This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(124.80000000000001, 59%, 22%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMD%3C/text%3E%3C/svg%3E)
I plan to use Form Recognizer to detect and convert tables data from pdf documents. I already trained and used the custom model as not all of the tables were properly recognized automatically. I use Azure Function and FormRecognizerClient to get and convert data to required format. It works, but the issue is that these tables sometimes contains units in mathematical formulas with subscripts and superscripts. The OCR doesn't recognize the power so f.ex 10^6 is recognized as 106, also some other signs are not recognized properly. Is there a way to improve this somehow? Maybe there is another Microsoft feature that I can use in parallel to detect things like this.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(128, 34%, 22%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EFF%3C/text%3E%3C/svg%3E)
I am facing the same issue, that subscript and superscript characters (e.g. a superscript ² referring to footnote "2") are not recognized and instead output as a normal 2 (instead of ²). If the values before the superscript character are also numbers, this actually results in incorrect extracted numbers.
It would be great, if susbcript and superscript characters would be recognized and rendered correctly in the output (guess this would hold for the read model as well as the lyout model).
For the built-in layout model, the top-notch feature would be to recognize such references and output it as new types (e.g. FootnoteReference and Footnote) in the analysis result.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(214.4, 45%, 30%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EGM%3C/text%3E%3C/svg%3E)
Hi, none that I'm aware of. I've forwarded your inquiry to the product group, will share updates soon as possible.