Extracting a text value from an embedded string value (OCR via attachment)

[Anonymous]

Hi guys,

This would be my first post here and one issue has me scratching and pulling all nighters. Its addictive!..

Here are my pics.

I have an email attachment for a pdf which is first fed into a variable (string type)

Then conditioned.

and finally passed trough AI for image recognition. 

The value is then embedded into a string variable.

However now I simply need to get the Text value from here in the above [0] string. It is not an array so I cannot pass it though filter array function. Tried many alternatives with split() function and item()? type as well but of no use. 

From a total of 1200 lines I need to only filter those starting with "text:" values and then append them as an array into a sharepoint file. Just..that..simple 😕

Would be really grateful if some of the experts here can shed some light on to it. Eagerly awaiting your replies!

[Nived_Nambiar]

Hi @Anonymous 

Once you got the output from AI as the output data for pdf file, 

Next step you can parse the json string obtained to Parse json action.

Next use the output from parse json and pass it to Select action, here specify value section as the output from parse json and select the highlighted option

Then after that in the same field specify item()?['text'] 

Output from select action would array having only text values 🙂

Hope it helps 🙂

Mark it as solution if it resolves your query 🙂

[Anonymous]

Hi @Nived_Nambiar  thank you for your reply. Would really appreciate if you can guide me on how to parse json in this case (a total newbie here unfortunately!)...Thanks a lot!

[Nived_Nambiar]

Hi @Anonymous 

Refer this video 

https://youtu.be/e0dzMqoJGtY

What you have to do is the output which you are getting in form of json you have to pass into Content Field 

In schema section, click Generate from sample and then paste the same json structure which would generate schema from sample.

Hope it helps !

Mark it as solution if this resolved your query

[Anonymous]

Hi @Nived_Nambiar 

How do I get a JSON output here that can be sent to JSON parser. (The one im sending as an input is generating null values)

I have this as an out of the appended variable output

On expanding the output it gives all in row, which is not what is perhaps required in the schema for parsing json?

[Anonymous]

as per the video, do I need to create another variable and place the previous appended string values of 'PDF_Content' into it before parsing?

[Nived_Nambiar · Accepted solution]

No, currently which dynamic content is giving the required value, place that in Content section of parse json, then if you have same json content with you [ may you can do a run and using compose function you can capture the json data] , you can generate schema from it 🙂

Video is just giving an idea on how to work with that 🙂

[Anonymous]

@Nived_Nambiar getting the below error here

[Nived_Nambiar]

Could you show me the flow (without running )?

[Anonymous]

Here is the flow without running

[Nived_Nambiar]

Hi @Anonymous 

See this 

Next to Recognize text in pdf or image document, place the parse json action, in the content, place the results instead of PDF_content.

Also make sure schema matches that of result dynamic content 🙂

Let me know if it works 🙂

[Anonymous]

its giving this error

[Anonymous]

It's like this now

[Nived_Nambiar]

it seems certain properties are missing from input content you have passed 

Just make sure schema contains full data from result content 🙂

[Nived_Nambiar]

Hi @Anonymous 

Refer this flow  which i have created using manual trigger using the same AI builder action

Output got 

Hope this helps you 🙂

[Anonymous]

Hey @Nived_Nambiar 

Thanks for putting those lines in. Some progress at least!
I have followed these steps you have outlined above, but schema is failing now in the parse JSON section.

Flow now.

That's the run time now. What could be wrong for the schema in failing?

[Nived_Nambiar]

How you are copying schema ? 

[Nived_Nambiar]

Hi @Anonymous 

To copy the output result.

First add the result content in compose, run the flow, then from output section of compose section, copy the complete data using ctrl +c and in parse json schema, paste the copied and click done.

This is the way you can generate schema

Or paste this schema 

{
    "type": "array",
    "items": {
        "type": "object",
        "properties": {
            "@@odata.type": {
                "type": "string"
            },
            "page": {
                "type": "integer"
            },
            "lines@odata.type": {
                "type": "string"
            },
            "lines": {
                "type": "array",
                "items": {
                    "type": "object",
                    "properties": {
                        "@@odata.type": {
                            "type": "string"
                        },
                        "text": {
                            "type": "string"
                        },
                        "boundingBox": {
                            "type": "object",
                            "properties": {
                                "@@odata.type": {
                                    "type": "string"
                                },
                                "left": {
                                    "type": "number"
                                },
                                "top": {
                                    "type": "number"
                                },
                                "width": {
                                    "type": "number"
                                },
                                "height": {
                                    "type": "number"
                                },
                                "polygon": {
                                    "type": "object",
                                    "properties": {
                                        "@@odata.type": {
                                            "type": "string"
                                        },
                                        "coordinates@odata.type": {
                                            "type": "string"
                                        },
                                        "coordinates": {
                                            "type": "array",
                                            "items": {
                                                "type": "object",
                                                "properties": {
                                                    "@@odata.type": {
                                                        "type": "string"
                                                    },
                                                    "x": {
                                                        "type": "number"
                                                    },
                                                    "y": {
                                                        "type": "number"
                                                    }
                                                },
                                                "required": [
                                                    "@@odata.type",
                                                    "x",
                                                    "y"
                                                ]
                                            }
                                        }
                                    }
                                }
                            }
                        }
                    },
                    "required": [
                        "@@odata.type",
                        "text",
                        "boundingBox"
                    ]
                }
            }
        },
        "required": [
            "@@odata.type",
            "page",
            "lines@odata.type",
            "lines"
        ]
    }
}

Please try any option and let me know

[VictorIvanidze]

Use this: 

outputs('Recognize_text_in_an_image_or_a_PDF_document')?['body/responsev2/predictionOutput/results'][0]?['lines'][0]?['text']

[NR2023]

Hi Victor,

Would you be able to elaborate on what this action is doing? I'm having a similar issue and I am trying to achieve something similar to what OP has created. However, I would want to be able to filter on a text value in the PDF and then I would like to trigger an action to split PDF and export. 

At the minute I am getting stuck on filtering the Array. So it would be good to know how I can search the text values to get a partial match based on text? In my document, the example is 'IT (Hero)' where (Hero) is the site code we want to find.