Instructions and supporting tools for training word embeddings for historical (ca. 1600 - 1900) German based on the texts of the Deutsches Textarchiv using fastText.

A pre-trained model can be obtained here [1,43 GB].

1. Follow the instructions for building fastText

2. Download the DTA normalized XML files
   wget -i dta_normalized.txt -P dta_normalized

3. Transform the DTA normalized XML files into plain text
   xsltproc tei2txt.xsl dta_normalized/* -o dta_normalized/*.txt

4. Concatenate all plain text files into a single text file
   cp dta_normalized/*.txt dta_normalized/dta_normalized_all.txt

5. Compute word embeddings using fastText
   fasttext skipgram -input dta_normalized/dta_normalized_all.txt -output dta_emb

1. Download wget.exe from https://eternallybored.org/misc/wget/

2. Download msxsl.exe from https://www.microsoft.com/en-us/download/details.aspx?id=21714

3. Download fasttext.exe from https://github.com/xiamx/fastText/releases

4. Download the DTA normalized XML files
   wget.exe -i dta_normalized.txt -P dta_normalized

5. Run a batch script to convert the XML files to a combined plain text file
   dta2txt.bat

6. Compute word embeddings using fastText
   fasttext.exe skipgram -input dta_normalized\dta_normalized_all.txt -output dta_emb