OCR Engine Failing to See Text (No Special Characters). What to Try Next?

[Ukiman1014]

I have a process that looks up information on a few hundred addresses in NYC.  In only a few instances, OCR can't "see" the address in the search results.  The text is clear with no special characters and the browser window is full screen.  So far, I've tried using the Tesseract engine if the Windows OCR engine fails.  I've also tried foreground window instead of entire screen.  The mouse just doesn't move, even though the text is there in black and white.

Here is the text the OCR engine is looking for, along with a link to the web page that it's analyzing:

4710 7 Avenue	HPDOnline - Search Results (nyc.gov)
4714 7 Avenue	HPDOnline - Search Results (nyc.gov)
1320 St Johns Place	HPDOnline - Search Results (nyc.gov)

What else can I try?  Is there normally some trick or workaround when OCR is failing for no good reason?  I should note that the mouse isn't going to the wrong place, it's not moving anywhere.

[UshaJyothi20]

Did you try with extract  data from web page or get details from webpage

if not please try.

[Agnius]

Why exactly are you using OCR here? It is the least reliable way of automating anything. You can use Extract data from web page to extract it, or Get details of element in web page. You can also use Get details of web page to get the entire page source of the page and then use Parse text to retrieve the value you are looking for using regex. Plenty of ways to do it, depending on what works best. 

I would absolutely advice against using OCR, unless everything else totally fails.

-------------------------------------------------------------------------
If I have answered your question, please mark it as the preferred solution. If you like my response, please give it a Thumbs Up.

I also provide paid consultancy and development services using Power Automate. If you're interested, DM me and we can discuss it.

[Ukiman1014]

I should have explained further.  Each time a block and lot is entered, there can be multiple results in various orders and each result can either be active or inactive. The goal is to click on the address that matches the one it's looking for and only on the active one (sometimes there are multiple matching entries, one active and one inactive).  At first I was trying to use the regular web page tools and I could easily determine if the text was on the page.  The problem was clicking on the right entry.  I couldn't figure out how to do that.  I worked with a consultant who directed me to use OCR.  It was finding the text and clicking on it.  It worked on over 200 properties, except for these three.

It currently works by clicking on the first match it finds, then I check if the word "active" appears on the resulting page.  If it does not, then it's inactive and it goes back to the last page and clicks on the next match.  It continues to do this until it finds an active entry or there are no more matches.

Can someone please briefly explain how I can do that with the regular web extraction tools?  Since "active" and "inactive" are listed next to each entry, I shouldn't have to click on them to determine that.  I'm not that great with the browser tools, so I wasn't sure how to read in the table of entries (I can do that) but then analyze them and translate them to clicks on the screen.  I probably am just ignorant but, as usual, I didn't have a lot of time to figure it out and OCR was working and recommended to me.

[Ukiman1014]

I had no idea that OCR was the "least reliable way of automating anything".  It was working well and simpler to use then the other tools that require more knowledge of web page coding and RegEx.  Not all of us are adept at this stuff, and it was kind of a rude way of telling me I was doing something "dumb". 🙁

[Agnius]

I don't see how my reply is rude. I told you the truth about the approach you're using and suggested better alternatives. I also never said what you're doing is "dumb". That's something you said yourself. I simply said it is the least reliable way to automate stuff. And that's true. It is. Perfectly fine that you didn't know that. Now you do, and hopefully will try to avoid using OCR when alternatives are available from now on.

Now, you can use Extract data from web page if you have several results in the page, or Get details of element in web page if there's only one element in the page and you want to verify the result. The samples you provided lead to only a single search result in all cases, so you might as well use Get details of element in web page and capture the element. Should work fine.

If you want to click on the item, you can also build a custom selector for the UI element you want to click on. I've right-clicked on the element and selected "Inspect" to view it in the developer tools of my browser (Chrome in my case) and here's what the element looks like: 

It actually has the text there, and this could be used in the selector. When I captured the element to click on, it actually automatically captured the text: 

You can then edit the selector and pass in a variable there for the text. If you use that in a Press button in web page action, it will press on the item with the address you're looking for.

I've tested and verified it to work. 

-------------------------------------------------------------------------
If I have answered your question, please mark it as the preferred solution. If you like my response, please give it a Thumbs Up.

I also provide paid consultancy and development services using Power Automate. If you're interested, DM me and we can discuss it.