Skip to main content

Parser failure report

CorpuscorpusThe BIO-labeled training data used to train Mailwoman's neural classifier. Assembled from real sources (OpenAddresses, National Address Database) and synthetic shards (boundary stress, order variants, negative space). Managed by @mailwoman/corpus. golden:dev · 4255 fixtures · modelsneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.' shipped, v263. Every gold labelground truthThe correct answer for an example, used as the standard a prediction is graded against. Mailwoman's ground truth is the hand-labeled golden set; its quality caps achievable accuracy. is graded (floor labelscomponent tagOne of the 33 labels in Mailwoman's address schema — street, locality, region, postcode, house_number, unit, po_box, country, venue, intersection, and others. Each parsed span carries exactly one component tag. compare the streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels./prefix/suffix family); a fixture "fails" a labelcomponent tagOne of the 33 labels in Mailwoman's address schema — street, locality, region, postcode, house_number, unit, po_box, country, venue, intersection, and others. Each parsed span carries exactly one component tag. when the parseaddress parsingThe process of decomposing a free-text postal address string into structured components — house number, street name, locality, region, postcode, and country — so a geocoder can resolve them to coordinates. disagrees with the hand gold.

Note: golden gold uses the flat pre-split street schema, so the street row reads confounded here (the modelneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.' splits prefix/streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels./suffix). country / region / locality / postcode / house_number are single-tag and valid — that is where a class trade shows.

Per-model failure count

modelneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.'fixtures with ≥1 labelcomponent tagOne of the 33 labels in Mailwoman's address schema — street, locality, region, postcode, house_number, unit, po_box, country, venue, intersection, and others. Each parsed span carries exactly one component tag. failurerate
shipped109526%
v263110126%

Failures by label — the class each candidate trades

A count that rises across candidates (bold, with the delta vs the first modelneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.') is a class the candidate silently traded. This is the row that flags a regression the floor gates miss — e.g. country degrading across the fragment lineage.

labelcomponent tagOne of the 33 labels in Mailwoman's address schema — street, locality, region, postcode, house_number, unit, po_box, country, venue, intersection, and others. Each parsed span carries exactly one component tag.shippedv263
region556580 (+24)
locality477481 (+4)
street432437 (+5)
postcode9298 (+6)
venue8786
house_number7476
country3424
dependent_locality3434
street_prefix66
street_prefix_particle66
po_box44
attention11
cedex11
unit11

Correlation — failure rate by structural class

Read across a row: a class that fails more on one modelneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.' is the shape of what that candidate traded (delimiter-free, non-ASCII, short inputs, per countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.).

structural classshippedv263
whitespace-only248/919 (27%)255/919 (28%)
has comma845/3312 (26%)844/3312 (25%)
non-ASCII396/827 (48%)390/827 (47%)
has countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head. gold127/224 (57%)117/224 (52%)
≤3 tokenstokenOne word or subword in the tokenized input. For the neural classifier, tokens come from SentencePiece (subword units); for the rule classifiers, tokens are whitespace- and punctuation-separated words.349/1251 (28%)356/1251 (28%)
US750/2527 (30%)761/2527 (30%)
FR276/1546 (18%)283/1546 (18%)
ZZ (synthetic)0/0 (—)0/0 (—)

Beyond reach — fails on every model (1073)

The persistent core: addresses no current candidate parsesaddress parsingThe process of decomposing a free-text postal address string into structured components — house number, street name, locality, region, postcode, and country — so a geocoder can resolve them to coordinates.. Per-countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head. breakdown (the correlation), then a sample.

countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.beyond-reach fixtures
US747
FR273
UN40
AD12
U.1

Sample (first 40 of 1073)

countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.inputsource
ADNew York, New York Steakhouse, 3790 Las Vegas Blvd S, Las Vegas, NV 89109adversarial
ADAthens, Georgia Café, 200 College Ave, Athens, GA 30601adversarial
ADBoston, Massachusetts Pizza Co., 50 Broadway, Brooklyn, NY 11201adversarial
ADMiami, Florida Surf Shop, 100 Ocean Dr, San Diego, CA 92101adversarial
ADNashville, Tennessee BBQ Pit, 75 Elm Ave, Atlanta, GA 30308adversarial
ADP'tit St. Denis Café, 500 Boul. Saint-Laurent, Montreal, QC H2X 2T6adversarial
ADSainte-Marie Bistro, 25 Rue Sainte-Catherine, Montreal, QC H3B 1A3adversarial
AD123Main St, Portland, OR 97214adversarial
AD50.Tremont.St,Boston,MA,02108adversarial
AD123 Main St c/o John Smith, Portland, OR 97214adversarial
AD123 Main St #4B, Portland, OR 97214adversarial
ADCenteradversarial
FR10 Rue de la République, 75008 Parisfr
FR45 Cours Lafayette, 69003 Lyonfr
FR1 bis Avenue des Champs-Élysées, 75008 Paris, Francefr
FR100 Rue de la Paix, 75008 CEDEX 08 Parisfr
FRParis 8efr
FRHôtel de Ville, 4 Place de l'Hôtel-de-Ville, 75004 Parisfr
FR7 Promenade des Anglais, 06000 Nicefr
FRÎle-de-France, Francefr
FR15 rue de la république, 75008 parisfr
FRШом, Creuse, Francefr
FRШом, Creusefr
FRCreuse, Шом, Francefr
FRFrance, Шом, Creusefr
FR富尔内勒, Lozère, Francefr
FRFrance, Lozère, 富尔内勒fr
FR富尔内勒 Lozèrefr
FRLOZÈRE 富尔内勒 FRANCEfr
FRFrance, Lozere, Runik yozuvfr
FRLozere, Runik yozuvfr
FRFrance, Creuse, Сен-Дизье-Леиренfr
FRФрессіне-де-Лозер, Lozère, Francefr
FRFRANCE, LOZÈRE, Фрессіне-де-Лозерfr
FRФрессіне-де-Лозер Lozèrefr
FRCreuse, Thauronfr
FRCreuse, Le Chauchet, Francefr
FRFrance, Creuse, Le Chauchetfr
FR卡萨尼亚, Lozèrefr
FR卡萨尼亚 Lozèrefr

Model-specific — where candidates disagree (50)

A ✓ under one modelneural classifierThe machine learning model at the core of Mailwoman's parser — a transformer encoder (~30M parameters) trained from scratch to do BIO token classification over addresses. It learns the 'grammar' of address formats; the gazetteer supplies the 'atlas.' and a failure under another = a fix or a regression between candidates — the diff you track release-over-release.

countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.inputshippedv263
ADPhone: 555-1234 — 200 Elm St, Springfield, IL 62701house_number→55-1234 200
FRManailly Creuse FRANCEregionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.Creuse FRANCE · countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
FRLizières Creuse FranceregionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.Creuse France · countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
FRMontredon Lozère FrancelocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Montredon Lozère · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
FRЛа-Нуай Creuse FrancelocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Ла-Нуай Creuse · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
FR47110 2 place de verdun sainte-livrade-sur-lotstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.Place DE · localitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Verdun Sainte-Livrade-Sur-Lot
FRWoustviller 57915 FrancepostcodepostcodeThe country-specific postal code (US ZIP, French code postal, etc.). Mailwoman handles postcode parsing entirely by rule classifier — a regex problem, not an ML one.57915 France
FR13940 Mollégès, FrancelocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.
FRIFS 14123 FrancepostcodepostcodeThe country-specific postal code (US ZIP, French code postal, etc.). Mailwoman handles postcode parsing entirely by rule classifier — a regex problem, not an ML one.14123 France
FR62119 Dourges FrancepostcodepostcodeThe country-specific postal code (US ZIP, French code postal, etc.). Mailwoman handles postcode parsing entirely by rule classifier — a regex problem, not an ML one.62119 France
FRLe Puy-Sainte-Réparade 13610localitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Le
FR59112 Ann ullin, FrancelocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.ullin
FRFrance, 59112 Ann ullinlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.France Ann ullin
FRWimereux, 62930localitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.
UNUnited States of America, Медоу Ејкерс, WYcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNПатни, Vermont, United States of AmericalocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Патни, Vermont · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
UNUnited States of America, ND, ЛорейнcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, ND, هاغوcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, Vermont, СодомcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNVermont, United States of AmericaregionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
UNUnited States of America, ND, MonroecountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNWyoming, ہلزڈیل وائیومنگ, United States of AmericalocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Wyoming ہلزڈیل وائیومنگ · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
UNUnited States of America, VT, УэтерсфилдcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, Wyoming, هوباک وایومینگcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNعلاالدین وایومینگ, Wyoming, United States of AmericalocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.علاالدین وایومینگ, Wyoming · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
UNUnited States of America, North Dakota, BloomfieldcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, اوپهام قوزئی داکوتا, NDlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.اوپهام وزئی داکوتا · countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, VT, ВотерфордcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, Vermont, МонтпилиерcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, لئونارد داکوتای شمالی, NDcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, Wyoming, بیر ریورcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, Wyoming, AntilopcountrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
UNUnited States of America, North Dakota, واتفورد سیتی قوزئی داکوتاlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.واتفورد سیتی وزئی داکوتا · countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.
USCider Mill Rd VT 05161, USstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.Cider VT Mill Rd · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
USSTATE RTE 100, VT 05350, USAstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.State Rte 100 VT · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
US122R Ave Southwest ND 58655streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.Ave Southwest
USVT 05051 Deerfield LnstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.VT Deerfield Ln · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
USFarson Branch Library, 30 Hwy 28 Farson-Eden School Bldg, Farson, WY 82932streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.Hwy 28 Farson-Eden School
USManchester Community Library, 138 CEMETERY AVE RTE 7A, Manchester VTstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.CEMETERY AVE RTE
USUS اوبران North DakotalocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.US اوبران
USبروکهورست وایومینگ, Wyoming, USlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.بروکهورست وایومینگ, Wyoming · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
USNorth Dakota, AntilopolocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.North Dakota Antilopo · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
USهئیل, North Dakota, USlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.هئیل, North · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.Dakota
USUS - ND - نوميlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.ND - نومي · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.
USUS, ND, AntilopalarlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.Antilo · regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.ND palar
USUS, Хак, NDlocalitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighbourhood in the hierarchy.US Хак
USSARDGEEN ARBON PLLC, 110 W FAYETTE ST STE 1110, SYRACUSE, NY 13202streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.W Fayette ST
USSeyfarth Shaw Fairweather, 1211 Ave of the America, New York, New YorkstreetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.Ave of the
USJo Ann Brinkley, PO BOX 7298, Glen Rose, TX 76043streetstreetThe named linear feature along which house numbers are ordered. Decomposes into a name plus street affixes; one of the Tier 2 fine labels.BOX 7298
USVirginia Stricklin, 9122 Heathcliff, San Antonio, TX 78254, USregionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality.TX US