Greater Sudbury data projectOne park, two languages, two answers: auditing the City's bilingual file
The City of Greater Sudbury publishes its parks as open data: 406 rows, each with an English and a French name, category and address. Before letting an agent answer questions about it in either language, we checked that the two halves of each record agree. We wrote the checks and made the counts ourselves, on a copy of the CSV taken on 23 October 2026.
The file at a glance (our tally of the City of Greater Sudbury Parks / Parcs file, 406 rows)| What we counted | Result |
|---|
| Largest English categories | Natural Park 92, Playground 83, Park 74, Tot Lot 70 |
| Community values | 23, with Sudbury on 206 rows |
| English name Unnamed / French name Sans nom | 94 / 93 |
| Category pairs off the usual English and French pattern | 3 |
| Rows where only one language says the park is unnamed | 9 |
| French names or categories with stray spaces or line breaks | 21 |
Why this matters for an agent: a tool that looks parks up by French category would return OBJECTID 169 as a natural park, while the English record calls it a plain Park. Asked in English, the agent says one thing; asked in French, another. Neither answer is invented, which is what makes it hard to notice.
The three checks, using only Python's standard library:# Sudbury parks, English half against French half: an audit we wrote for this page
from collections import Counter
import csv
rows = list(csv.DictReader(open("sudbury-parks.csv", encoding="utf-8-sig")))
clean = lambda s: " ".join(s.split())
# Learn the usual English-to-French category pairing from the file itself (the majority wins).
pairs = Counter((clean(r["Category_E"]), clean(r["Category_F"])) for r in rows)
usual = {}
for (en, fr), n in pairs.most_common():
usual.setdefault(en, fr)
# Check 1: categories whose French label is not the usual partner of the English one.
odd_category = [(r["OBJECTID"], clean(r["Category_E"]), clean(r["Category_F"])) for r in rows
if usual[clean(r["Category_E"])] != clean(r["Category_F"])]
# Check 2: one language says the park has no name, the other gives one.
EN_NONE, FR_NONE = "Unnamed", "Sans nom"
odd_name = [(r["OBJECTID"], clean(r["Name_EN"]), clean(r["Name_FR"])) for r in rows
if (clean(r["Name_EN"]) == EN_NONE) != (clean(r["Name_FR"]) == FR_NONE)]
# Check 3: text that needs trimming before any lookup can match it.
untrimmed = sum(1 for r in rows for k in ("Name_FR", "Category_F") if r[k] != r[k].strip())
print("parks", len(rows), "| category pairs that break the usual pattern:", odd_category)
print("name pairs where only one language says unnamed:", len(odd_name))
print("French name or category values with stray spaces or line breaks:", untrimmed)
What it printed:parks 406 | category pairs that break the usual pattern: [('115', 'Tot Lot', 'Rue Redfern'), ('123', 'Natural Park', 'Parc'), ('169', 'Park', 'Parc naturel')]
name pairs where only one language says unnamed: 9
French name or category values with stray spaces or line breaks: 21
The odd category pairs
OBJECTID 115 is a Tot Lot in English and Rue Redfern in French, a street name sitting in the category column. OBJECTID 123 is a Natural Park paired with Parc, and 169 a Park paired with Parc naturel. The pattern each one breaks was learned from the file itself, by majority.
The odd name pairs
OBJECTID 65 is Unnamed in English but croissant Douglas in French. Five parks, including 265 Lively Greenspace and 273 Bell Grove, have English names and Sans nom in French. Three more, 364 to 366, say Unnamed - Lake Access and Sans nom, which a person may decide is consistent after all.
Why a person decides
The checks find candidates, not verdicts. Whether Unnamed - Lake Access matches Sans nom is a judgement about meaning. A good audit lists every case with its ID and leaves that call to someone who can ask the City.
What an honest agent answer looks like
In our adult course, the lookup tool returns both languages and the OBJECTID, and the answer must quote the ID it used. When the two languages disagree, the agent says so: OBJECTID 169 is listed as Park in English and Parc naturel in French. An answer without an ID is rejected by the check, and so is one that reports only the language that was asked.
Reused under the City of Greater Sudbury Open Data Licence, an adaptation of Ontario's Open Government Licence 1.0 whose grant covers commercial purposes (licence, dataset). Contains information licensed under the Open Data Licence - City of Greater Sudbury. No endorsement by the City is implied.
