Improves Typo recognition for autocomplete #4690
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…oves the score increases to be on the outer loop.
dwnoble
approved these changes
Oct 25, 2024
hqpho
pushed a commit
to hqpho/dc-website
that referenced
this pull request
Oct 29, 2024
This PR modifies the scoring algorithm for place autocomplete to count a small score for non-exact matches, to account for one typo. With these changes, we will favor "San Diego" over "Dieppe" for the query "Sna Die". Prod: https://screenshot.googleplex.com/Bsx2BbyLZArbQuX Local with this change: https://screenshot.googleplex.com/9jHqKb2uHJLz37k Note that "Sne Die" will still go back to "Dieppe" because that's 2 typos, so San Diego is out even if it was returned by google Maps predictions: https://screenshot.googleplex.com/9LViJoVFni3Lui6 Typo check done as a bag of letters with at most off by one. We do this check on top of the Google Maps Predictions which already take into account typo correction. This part is just to choose the best prediction from google maps. Doing this as part of gaps identified in place autocomplete: https://docs.google.com/document/d/15RVckX9ck5eyyhBHW8Nb9lmxPBDPMIeLbax14HbN-GI/edit?tab=t.0
hqpho
added a commit
to hqpho/dc-website
that referenced
this pull request
Oct 29, 2024
* update submodule for release (datacommonsorg#4681) * update NL goldens after mixer push (datacommonsorg#4680) * Adds logging for autocomplete responses. (datacommonsorg#4678) Logs the response count for autocompletion. Staging is not showing any responses. Would like to better understand where the breakdown is occurring. * Exit cdc_services/run.sh when any background process exits (datacommonsorg#4682) This makes startup errors in Mixer or NL servers more obvious. Bug: b/374820494 Reference: https://docs.docker.com/engine/containers/multi-service_container/#use-a-wrapper-script * update nodejs goldens (datacommonsorg#4685) goldens needed to be updated because of a bunch of recent data updates (data diffs can be seen here: datacommonsorg/mixer#1438, datacommonsorg/mixer#1439) * Update submodules (datacommonsorg#4688) * Pin transformers to 4.45.2 (datacommonsorg#4689) Also updates nl goldens * Support schema update mode for data docker (datacommonsorg#4686) Use the `DATA_RUN_MODE` environment variable to decide what mode to pass to run_stats.sh and whether to build embeddings. The mode `schemaupdate` for run_stats.sh is added by datacommonsorg/import#344, which this PR updates the import submodule to include. A docsite page will describe how to pass in this environment variable: datacommonsorg/docsite#527 * Improves Typo recognition for autocomplete (datacommonsorg#4690) This PR modifies the scoring algorithm for place autocomplete to count a small score for non-exact matches, to account for one typo. With these changes, we will favor "San Diego" over "Dieppe" for the query "Sna Die". Prod: https://screenshot.googleplex.com/Bsx2BbyLZArbQuX Local with this change: https://screenshot.googleplex.com/9jHqKb2uHJLz37k Note that "Sne Die" will still go back to "Dieppe" because that's 2 typos, so San Diego is out even if it was returned by google Maps predictions: https://screenshot.googleplex.com/9LViJoVFni3Lui6 Typo check done as a bag of letters with at most off by one. We do this check on top of the Google Maps Predictions which already take into account typo correction. This part is just to choose the best prediction from google maps. Doing this as part of gaps identified in place autocomplete: https://docs.google.com/document/d/15RVckX9ck5eyyhBHW8Nb9lmxPBDPMIeLbax14HbN-GI/edit?tab=t.0 --------- Co-authored-by: chejennifer <[email protected]> Co-authored-by: Gabriel Mechali <[email protected]> Co-authored-by: natalie <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR modifies the scoring algorithm for place autocomplete to count a small score for non-exact matches, to account for one typo.
With these changes, we will favor "San Diego" over "Dieppe" for the query "Sna Die".
Prod: https://screenshot.googleplex.com/Bsx2BbyLZArbQuX
Local with this change: https://screenshot.googleplex.com/9jHqKb2uHJLz37k
Note that "Sne Die" will still go back to "Dieppe" because that's 2 typos, so San Diego is out even if it was returned by google Maps predictions: https://screenshot.googleplex.com/9LViJoVFni3Lui6
Typo check done as a bag of letters with at most off by one. We do this check on top of the Google Maps Predictions which already take into account typo correction. This part is just to choose the best prediction from google maps.
Doing this as part of gaps identified in place autocomplete: https://docs.google.com/document/d/15RVckX9ck5eyyhBHW8Nb9lmxPBDPMIeLbax14HbN-GI/edit?tab=t.0