FineType

FineType classifies strings into a rich taxonomy of 151 semantic types — each type is a transformation contract that guarantees a DuckDB cast expression will succeed.

$ finetype infer "192.168.1.1"
technology.internet.ip_v4

$ finetype infer "2024-01-15T10:30:00Z"
datetime.timestamp.iso_8601

$ finetype infer "[email protected]"
identity.person.email

Why FineType?

Go beyond primitive types. Detect the real meaning of your data.

151 Semantic Types

Classify across 6 domains: datetime, technology, identity, representation, geography, and container types.

Transformation Contracts

Every prediction maps to a DuckDB SQL expression guaranteed to parse successfully. Types you can act on.

Pure Rust, No Python

Built with Candle ML. 600+ classifications/sec, 8.5 MB memory, 66ms cold start. Ship without a runtime.

Taxonomy

FineType recognizes 151 types across 6 domains:

Domain	Types	Examples
`datetime`	46	ISO 8601, RFC 2822, Unix timestamps, timezones
`technology`	34	IPv4, IPv6, MAC addresses, URLs, UUIDs, hashes
`identity`	25	Names, emails, phone numbers, passwords
`representation`	19	Integers, floats, booleans, hex colors, base64, JSON
`geography`	16	Latitude, longitude, countries, cities, postal codes
`container`	11	JSON objects, CSV rows, query strings, key-value pairs

Label format: {domain}.{category}.{type} — e.g., technology.internet.ip_v4. Locale-specific types append a suffix: identity.person.phone_number.EN_AU.

CLI Usage

FineType provides 9 commands covering the full ML pipeline:

# Classify a single value
finetype infer -i "bc89:60a9:23b8:c1e9:3924:56de:3eb1:3b90"

# Classify from file (one value per line), JSON output
finetype infer -f data.txt --output json

# Column-mode inference (distribution-based disambiguation)
finetype infer -f column_values.txt --mode column

# Profile a CSV file — detect column types
finetype profile -f data.csv

# Generate synthetic training data
finetype generate --samples 1000 --output training.ndjson

# Train a CharCNN model
finetype train --data data/train.ndjson --epochs 10 --batch-size 64

# Evaluate model accuracy
finetype eval --data data/test.ndjson --model models/char-cnn-v2

# Evaluate on GitTables benchmark
finetype eval-gittables --dir eval/gittables

# Validate data quality against taxonomy schemas
finetype validate -f data.ndjson --strategy quarantine

Column-Mode Inference

Single-value classification can be ambiguous: is 01/02/2024 a US date (Jan 2) or EU date (Feb 1)? Is 1995 a year, postal code, or plain number?

Column-mode analyzes the distribution of values in a column and applies disambiguation rules:

Date format — US vs EU slash dates, short vs long dates
Year detection — 4-digit integers predominantly in 1900-2100 range
Coordinate resolution — latitude vs longitude based on value ranges
Numeric types — ports, increments, postal codes, street numbers

# CLI column-mode
finetype infer -f column_values.txt --mode column

# CSV profiling (uses column-mode automatically)
finetype profile -f data.csv

DuckDB Extension

-- Install and load
INSTALL finetype FROM community;
LOAD finetype;

-- Classify a single value
SELECT finetype('192.168.1.1');
-- → 'technology.internet.ip_v4'

-- Detailed output (type, confidence, DuckDB broad type)
SELECT finetype_detail(value) FROM my_table;
-- → '{"type":"datetime.date.us_slash","confidence":0.98,"broad_type":"DATE"}'

-- Normalize values for safe TRY_CAST
SELECT finetype_cast(value) FROM my_table;

-- Recursively classify JSON fields
SELECT finetype_unpack(json_col) FROM my_table;

-- Check extension version
SELECT finetype_version();

The extension embeds model weights at compile time — no external files needed.

Performance

Model Accuracy

Model	Accuracy	Test Samples
Flat CharCNN v2	91.97%	15,100

Post-processing rules improve Macro F1 from 87.9% to 90.8% (+2.9 points without retraining).

Real-World Evaluation (GitTables)

Evaluated against 2,363 annotated columns from 883 real-world CSV tables:

Type Category	Accuracy	Example Types
Timestamps	100%	`datetime.timestamp.*`
Country names	100%	`geography.location.country`
URLs	89.7%	`technology.internet.url`
Dates	88.2%	`datetime.date.*`
Person names	80-85%	`identity.person.*`

Column-mode inference improves accuracy for ambiguous types: geography +9.7%, datetime +4.8%.

Latency & Throughput

Metric	Value
Model load	66 ms cold, 25-30 ms warm
Single inference	p50 = 26 ms, p95 = 41 ms
Batch throughput	600-750 values/sec
Memory footprint	8.5 MB peak RSS

Installation

Homebrew (macOS)

brew install noon-org/tap/finetype

Cargo

cargo install finetype-cli

From Source

git clone https://github.com/noon-org/finetype
cd finetype
cargo build --release