Skip to main content
Question

Geocoding from Text Blocks


braggken
Supporter
Forum|alt.badge.img+6

Looking for ideas on geocoding from text blocks. Each feature has a block of text (~50 to 100 words) describing an incident. More than one place could be mentioned so results would likely include more than one place. But in any case the goal is to geocode based on places mentioned in the text.

An example might look like this:

"...Something happed in Streetsville and people were walking toward Placeville."

And we would want geocoding results from Streetsville and Placeville.

4 replies

milo89
Enthusiast
Forum|alt.badge.img+16
  • Enthusiast
  • November 14, 2023

Hi @braggken​,

Not sure if this is too helpful but a while ago I had a similar requirement with address being stored in blocks with a variety of other attributes and text. Admittedly, probably not as long as 50-100 words though.

At first, as we were only looking for a general geocode (Suburb/City level), we split the text with an attribute splitter (and list exploder from memory I believe) and attempted to geocode all values.

In the end, It was actually easier to split the text block and then attempt to join to administrative data (city, suburb, town etc.) using a feature merger.

 

Also interested to see if there are other thoughts on this one.

 


redgeographics
Celebrity
Forum|alt.badge.img+48

Do you have a list of locations that are likely included in the text?


braggken
Supporter
Forum|alt.badge.img+6
  • Author
  • Supporter
  • November 15, 2023
redgeographics wrote:

Do you have a list of locations that are likely included in the text?

Yes I can create a list of most of locations that are likely to be in the text.

In fact this the road I was starting down. Short story: I look for each location in each block with a StringSearcher. This kind of works but two limitations:

  • the names must match exactly
  • my list is likely not complete.

redgeographics
Celebrity
Forum|alt.badge.img+48
braggken wrote:

Yes I can create a list of most of locations that are likely to be in the text.

In fact this the road I was starting down. Short story: I look for each location in each block with a StringSearcher. This kind of works but two limitations:

  • the names must match exactly
  • my list is likely not complete.

Yeah, I fear that that may be the case. I wonder if you can use ChatGPT (or other AI, I don't really work with it that much) to extract locations out of the text and then a geocoding API.


Cookie policy

We use cookies to enhance and personalize your experience. If you accept you agree to our full cookie policy. Learn more about our cookies.

 
Cookie settings