Improve geocoder matches for numeric adjectives #5997
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## dev-2.x #5997 +/- ##
===========================================
Coverage 69.73% 69.73%
- Complexity 17297 17317 +20
===========================================
Files 1954 1960 +6
Lines 74160 74263 +103
Branches 7595 7603 +8
===========================================
+ Hits 51717 51791 +74
- Misses 19806 19832 +26
- Partials 2637 2640 +3 ☔ View full report in Codecov by Sentry. |
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
acf9ada to
24975b4
Compare
|
|
||
| @ParameterizedTest | ||
| @ValueSource( | ||
| strings = { "Meridian Ave N & N 148th", "Meridian Ave N & N 148", "Meridian Ave N N 148" } |
There was a problem hiding this comment.
Just curious, is this supposed to work with "Meridian & N 148"?
There was a problem hiding this comment.
I had to check this, but the answer is yes!
I wrote a few more test cases to make sure the it's really the case.
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
8254797 to
e471aaa
Compare
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
dc686f9 to
45d6b68
Compare
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
45d6b68 to
edea05e
Compare
|
Today I had an idea how I can solve this problem even better: I am now stripping the number suffixes like "th" from the numbers so that "148th" becomes just "148". This is really effective at matching numbers (even one or two digit ones) without having to become too fuzzy for regular text search. This means that "arts" no longer matches "Arthur Place" so a good solution all around. |
leonardehrenfried
changed the title
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
70f99d4 to
f0a2c6c
Compare
src/ext/java/org/opentripplanner/ext/geocoder/EnglishNGramAnalyzer.java
Outdated
Show resolved
Hide resolved
leonardehrenfried
force-pushed
the
th-geocoder
branch
from
a8573c7 to
bf02dae
Compare
leonardehrenfried
changed the title
Improve geocoder matches for numeric adjectives
Summary
This PR improves the sandbox geocoder by shortening the minimal n-gram to 3 from 4.
This is so that the stop name of "Meridian Ave N & N 148th St" is matched when the users search for "Meridian Ave N & N 148". Previously "148th" would only be tokenised to "148t" but not "148".
This also changes one of the regression test cases: When you search for "arts" you now also get "arthur place" which before you didn't. @miles-grant-ibigroup said that this is fine.Unit tests
Added.