Clone from jonathanagustin/courtlistener-bulk

.claude/hooks/session-start.sh

@@ -0,0 +1,71 @@
+#!/bin/bash
+set -euo pipefail
+
+# Only run in Claude Code on the Web (remote environment)
+if [ "${CLAUDE_CODE_REMOTE:-}" != "true" ]; then
+  echo "Skipping session-start hook (not running in Claude Code on the Web)"
+  exit 0
+fi
+
+echo "========================================"
+echo "CourtListener Bulk Sync - Session Start"
+echo "========================================"
+echo ""
+
+# Navigate to project root
+cd "${CLAUDE_PROJECT_DIR}"
+
+# Install dependencies with uv
+echo "Installing dependencies..."
+if command -v uv >/dev/null 2>&1; then
+  uv sync 2>&1 || echo "Warning: uv sync had issues"
+else
+  echo "Warning: uv not found. Install with: curl -LsSf https://astral.sh/uv/install.sh | sh"
+fi
+
+# Check for HF_TOKEN
+if [ -z "${HF_TOKEN:-}" ]; then
+  echo ""
+  echo "WARNING: HF_TOKEN not set!"
+  echo "Set it with: export HF_TOKEN=<your-token>"
+  echo ""
+fi
+
+# Show sync status
+echo ""
+echo "Checking sync state..."
+if [ -f "sync_state.json" ]; then
+  echo "Sync state found. Currently synced tables:"
+  python3 -c "
+import json
+try:
+    with open('sync_state.json') as f:
+        state = json.load(f)
+    for table, info in state.get('tables', {}).items():
+        rows = info.get('rows', 'unknown')
+        synced = info.get('synced_at', 'unknown')[:10] if info.get('synced_at') else 'unknown'
+        print(f'  - {table}: {rows} rows (synced {synced})')
+except Exception as e:
+    print(f'  Error reading state: {e}')
+" 2>/dev/null || echo "  (unable to parse sync_state.json)"
+else
+  echo "No sync_state.json found - fresh start."
+fi
+
+# Show available tables
+echo ""
+echo "Available tables (by size):"
+echo "  Small:  courts, people, financial-disclosures"
+echo "  Medium: citations, parentheticals, citation-map"
+echo "  Large:  clusters, dockets, opinions"
+
+echo ""
+echo "========================================"
+echo "Ready to sync!"
+echo "========================================"
+echo ""
+echo "Commands:"
+echo "  uv run python sync.py --dry-run              # See what will sync"
+echo "  uv run python sync.py --tables courts        # Sync one table"
+echo "  uv run python sync.py --xet-high-perf        # Sync all (fast mode)"
+echo ""

.gitattributes

@@ -1,59 +1,3 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.lz4 filter=lfs diff=lfs merge=lfs -text
-*.mds filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
+# Xet storage for large data files
 *.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
-# Audio files - uncompressed
-*.pcm filter=lfs diff=lfs merge=lfs -text
-*.sam filter=lfs diff=lfs merge=lfs -text
-*.raw filter=lfs diff=lfs merge=lfs -text
-# Audio files - compressed
-*.aac filter=lfs diff=lfs merge=lfs -text
-*.flac filter=lfs diff=lfs merge=lfs -text
-*.mp3 filter=lfs diff=lfs merge=lfs -text
-*.ogg filter=lfs diff=lfs merge=lfs -text
-*.wav filter=lfs diff=lfs merge=lfs -text
-# Image files - uncompressed
-*.bmp filter=lfs diff=lfs merge=lfs -text
-*.gif filter=lfs diff=lfs merge=lfs -text
-*.png filter=lfs diff=lfs merge=lfs -text
-*.tiff filter=lfs diff=lfs merge=lfs -text
-# Image files - compressed
-*.jpg filter=lfs diff=lfs merge=lfs -text
-*.jpeg filter=lfs diff=lfs merge=lfs -text
-*.webp filter=lfs diff=lfs merge=lfs -text
-# Video files - compressed
-*.mp4 filter=lfs diff=lfs merge=lfs -text
-*.webm filter=lfs diff=lfs merge=lfs -text
+data/**/*.parquet filter=lfs diff=lfs merge=lfs -text

CLAUDE.md

@@ -0,0 +1,225 @@
+# CLAUDE.md
+
+Instructions for Claude Code to continue syncing CourtListener bulk data.
+
+## Project Overview
+
+This is a HuggingFace dataset repository that mirrors CourtListener's bulk legal data as Parquet files. The project has two main scripts:
+
+1. **sync.py** - Downloads CSV data from CourtListener's S3 bucket, converts to Parquet with zstd compression, and uploads to HuggingFace with Xet storage.
+
+2. **embed.py** - Generates embeddings for legal documents using BGE-large via HuggingFace Inference API. Stores embeddings as Parquet files in `embeddings/`.
+
+## Quick Start
+
+```bash
+# Install dependencies
+uv sync
+
+# Set HuggingFace token
+export HF_TOKEN=<your-token>
+
+# Run sync (dry-run first to see what will sync)
+uv run python sync.py --dry-run
+
+# Sync specific tables (start with smaller ones)
+uv run python sync.py --tables courts,people,financial-disclosures
+
+# Sync with high-performance Xet mode
+uv run python sync.py --xet-high-perf --tables citations
+```
+
+## Tables by Size (smallest first)
+
+| Table | Compressed Size | Priority |
+|-------|----------------|----------|
+| courts | ~81 KB | 1 |
+| people | ~455 KB | 2 |
+| financial-disclosures | ~516 KB | 3 |
+| citations | ~127 MB | 4 |
+| parentheticals | ~271 MB | 5 |
+| citation-map | ~515 MB | 6 |
+| clusters | ~2.4 GB | 7 |
+| dockets | ~4.9 GB | 8 |
+| opinions | ~53.6 GB | 9 |
+
+## Sync State
+
+The script maintains `sync_state.json` to track what has been synced. Each table entry includes:
+- `source_key`: S3 key of the synced file
+- `etag`: S3 ETag for change detection
+- `synced_at`: Timestamp of last sync
+- `rows`: Number of rows in the table
+
+The script is fully idempotent - running it again will skip already-synced tables unless `--force` is used.
+
+## Disk Space Requirements
+
+The script processes one table at a time and cleans up after each:
+1. Downloads compressed CSV to temp dir
+2. Decompresses and converts to Parquet
+3. Uploads to HuggingFace
+4. Deletes local files
+
+Minimum disk space needed: ~2x the largest table you're syncing.
+- For citations (127MB): ~300MB free
+- For opinions (53.6GB): ~120GB free
+
+## Environment Variables
+
+| Variable | Required | Default | Description |
+|----------|----------|---------|-------------|
+| `HF_TOKEN` | Yes | - | HuggingFace token with write access |
+| `HF_XET_HIGH_PERFORMANCE` | No | off | Saturate network + all CPU cores |
+| `HF_XET_NUM_CONCURRENT_RANGE_GETS` | No | 16 | Parallel downloads per file |
+| `HF_XET_CHUNK_CACHE_SIZE_BYTES` | No | 0 | Chunk cache for cross-file dedup |
+| `HF_XET_SHARD_CACHE_SIZE_LIMIT` | No | 4GB | Shard cache to avoid re-uploads |
+| `HF_XET_RECONSTRUCT_WRITE_SEQUENTIALLY` | No | off | Sequential writes for HDD |
+| `HF_XET_CACHE` | No | ~/.cache/huggingface/xet | Xet cache directory |
+
+## CLI Options
+
+```
+--tables TABLE1,TABLE2   Sync specific tables (comma-separated)
+--dry-run                Show what would sync without doing it
+--force                  Re-sync even if already complete
+--list                   List available tables and exit
+--work-dir PATH          Custom temp directory
+
+Xet Performance Options:
+--xet-high-perf          Max network/CPU utilization
+--xet-concurrent N       Parallel downloads (default 16, try 32-64)
+--xet-chunk-cache SIZE   Enable chunk cache (e.g., "10GB") for dedup
+--xet-hdd                Sequential writes for spinning disks
+```
+
+## Performance Tuning Examples
+
+```bash
+# Fast network, SSD storage (recommended for cloud VMs)
+uv run python sync.py --xet-high-perf --xet-concurrent 64
+
+# Enable chunk cache for better dedup across legal datasets
+uv run python sync.py --xet-chunk-cache 10GB --tables citations,parentheticals
+
+# Slow network or HDD storage
+uv run python sync.py --xet-hdd --xet-concurrent 8
+
+# Maximum performance (fast network + SSD + lots of RAM for cache)
+uv run python sync.py --xet-high-perf --xet-concurrent 64 --xet-chunk-cache 20GB
+```
+
+## Resuming a Sync
+
+If a sync is interrupted:
+1. The script automatically resumes from where it left off
+2. Completed tables are tracked in `sync_state.json`
+3. Partially uploaded files are not tracked (will re-sync)
+
+To check current state:
+```bash
+cat sync_state.json
+```
+
+## Source Data
+
+- **S3 Bucket**: `com-courtlistener-storage`
+- **URL Pattern**: `https://com-courtlistener-storage.s3.amazonaws.com/bulk-data/{table}-{date}.csv.bz2`
+- **Update Frequency**: Daily (files are dated)
+
+## HuggingFace Repository
+
+- **Repo ID**: `jonathanagustin/courtlistener-bulk`
+- **Visibility**: Private
+- **Storage**: Xet (chunk-level deduplication)
+
+---
+
+## Embeddings
+
+### Embedding Model
+
+The project uses **BAAI/bge-large-en-v1.5** for generating embeddings:
+- **Dimensions**: 1024
+- **Context window**: 512 tokens
+- **License**: MIT (commercially usable)
+- **Provider**: BAAI (Beijing Academy of AI)
+
+### Quick Start (Embeddings)
+
+```bash
+# First sync the source data
+uv run python sync.py --tables opinions
+
+# Generate embeddings (dry-run first)
+uv run python embed.py --dry-run
+
+# Embed specific tables
+uv run python embed.py --tables opinions,clusters
+
+# Custom batch size (lower if hitting rate limits)
+uv run python embed.py --batch-size 16 --tables parentheticals
+```
+
+### Embedding Tables
+
+| Table | Source | Text Columns | Priority |
+|-------|--------|--------------|----------|
+| opinions | data/opinions.parquet | plain_text, html_lawbox, html_columbia, html | 1 |
+| clusters | data/clusters.parquet | case_name, syllabus, attorneys, posture | 2 |
+| parentheticals | data/parentheticals.parquet | text | 3 |
+| dockets | data/dockets.parquet | case_name, cause, nature_of_suit | 4 |
+
+### Embedding State
+
+The script maintains `embed_state.json` to track progress:
+- `rows_embedded`: Number of rows processed
+- `source_rows`: Total rows in source table
+- `model`: Model used for embeddings
+- `status`: complete/in_progress/failed
+
+Embedding is resumable - if interrupted, it picks up from the last checkpoint.
+
+### Output Format
+
+Embeddings are stored as Parquet files in `embeddings/`:
+```
+embeddings/
+├── opinions.parquet      # id (int), embedding (list<float32>)
+├── clusters.parquet
+├── parentheticals.parquet
+└── dockets.parquet
+```
+
+Each row contains:
+- `id`: The source document ID
+- `embedding`: 1024-dimensional float32 vector
+
+### Embedding CLI Options
+
+```
+--tables TABLE1,TABLE2   Embed specific tables (comma-separated)
+--dry-run                Show what would embed without doing it
+--force                  Re-embed even if already complete
+--batch-size N           Texts per API call (default: 32)
+--model MODEL            Override embedding model
+--rate-limit N           Max requests per minute (default: 30, 0=unlimited)
+--min-interval N         Minimum seconds between requests (default: 0)
+--list                   List available tables and exit
+```
+
+### Rate Limits
+
+The HF Inference API has rate limits. If you hit them:
+1. Reduce batch size: `--batch-size 16` or `--batch-size 8`
+2. The script automatically retries with exponential backoff
+3. Progress is saved periodically, so interruptions are safe
+
+### Use Cases
+
+Once embeddings are generated, they enable:
+- **Semantic search**: Find cases by meaning, not just keywords
+- **Similar case discovery**: Find related precedents
+- **Citation recommendation**: Suggest relevant citations
+- **Clustering**: Group similar legal documents
+- **Classification**: Categorize documents by topic

README.md

@@ -1,50 +1,39 @@
 ---
-license: cc-by-4.0
+license: pddl
 task_categories:
-  - text-classification
+  - text-generation
   - question-answering
 language:
   - en
 tags:
   - legal
-  - law
   - court-opinions
+  - case-law
   - courtlistener
-  - free-law-project
+  - pacer
+  - recap
 size_categories:
-  - 10M<n<100M
-configs:
-  - config_name: courts
-    data_files:
-      - split: train
-        path: data/courts/*.parquet
+  - 1M<n<10M
 ---
 
 # CourtListener Bulk Data
 
-This dataset contains bulk data from [CourtListener](https://www.courtlistener.com/),
-maintained by the [Free Law Project](https://free.law/).
+Private mirror of [CourtListener](https://www.courtlistener.com/) bulk data, converted to Parquet format for efficient querying and ML workflows.
 
-## Dataset Description
+## Data Sources
 
-CourtListener is a free legal research website containing millions of legal opinions
-from federal and state courts. This dataset provides structured access to:
+This dataset syncs from CourtListener's official bulk data exports:
 
-- **Courts**: Metadata about courts in the US legal system
-- **Clusters**: Opinion clusters (cases) with metadata
-- **Dockets**: Case docket information
-- **Opinions**: Full text of judicial opinions
-- **Citations**: Citation relationships between opinions
-- **People**: Judge and attorney information
-- **Financial Disclosures**: Judicial financial disclosure data
-
-## Statistics
-
-| Table | Rows | Files | Size (MB) |
-|-------|------|-------|-----------|
-| courts | 3,355 | 1 | 0.2 | ✅
-
-**Total**: 3,355 rows, 0.2 MB
+| Table | Description | Source |
+|-------|-------------|--------|
+| `courts` | Court metadata (700+ courts) | Core reference table |
+| `clusters` | Opinion clusters with case metadata | Case law groupings |
+| `opinions` | Full opinion text | Largest table (~9M decisions) |
+| `citations` | Citation graph between opinions | Network analysis |
+| `parentheticals` | Court-written case summaries | Compact case descriptions |
+| `people` | Judge information | 16,000+ judges |
+| `financial_disclosures` | Judge financial disclosures | 1.7M investment records |
+| `dockets` | RECAP docket metadata | Federal court filings |
 
 ## Usage
 
@@ -52,51 +41,57 @@ from federal and state courts. This dataset provides structured access to:
 from datasets import load_dataset
 
 # Load a specific table
-courts = load_dataset("drengskapur/courtlistener", "courts", split="train")
-clusters = load_dataset("drengskapur/courtlistener", "clusters", split="train")
-
-# Query via DuckDB (recommended for large tables)
-import duckdb
-conn = duckdb.connect()
-conn.execute("INSTALL httpfs; LOAD httpfs;")
-result = conn.execute("""
-    SELECT * FROM 'hf://datasets/drengskapur/courtlistener/data/clusters/*.parquet'
-    WHERE case_name ILIKE '%miranda%'
-    LIMIT 10
-""").fetchdf()
+courts = load_dataset("jonathanagustin/courtlistener-bulk", data_files="data/courts.parquet")
+opinions = load_dataset("jonathanagustin/courtlistener-bulk", data_files="data/opinions/*.parquet")
+
+# Stream large tables
+from datasets import load_dataset
+opinions = load_dataset(
+    "jonathanagustin/courtlistener-bulk",
+    data_files="data/opinions/*.parquet",
+    streaming=True
+)
+for opinion in opinions["train"]:
+    print(opinion["case_name"])
 ```
 
-## HuggingFace Datasets Server API
+## Sync Script
 
-Query directly via HTTP without downloading:
+This dataset includes an idempotent sync script that can be run anywhere:
 
 ```bash
-# Get rows
-curl "https://datasets-server.huggingface.co/rows?dataset=drengskapur/courtlistener&config=courts&split=train&length=10"
+# Sync all tables
+python sync.py
 
-# Search
-curl "https://datasets-server.huggingface.co/search?dataset=drengskapur/courtlistener&config=clusters&split=train&query=miranda"
+# Sync specific tables only
+python sync.py --tables courts,citations
+
+# Check what would be synced (dry run)
+python sync.py --dry-run
 ```
 
-## Data Source
+The script:
+- Downloads from CourtListener bulk data API
+- Converts CSV to Parquet with optimal compression
+- Uploads to HuggingFace with Xet deduplication
+- Tracks sync state to resume interrupted runs
+- Cleans up local files after upload to minimize disk usage
+
+## Storage
 
-Data is sourced from CourtListener's public S3 bulk data exports:
-`s3://com-courtlistener-storage/bulk-data/`
+This repository uses [Xet storage](https://huggingface.co/docs/hub/en/xet) for:
+- 10x faster uploads/downloads
+- Chunk-level deduplication (efficient incremental updates)
+- Support for files >1TB
 
-Updated daily via automated sync.
+## Update Frequency
+
+CourtListener updates their bulk data regularly. Run `sync.py` to pull latest changes.
 
 ## License
 
-This data is provided under the [CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/) license.
-Original data is from CourtListener / Free Law Project.
+Public Domain Dedication and License (PDDL) - same as CourtListener source data.
 
-## Citation
+## Attribution
 
-```bibtex
-@misc{courtlistener,
-  title={CourtListener},
-  author={Free Law Project},
-  year={2024},
-  url={https://www.courtlistener.com/}
-}
-```
+Data provided by [Free Law Project](https://free.law/) via [CourtListener](https://www.courtlistener.com/).

data/citation-map/shard_00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7d347e7fb4a888d32fa52ae94dfb3b34e8cf4f14c2f1d2b0660e9587c9fa9da
+size 4031405

data/citation-map/shard_00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd1f7aa8109c020bb9430b5d1ca6d3e650234050fce1c0bb0396018f4ae8664b
+size 4049755

data/citation-map/shard_00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce2d769eae89c75b53e9bcdfe57171e94cbb9523c8afac5e783981d206bae660
+size 4003163

data/citation-map/shard_00003.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20bc53522040bd4945574953e20c5a8553bf6186c013f16def62fbde35f0f317
+size 4039647

data/citation-map/shard_00004.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:153c33d7272ae91352e09df0404928358e3de104cbe6c4876ae4feabcad5b02c
+size 4056931

data/citation-map/shard_00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d58271348aabe1895a0fe2af3b174d2b3fc7f4da9fefbdff0ab0635a6e9821f
+size 4046229

data/citation-map/shard_00006.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90469c630b5e238190cd872fd2027fdd5ab5868d4a9ca8cc3bce18613f3b796f
+size 3990822

data/citation-map/shard_00007.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff11c6c402fe2f6a0ea2214ee45723fcd6c7de85c22bb975d5231f4116f9ed5c
+size 3955459

data/citation-map/shard_00008.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a769ce334b85e9a33499e3d7ab1a13fe9e87ffbdb9204df8b881a295bcc117cf
+size 4020589

data/citation-map/shard_00009.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b5020497ab004f71d8c1986196d1fc097408452ea292c48bfd42dafc6f0abac
+size 4014322

data/citation-map/shard_00010.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:016273e56734fc457599b45aff868818e1aaddadd84f7e0d49e5e84fe02c9ac2
+size 4023483

data/citation-map/shard_00011.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8d46286054a0258bd8c30b57357312467a542ca4590dd9bdf0de3c3c8c81001
+size 4006521

data/citation-map/shard_00012.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b22658fa0f21a045b9ffe45ac068140b0d0936943dd9e2db657815004a0f128
+size 4048528

data/citation-map/shard_00013.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efd96d023b58f550ffba635b427cd86f0e8684afb3fc49ffdcaf9ad8b4cef0cb
+size 4011946

data/citation-map/shard_00014.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168d9a1d43fd96645b113d69a8d36a5be0c80391891741f4f5a20cc897edddc6
+size 3990585

data/citation-map/shard_00015.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edc4975e5d3c65d7088ed4f459fe13458b2ba9f68e02afdcee5aaedf0ad6ceeb
+size 3956004

data/citation-map/shard_00016.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3553176758593ce938e462b0107cfec6ef8e5e7628ef45a7ebe58bace156890
+size 4019564

data/citation-map/shard_00017.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba32e44adb7d13d4dedfda8f3806dc276d26250532c28b34c9c188b789ff4a85
+size 4006936

data/citation-map/shard_00018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d87ebd034ea65d93717d5dec1de50d674ec0fc2e2bf0b111b244e96e85863484
+size 4002442

data/citation-map/shard_00019.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cce0458a41c4fc04d5136395380189b0da7956801d95449d32ccbc6894592f2
+size 3954754

data/citation-map/shard_00020.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78dd3c1b4e546e2cdb3a0aecf1958cf19dd0a2a6b5849652ee10f0e199cc03fb
+size 3923510

data/citation-map/shard_00021.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7101661b52dc6987a083205cdb47f290e5bf9980d597d1d7dbe291aecfacc988
+size 3883211

data/citation-map/shard_00022.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba7b9fbbc5966818c45fc61e8b4c311e42624aa4185416cd2917024da9afe40d
+size 3846976

data/citation-map/shard_00023.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:526521600841b720a5dda6185da0e9f870b7b5f4efe43ce0c329b4b46741167a
+size 3736283

data/citation-map/shard_00024.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e75af009db598f8d6fb63571b44fded3c7381cf1f2e46f0b20d2300f4a6cefa
+size 3742760

data/citation-map/shard_00025.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fce57a7a7946b3a376a3458e3e1a7ce5240d3975f072c5f1879a4dd3c36625a
+size 3825918

data/citation-map/shard_00026.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f03185ab848bacb4a3470f401601e4a357405d78d03034de69d1faa3abf1af6
+size 3806457

data/citation-map/shard_00027.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2b7ac6f5db73e1df0a2a3995cd4d8fbf4bd3de1a01618b2f79a10540a21f3fd
+size 3818699

data/citation-map/shard_00028.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5a1d7cba4f8a2cd14b7025043be4bad421db1beadd39255b0980217a9b239cd
+size 3729960

data/citation-map/shard_00029.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97a3e1b192ea45c2e22bb0a60000c4084424cb1da733601aa6eaa9c6dd9ff84f
+size 4055647

data/citation-map/shard_00030.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31b044421f4f52281efddc0f91834799b0a21831e39017622af76c595080c16a
+size 3666082

data/citation-map/shard_00031.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebc2350fef433e5333608949d97f78e600a3077955319d848056fdc8826a99cf
+size 4179088

data/citation-map/shard_00032.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22af90507e2c40656ace3c05ef74d14cf219c6dc153d9b32f68a22193fc00a0e
+size 3875198

data/citation-map/shard_00033.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7692282c0e6808c130ee4e25ef8d4267451dd9ac60ba66e198fccb82b0517922
+size 3851605

data/citation-map/shard_00034.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:424f29c25b3e7ae510ca18a5c1658691cfba6a23df3f7238d0ab695a44d0175a
+size 3851584

data/citation-map/shard_00035.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:447d348231d865982230f1e2227ff031c62064cf77724ac9c33f9c977deed459
+size 3815527

data/citation-map/shard_00036.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdf4134d1cf61289ea2636b9d22e01cb61e1d340ae0b0536f1a942f94d76fec9
+size 3811038

data/citation-map/shard_00037.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3497158d2afae75573941470ae5dc6315c8ec7f398a56fa2a07b08b7dd54b46
+size 3872711

data/citation-map/shard_00038.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d35c19f02db16e5f4ec52d7d9aff173d8deff1caf7d0e90a25fbb3aab21c3347
+size 3874715

data/citation-map/shard_00039.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f0c37960bddd6b0f64da08a235b1e4aeba1afb1826f87edf6527bf0f1870070
+size 3961661

data/citation-map/shard_00040.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da705471e7f6d8852fc46056bddd2ef3b2a72669d7f31913b92e5461d16723eb
+size 3992810

data/citation-map/shard_00041.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c593139a3920683708f453e0a9b97594af18805dd2c9b58a43a67e3fb7cc6ca5
+size 3969148

data/citation-map/shard_00042.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30556304cc5f72991e4aae5f9f81d9abd9fa557b0a74cee48ea0633975a7dcba
+size 3917610

data/citation-map/shard_00043.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfdf7cd639e113d61ee8013a41ebd9335c528bf1da2efebe322e2cea2d184417
+size 3864201

data/citation-map/shard_00044.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d08b6683394990bb13dc74bcaea50da166be455573226bc289a3d42e0a26da7d
+size 3722499

data/citation-map/shard_00045.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d27e22eb06d29c9c31c2e5df7f8f90547bf8bd59e568281e659871d547137a0e
+size 3873227