A bot using the GCP Document AI - no coding needed | C2C Community

A bot using the GCP Document AI - no coding needed

  • 14 December 2022
  • 6 replies
  • 67 views

Userlevel 5
Badge +5

Have you ever thought of utilizing the capabilities of AIs in a business process without writing a line of code? Here is an example process I built in an hour:

Requirement: Use the GCP Document AI to recognize text in an image and save the recognized text to a file.

Steps:

  1. Activate the API for the GCP Document AI and add a standard OCR processor.
  2. Create a Service Account and key file
  3. Download the Automation Anywhere Community Edition (free) and follow the setup instructions.
  4. Connect to the API using the Document AI package and provide the path to your test image.
  5. Use the JSON Parser to get the AI results.
  6. Parse the text field and write it into a file.   

And here are the results:

Below is my example image

This is the complete API response:

And here is the text key value: 

And finally the text file:

 

As you can see, the AI most likely considered the cloud logo to be a ‘2’ but otherwise recognized the text correctly.

The bot could have repeated the actions for multiple files, written the result into a Google Sheet or database, or sent the file as an email attachment.

I think the combination of bots and AIs is a very powerful, fast, and easy way to get processes automated - not only rule-based processes. And without writing a line of code.

What do you think?


6 replies

Userlevel 7
Badge +35


Thank you, @MarkusK  for sharing  Document AI implementation of sucess, You did great but similar to google keep. Try to extend more features like extracting text with the format, color and table, and graph.

Userlevel 7
Badge +29

Amazing! What happens in the case of working with different languages, @MarkusK ? What if a single image has both English and Japanese for instance, with different character sets?

Userlevel 5
Badge +5

@Dimitris Petrakis that is a good question. I tried this image:

The Document AI returned the following result:

森波拉酒店→
Shampoole Hotel
丛林餐厅 ➔
Forest Restaurant
177
停车场
Parking lot
土特产 街
Specialties Street

 

And here is a screenshot from a Google translation (I only pasted the Chinese characters returned from the Document AI):

 

 

Userlevel 7
Badge +35

Well done! @MarkusK,  You took good content to do the experiment for Document AI and get a good result.

Userlevel 7
Badge +35

I hope @Dimitris Petrakis will like @MarkusK experiment based on his question. Thank you @Dimitris Petrakis for the question. It was a good question based on Document AI.

Userlevel 7
Badge +29

Wow! Impressive! And also correct! This is amazing, @MarkusK ! Shampoola / Sempora is also correct as it is a matter of pronunciation (I know from Japanese and the same rules apply with Chinese). For instance R and L are the same. 

 

I really didn't expect it to work so perfectly 😯

 

I am wondering now: Is it DocAI used on Google Translate too? Most obviously, I guess... Right? 

 

It has been really useful when I was traveling to other countries who had no English menu (happened in Japan and in Denmark). You use your camera and it translates on-the-fly! ❤️

Reply