Extract suburb name from address string in Bigquery - sql

Question

I have a table of addresses (property) from which I need to extract just the suburb name. I have another table (suburbs) that contains all of the suburb names.
I'm having a problem with the multi-word suburb names, where a match is found on one, and both words. I need it to match with the longest suburb name, eg. an address with "North Bondi" should only match to suburb "North Bondi" and not suburb "Bondi".
I've found some examples online that use the MAX function in the join but Bigquery won't let me use that function in the join.
Would appreciate if someone could please suggest corrections, or provide guidance on other solutions (eg. sorting the suburb table and retrieving only one result?) Thank you!
Table: property
address
12 Smith Street Surry Hills NSW
34 Jones Street Bondi NSW
15 Sunny Road North Bondi NSW
Table: suburbs
suburb
state
Surry Hills
NSW
Bondi
NSW
North Bondi
NSW
Current code used:
Select * from ( SELECT p.address, s.suburb
FROM `property` p
JOIN `suburbs` s
ON INITCAP(p.address) LIKE CONCAT('%', INITCAP(s.suburb),' ', INITCAP(s.state), '%')
GROUP BY p.address, s.suburb
) x
join `property` p
ON p.address = x.address
where p.address is not null;
Actual result:
address
suburb
12 Smith Street Surry Hills NSW
Surry Hills
34 Jones Street Bondi NSW
Bondi
15 Sunny Road North Bondi NSW
Bondi
15 Sunny Road North Bondi NSW
North Bondi
Desired result:
address
suburb
12 Smith Street Surry Hills NSW
Surry Hills
34 Jones Street Bondi NSW
Bondi
15 Sunny Road North Bondi NSW
North Bondi

Answer 1

Try this:
select v1.address,
string_agg(v1.addr_part, " ") as suburb,
from (
select t.address,
addr_part,
from property t
cross join unnest(split(t.address, " ")) addr_part WITH OFFSET AS ofst
where ofst > 2
qualify row_number() over(partition by t.address order by ofst desc) > 1
) v1
group by v1.address
;
But! This approach assumes:
The first 3 words in every address are not belong to suburb name;
Every state is one word.

Answer 2

I've found some examples online that use the MAX function in the join but Bigquery won't let me use that function in the join.
Using a window function Instead of a MAX function,
Select * from ( SELECT p.address, s.suburb
FROM `property` p
JOIN `suburbs` s
ON INITCAP(p.address) LIKE CONCAT('%', INITCAP(s.suburb),' ', INITCAP(s.state), '%')
) x
QUALIFY RANK() OVER (PARTITION BY address ORDER BY LENGTH(suburb) DESC) = 1
Query results:
address
suburb
12 Smith Street Surry Hills NSW
Surry Hills
15 Sunny Road North Bondi NSW
North Bondi
34 Jones Street Bondi NSW
Bondi

Answer 3

Consider below approach
SELECT p.address,
STRING_AGG(s.suburb ORDER BY LENGTH(s.suburb) DESC LIMIT 1) suburb
FROM `property` p
JOIN `suburbs` s
ON INITCAP(p.address) LIKE CONCAT('%', INITCAP(s.suburb),' ', INITCAP(s.state), '%')
GROUP BY p.address
if applied to sample data in your question - output is