Package index • joinery

Strategies

Declare how matching works. A strategy holds preparation pipelines, blocking, weights, and a threshold; it runs nothing itself.

search_strategy(): Define a Search Strategy for Record Linkage
exact_strategy(): Define an Exact Matching Strategy
embedding_strategy(): Create an Embedding Strategy
block_on_tokens(): Block on a Column's Rare Tokens (region-free blocking)
smooth_rip_identity() smooth_rip_log() smooth_rip_offset() smooth_rip_softmax(): Configure rIP smoothing for a search strategy

Text preparation

Turn a column’s text into tokens. Used inside a strategy formula (column ~ step1 + step2).

normalize_text(): Normalize text for matching
normalize_street(): Normalize street names across languages
normalize_date(): Normalize dates to ISO 8601 format (YYYY-MM-DD)
approximate_date(): Approximate dates by rounding to coarser time units
word_tokens(): Split text into word tokens
numeric_tokens(): Tokenize numeric fields, expanding ranges into individual numbers
date_tokens(): Extract date components as tokens
fuzzy_tokens(): Collapse near-duplicate tokens to a canonical form
generate_ngrams(): Generate character n-grams from text
token_shapes(): Convert tokens to shape signatures
extract_initials(): Extract initials from tokens
as_metaphone(): Encode text phonetically with Metaphone
as_soundex(): Encode text phonetically with Soundex
as_cologne(): Encode text phonetically with the Cologne procedure
strip_vowels(): Strip vowels from text (consonant skeleton)
filter_stopwords(): Filter out stopwords from token lists
find_stopwords(): Discover candidate stopwords from a prepared token table
drop_numeric_tokens(): Drop numeric (house-number) tokens from token lists
drop_short_tokens(): Drop short tokens from token lists
use_dictionary(): Map tokens to canonical groups with a lookup table

Matching

Find duplicates within a table, candidates across tables, resolve pairs into entities, and stage passes together.

detect_duplicates(): Detect Duplicate Records
deduplicate_table(): Deduplicate a Table
search_candidates(): Search for Candidate Matches Between Tables
extract_unmatched(): Extract Unmatched Records
materialize_records(): Materialize Records by ID
resolve_entities(): Group Matched Pairs into Entities
multi_stage_dedup(): Staged Duplicate Detection (within one table)
multi_stage_search(): Staged Search Across Tables or Sources
prepare_search_data(): Prepare Data for Record Linkage Search
compute_rarity(): Compute Token Rarity for Record Linkage

Diagnostics

Check a strategy before you run it, and inspect what it found.

inspect_tokens(): Inspect Tokens for a Specific Column
plan_strategy(): Plan a Search Strategy from Raw Inputs
audit_strategy(): Audit a Search Strategy Against Data
rarity_distribution(): Read the Token Rarity Distribution
summarise_matches(): Summarise a Match Result
explain_match(): Explain a Single Match
sample_matches(): Sample Matches for Review
compare_stages(): Compare Stages of a Multi-Stage Match
recommendations(): Recommendations from a Diagnostic Object

Plots

Visualise a diagnostic result. Each verb returns data; these draw it. The default plot() method of each diagnostic class calls the most useful one of these.

score_histogram(): Bar chart of the pre-binned score distribution
score_density(): Kernel density of the score distribution
coverage_plot(): Bar chart of match coverage (base and/or target)
cluster_size_plot(): Bar chart of cluster-size distribution (duplicates only)
ambiguity_plot(): Bar chart of candidates-per-record distribution (candidates only)
top_gap_density(): Bar chart of top-1 vs top-2 score gap distribution (candidates only)
rarity_histogram(): Bar chart of median token rarity per column
token_frequency_plot(): Bar chart of average tokens per record per column
block_size_plot(): Bar chart of block sizes (requires block_by on strategy)
vocab_overlap_plot(): Bar chart of vocabulary overlap between base and target per column
similarity_histogram(): Histogram of sampled pairwise cosine similarities
norm_plot(): Bar chart of embedding norm quantiles
contribution_plot(): Horizontal bar chart of per-column score contributions
token_contribution_plot(): Horizontal bar chart of per-token score contributions, coloured by column
stage_coverage_plot(): Line plot of cumulative base coverage by stage
stage_score_plot(): Grouped bar chart of score distributions by stage
frontier_plot(): Cost/recall frontier scatter for a strategy plan

Calibration

Train and evaluate a false-positive filter on labelled pairs.

export_for_labelling(): Export a match sample to CSV for manual labelling
import_labels(): Import a labelled CSV back into a feature/label table
match_features(): Build a per-pair feature table for calibration
fit_filter(): Fit a false-positive filter on labelled match pairs
apply_filter(): Apply a fitted filter to match features
calibrate_matches(): Calibrate matches end-to-end (features -> filter -> apply)
calibrate(): Evaluate a fitted filter on labelled pairs
joinery_recipe(): Build a tidymodels recipe for calibration features

Embeddings

Vector-based matching for the embedding strategy.

compute_embeddings(): Compute Embeddings for Records
score_embeddings(): Score Embedding Pairs Using Cosine Similarity
clear_embedding_cache(): Clear the embedding reuse cache

DuckDB backend control

Tuning for the out-of-core DuckDB backend.

duckdb_control(): DuckDB Execution Control
duckdb_batch_plan(): Create a Batch Plan for DuckDB Table Processing
batch_map(): Apply a function to DuckDB table batches
drop_joinery_temp_tables(): Drop all temporary DuckDB tables created by joinery

Example data

Synthetic datasets used throughout the documentation.

base_example: Base dataset for record linkage example
target_example: Target dataset for record linkage example
workshop_register: Workshop guild register (base) for record linkage examples
workshop_listings: Workshop external directory (target) for record linkage examples
workshop_panel: Multi-year workshop panel for cross-year linkage examples
match_labels_example: Labelled candidate pairs for calibration examples
street_types: Multilingual Street-Type Normalization Dictionary
street_stopwords: Multilingual Street-Name Stopwords