presentation

itpACDH_20200827_notes.md

@@ -0,0 +1,35 @@
+hello I'm Walter,
+I was lucky enough to be assigned to project Arthur Schnitzler Briefe conducted by Martin.
+
+use the hashtag to explore more on twitter.
+
+correspSearch dot net is a platform hosted by 
+Berlin-Brandenburg Academy of Sciences and Humanities intended to facilitate correspondence search, who wrote when and where a letter to whom.
+to do so, a TEI variation called CMIF has to be prepared.
+Correspondence Metadata Interchange Format
+
+there is an online editor available, as you see here.
+GND, gemeinsame Normdaten geo locations are supported and resolvable.
+
+input data were letter in print, scanned already.
+the larger volumes contained a index, three columns starting with recipient and dates.
+
+there are multiple approaches to tackle task to convert index into an CMIF.
+
+Martin inspired me to following workflow
+
+first we use transkribus to perform a text recognition.
+important to mention is that text area recognition needed a little help, manual selection was unavoidable.
+
+now all lines of the index were in a text file, line by line.
+
+with OpenRefine this messy data was shaped into a table.
+during this process, problems in text recognition surfaced.
+on first glance correct dates, couldn't be, because they weren't chronological.
+
+to detect such problems, a little python script in jupyter notebook was created.
+
+as final step, by means of OpenRefine suspect dates were edited manually.
+and shaped into CMIF json file, which CMIF Creator can load.
+GND has been adjusted manually by Martin.
+