gcc_eval_candidate_size: 120
  deduplicate: true
  stratify_by:
    - "Severity"

generation:
  default_max_input_tokens: 512
  default_max_new_tokens: 72
  default_min_new_tokens: 18
  num_beams: 2
  length_penalty: 1.0
  no_repeat_ngram_size: 3
  early_stopping: true

models:
  bart_large_cnn:
    hf_name: "facebook/bart-large-cnn"
    enabled: true
    prompt_prefix: ""
    max_input_tokens: 512
  flan_t5_small:
    hf_name: "google/flan-t5-small"
    enabled: true
    prompt_prefix: "summarize: "
    max_input_tokens: 512
  pegasus_cnn:
    hf_name: "google/pegasus-cnn_dailymail"
    enabled: false
    prompt_prefix: ""
    max_input_tokens: 512