 Many things also very welcome from my side to this nice conference. My name is Stefan von der Heide, I'm from a company CCS content conversion specialist to be allocated in Hamburg, Germany. So, it's my pleasure to present you our let's say experience we have made with a new metagraphy, metagraphy API and which we have used in the fact Ukraine handwritten project. What I do now is to provide you an overview about the project I will tell you about this solution architecture we have used and of course I will tell you something about our experience about the new metagraphy API and finally present some results. The project itself is about digitizing handwritten birth, marriage and death registries. Well, the job was to to collect let's say the main information like names, dates and places, which were from the second half of the ninth century until the beginning of the 20th century. And there's a form based handwritten material with some variants and layout writing style and of course like typical like scanning quality. And you can see for example, here's a typical sample some tables filled in handwritten form. Then might be the scanning is not that good, or you have some, let's say menu over writing in the lower left part. Another form you see let's say more columns on the right. And so on, you know, so it's more or less the same on the one hand on the other side, it is pretty different in variances. So there's something to do. We did it in the following way, and the scans were provided by our customers or they are as they are. Then we as a company have our own digitization software which we call dog dog with and with this. We have done the copying and de screwing and let's say as main job is the, the segmentation part and the classification part. So that we get in the end, something like like like this year so we can get text regions with the dates. Well, here not the name but let's say candidates for the name but typically this case and names are in the yellow let's say a bigger paragraphs and a location maybe or an age is on the left column and so on. So this let's say this is the result from our own software. And then it is put on the new metagrapho API. So let's say the transcribers part here is now coming into place which and this is doing the line segmentation as you have seen we have recognized the whole paragraph. And of course the main job is to do the HDR with the text we have got back. We really important to our own software using additional some named recognition really to identify the names and the full text thing. Some post processing which you need in this case and some export and then we had it. So this is how we process the workflow and this job. And maybe an important general hint is here and let's say our dog with software is a kind of it's a workflow software which runs all steps automatically but it has a chance to let's say for manual QA, which is important in this case because it's not none of these steps are perfect you know. So you really need some persons if you want to reach some quality in the end to make some manual checks or even corrections. Okay, really coming now to the metagrapho API and a bit more detail. Maybe for the one you're not very know what this is so we use it and let's say in our case in a cut let's say simple way at least we think this is this is simple. So you have some model and you provide via this API, the image, the model, the idea of the model, and in our case some layout information this case the text region. And we will get back the next detailed level of layout this is text lines here and the main topic for our project here was a text itself of course. The training is independent of the API was done with the usual transcribers tools so training was independent. We have this model and then we really have uses this API. The purpose is mainly really mass production you know we want to produce really a lot of this kind of data. We were in the good position that we are some how let's say early adopters so we had a chance to discuss and define let's say, or get some influence let's say on the API definition which was very helpful for us but I hope also a bit for the retranscribers team. And I can say really that we appreciate very much so how we work together so the most kind of working many thanks for the transcribers team here again. And we can tell you really openly and honestly that the API is already pretty good instability and performance so this is really ready now for using it. And maybe just a result from the overall thing just the HDR result we have, we could reach was a character up about 8% which was reasonable in our context. And our purpose was to identify let's say some names dates and so on so so even as this is character is for the general character at the rate let's say we have to evaluate. I do not have a number unfortunately about the important data, of course, but I can tell you let's say from what we want to reach and how satisfied the customers and from commercial perspective. It was very successful. Let's summarize it in our opinion. This is a good working API and what it was very useful in our context. Yeah, that's it. Make contact me later if you like. Thank you. Thank you Stefan for this interesting insights into the practical use of one of our more recent products. Thank you for processing API. Are there any questions from the audience. Anything you would like to know or comment on a little comment from my side. Yeah, as I said in my earlier presentation or my part of the next gen presentation. One of the future goals for transcribers is to further facilitate large scale processing. And this is basically what we have done. And for which we are providing a tool now. And yeah, we're really happy that it seems to be working and hope that the collaboration will be as fruitful because I can really come return the compliment and it's been a pleasure to work with you guys. Because then I think it's time for our next talk, but not before I give you a mark.