Issue with the CSV importer data

[simardeep1792]

Hi @gagandeepsingh1105 ,

Could you investigate an issue with our current Pelias instance? According to @NickZachary , there seems to be a problem when API requests are sent to the CSV data. Although the data exists, the response we receive is the centroid of the data instead of the original geocoding locations. Additionally, the accuracy appears to be incorrect.

I have attached some screenshots that @NickZachary shared with me for your reference.

[simardeep1792]

@gagandeepsingh1105 if you need more insights feel free to connect with me or @NickZachary.

[gagandeepsingh1105]

Sure @simardeep1792 , I'll start working on it and let you know if I need any support.

[NickZachary]

My thoughts on this:

• we confirmed that the address queried exists in the original StatCan CSV and the lat/long is rooftop.
• The same address resolves to rooftop point using the OSM data in the Pelias instance
• Looking at the returned JSON response from the souce=other (i.e. StatCan) you can see that the address was normalized and ought to just be a straight up textual match in ElasticSearch. We know that it should exists if the input csvs from csv-importer worked correctly.

[simardeep1792]

@gagandeepsingh1105 do we have any updates on the testing? Can we do a quick test run on this as @NickZachary is recommending?

[gagandeepsingh1105]

Hi @NickZachary , can we connect to quickly discuss the approach for further testing the csv-import functionality.

Thanks

[gagandeepsingh1105]

As discussed with the team, after trying a couple of configurations changes to resolve the issue of confidence level dropping to 0.6 when city name is included in the text search , we decided to open a git hub issue for Pelias open community: pelias/csv-importer#107

[simardeep1792]

@gagandeepsingh1105

Could you please provide an update on this issue? It appears that members of the community have responded.

[gagandeepsingh1105]

Hi @simardeep1792 , the response on the git hub issue by Pelias community doesn't provide enough explanation for the behavior we have noticed when including city names in the search. So, James Logan (from Zac's team) has put in another comment in the issue seeking for a more relevant explanation/reasoning.

[simardeep1792]

@gagandeepsingh1105 Thanks for the update.