Collaborate on Invoice Document for Research and Development with Ease Using airSlate SignNow

Invoice document for Research and Development

in this lesson we will see how we can extract any invoice with uipath document understanding the machine learning let me present the case we have a folder here with a lot of invoices in today we only have three because of the example but you could have a imagine that you have a thousand here they look like this let me open two of them it is just random invoices that I found on Google by the way so they look like this you can see that they got the same structure sort of they got an invoice numbers and dates totals and some lines here but they are not the same so if we try to extract them with one group we will run into a problem that's why we're used to document understanding packages with machine learning let me close this one down and now let us create the case in uipath the first thing we will go to is the manage packages in July path go to all packages we need to three packages that is the document understanding here take this one click and save I accept then take the intelligent OCR click the manage packages again you have to do three times so in telleth James Bond she are here click click save click accept then we need to final package soul patch again and then we take the omni pack only the page activities here click and click Save the first thing that we need to do is to load in taxonomy so let me just dragon sequence that we could place it all into so it will be fit more Ora so we have the mean then we'll need to load a taxonomy solo taxonomy and we will create a taxonomy that we will load here so go up to taxonomy manager take this one now we need to talk about which data do we want out of our invoice for this example well let's just take maybe invoice number invoice date and maybe total so that will be three data but you could have a lot of other data so in Syria that we have a group will create group this one could be the company so click plus here we can call this a lot of variants that will be the company there of ours today yes please the category that could be the department that will handle the invoices maybe most often it will be account payable Department so click plus here again so a department like this now we need to add the new document type that will be the type invoice call it invoice and then we will add the fields the first thing that we need to extract is the English number so put it in here as often you will most often be waiting with text because we pass them to excel or all the files and it doesn't make sense to take numbers so just take two texts per default unless you have a reason that you click Save then you add a new field and we'll have the invoice date the type that will be text again click Save new field that will be the total text a reason for this is also that we don't know sometimes we have a text in our amount here you can see the dollar sign or text in our numbers up here so for to be saying just choose text then we click Save now we created our taxonomy we can click close we load our taxonomy that we just created and we need to load it into an output Hannibal and so ctrl K this one we can just called Texas Tech's son like this now we have loaded the taxonomy we need to read our document that is we only have it here as an image we need to read it into some sort of data that we can use the close this one down so let's find a digitize document this one will will read our document into something that we can use the first thing that we will use is document path I placed mine and the folder on the desktop core data and invoices you can see it here I will first just read one invoice and then we will create a solution to have any invoice in the folder so take this one shift right click on the invoice copy as path we will go back to UI path instead of just hard coding in it up here we create a variable that has always best practice so let's say STR dark path we will paste it in here as default value so that one will be coloring loss now we can go up here instead and for the document text and document object model that is both output variables this one the document text will contain our text from our invoice and a document object model will contain information properties of our text so if control K STR doc text the document object model will be control kan de ok so now we created two levels control K for creating variables when you are here in the field we can see that we created it down here practice now we need an OCR engine we will use two on the page here Toshio drag it in that's it now we are done with the digitizing so we can move down so we need to classify so let's find a classified document classified documents cool yeah directors in and here we got some sort of different fields that beneath the document path will that one is easy STR document path the taxonomy that's just a taxonomy the document text is text and then we have the document object model what this scope do is that we need to classify it because he meant that we have several documents I put in yourseives receipts and invoices we need to shorten so we need to know that every time there's an invoice we need to treat them as an invoice so that's what we're doing here the classification resolved that is after we made the classification we store them into a bear so control okay here again and we can just say class see the occasion results like this then we need to have some sort of a classifier and we can just choose the keyword based classifier and one will be search for keywords that could be likes maybe for the invoice type that could be invoiced because I think that's always or the invoice text and then English you can see that is here and probably in all our in watches he is doesn't what she as well so what we're doing is that we will need to create a learning fine path let us go to go to know plan plus plus and created a blank jason you could create it wherever you want I just like to create it so let's create a new it won't be blank no just say mass and let's see your data I will just say that in my save it in my phobia then I will go down to and choose the document type that is a JSON file here and click Save then we will use this file path and here you cannot use a variable you need to hard-coded in yep so thank you we can see that we call it new one we could have made an appropriate name but this one will be off blank and you can see that indeed a Jason so shift right-click Caviar's path and paste it in now we have a blank learning path we need to manage learning so what do we want to we have the invoice you we need to add some key words we will just choose invoice so whenever our classifier will come to the name it Moi's it will know that there's an invoice safe if we have more documents up in our taxonomy manager we could create more learnings but this one book is sufficient to the invoices then we can configure a classifier so yeah and we will just mark this as English then we will click Save now we are done with a classified document document scope we can move ahead we need to extract the data so let's find a data extraction scope here we will track the skin underneath so again we will have the STR document path here into document path we will have the texana me here we will have the document text so and here it's nice that we have created all these variables we can just easy put them in this one listed down here and then we'll have the classification resolved this one is is an array so we will just type in class in the case and results sorry 0 like this then we will extract it to something and this one will just be the extraction salt here so again we create a new babel control cave we will call this one extract songs like this now we need to tell you I pass the data extraction scope what extracts us do you want to use and here our machine learning extractor comes into play so search for a machine learning extract them drag it in fresh battery now we need an element and an API key ml skill is the mandatory so you don't have to fill that in the end point will go to search on google for machine learning extractor uipath then scroll a little bit down you can see a click here for information about a document understanding endpoints take you then this one is a community license and you can see yeah that we here in the model we will use an invoice so that's here if you use a community use this endpoint and if it's an enterprise solution use this endpoint I will use take this one I'll copy this one you copy it into endpoint iam 20 api key well go to platform like uipath calm it's always greater and login and after you login you will see this sub click on admin then click on licenses other services and then under document understanding click generate new that one will generate a new API key for document understanding and then just click copy API key like this now paste it in here and we can click get capabilities now we need to configure extractors so click here what we need to do is press this one that one will expand all the English data fields that we are we have invoice number we have in my state and children now these dropdowns that is what our machine learning extracted will extract so the invoice number that is the invoice no and this one is done automatically that's the smart path then we will date here and we total like this then we click save now we have extracted our data into the extraction Vishal's yep then we need to some sort of a validate that they are indeed the right results you don't have to do this step but it's nice so it will find a present validation station that is that could be data missing that could be we could not be one admission sure that the machine learning will extract the right day so this one is almost always a good idea so the document pass well yeah I know we will have to put this one in again dark text Tom Cech sodomy the extraction was Charles that was from up here and now we will have the validated results so click ctrl K Valley dated results like this so this one will prompt the user to verify that the data and their right you will see that in a few minutes I will present a solution for you now we just need to export the extraction results into a data table actually is a dataset but let's find a export extraction of salt here its extraction of salts well the input that is the validated results from appear and if dataset so control K we can call this data set results now we have either s a data set we get maybe for each so we want a data out we want to upload it to an excel sheet so find a for each find it here so for each item that's fine that'll be the data set with salts and then we'll take the attribute tables so take it like the tutor spits low tables then in the type argument change it from object to tentative like this just delete this body now we could it is this one that will only be one but we will write the range to an excel sheet so let's go back to our Thor now I will create some sort of a report or so new Microsoft Excel worksheet will you call this yeah maybe report or this could also be where we stored it before we put it into an efficient or suck so we have a report here nothing more than this and let me just shift right click and copy his path so we go back here and then we find a wide range drag it in from the white book so we don't need the Excel application scrub yep the white book path will let's just destroy it here STR salt sheet white book actually like this paste it in here so then we just say the STR the salt white book we eat the data table that is just the item is we are creating the data table up here and then we will add the headers so now we have created a workflow that will only read our first a noise remember that but it will work so shouldn't be drab let's do that so we run the file do I possible run now and in a few seconds we will have the validation station that was the one I told you about where we could validate our results yeah a validation station and you can see that we have extracted our invoice number that is us shuo shuo one from here we have the invoice date that is in that's true as well and we have the total that's true as well however we can see that how confident the machine learning extractor is that we have the right things that they actually totally right but this one could be a little bit off so we need such a Hume to validate this if you don't want to do this you can just skip the present validation station but we need to mark these trees in there okay let's say that we wanted to this one was not correct we could just to myself so 150 4.08 and stand whatever just click save and now we should write them into our excel sheet in a few seconds we can open up our forum we can open our report that which is created every can see that we did have created our reply that's fine so now we have created the solution now we just need to scale it so we can have any invoice in our folder let me move this one plus this one down let's think about a solution well I wanna have a data table where we can so when we we build a data table in the beginning then we have a for each that will loop through all our files in our invoice folder and then as we get to a source we will add them to take the table and then we will have this right range outside of the four inch but let's start building a data table so let's find a field take the table up here click to take the table and delete everything that's in it I don't know why this shouldn't have these predefined fields in you are almost or never using I have never used them but that was just a bit of a gambling so we got the invoice number yeah then we got the invoice date and a total so let us add some more cards invoice date here and we have a total sorry total here that it's deprecated okay then we need to create a data table for the output so - okay DT output like this now we need to have all all these steps into a for each where we loop through each of their files so let's find a voyage right in here the first thing that we'll do is say for each file because now we need to get each file name and we can do that quite easily with some simple repeat coding so we'll just say directory you notice it get files and then we'll put in the directory where we store our invoices and we start here so it is just shift right click copy as path then maybe we should put it in our variable that we created so the HTR docpath so instead of actually we could just delete this finally we have it here so this one will get the fires from the STR dock then we just repeat what we're doing here we get all the files from this folder the file names and then we can do something to it and we just need to do all these things to him so let me delete this file here we will go a little bit down so what we do is that we will happy this this this is this bracket in thanks now we need to change it a bit because we are not at the document path that isn't updated this one because this one will be our parent folder we need to file so everywhere we put in the document path we will say file and now you can see we have an area that is because this for each that is instead of the object up here in the tab argument we need to have a string like this then we saw it so we have the file here as well we have it here we have it here and that should be like this yeah that's it so now we we're doing all the steps however we need to we have our data table of view so instead of having this for each we will just do a merge data table that is let me show you very data table here what we'll do listen did our destination that was the output that we created of you so go over you dt output here then the source well we will just say that as is the data set result tables that one is our data table so we will go here and say the source that is the data set with salt tables and then let me click the tree dodger and then we will say the first one here so that one will be su here like this so now we merge it into the DT output and we will loop through the next one again so now we just need to have to have the right range let me crack this outside the for each we can delete the for each yeah what we need to fix here is that we will still write it to our basalt web hook but instead of the item will have the DTI output then we repeat what we did here we created a data table with the columns that we wanted we dragged our sequence into a voyage that got each invoice file so now we're just looking through each invoice and as we get one of them we will merge it into our data table a great so we've simply just added the lines and then finally when we are done with the voyage we make the right range let's see that in action we run the file now we have the document validation station for the first one well we know that is fine so we mark every one click Save then we will have their second one in a few seconds that is for the next illinois's my computer is a bit slow but if you have a faster computer this will run very smooth on my workstation that one will be so this is the second one we can verify that we have the right results invoice number that is ten well that's correct and we can see here the document type that is invoice so he had recognized that it's an invoice so this one is correct they're Illinois State well that's correct - and the total that is correct - - even though it's grand total so that's fine and now we will have our third invoice here so right now we have the invoice number that's 14 yep the invoice date that's here and then the total that is fine as well and let's see imagine that we have a new day to you then we can see here that this one is blank it's not a date then we would have had an error here so yeah that's just to show that we can easily correct them or we can see if anything is missing then just click Save now our workflow is done we will have our results here then to report so open up the report that's it we extracted the data that we wanted we could easily have gotten more and we had a thousand invoices we will have it extracted them as well that's it