Embeddings process

1. cleanup: delete the following files to ensure that they are re computed
   • html_corpus.txt.gz ... contains the extracted training xpaths
   • html_vocabulary.cvs.gz ... the vocabulary to id mapping
   • html_corpus.bin.gz ... the binary version (translated using the vocabulary) of the html xpath corpus
2. generate a file with the Xpath representations using generate-html-corpus-texts.py. the corresponding XPaths are stored in html_corpus.txt.gz.
3. use triinput.py to generate (a) the html vocabulary file and (b) the html corpus file as well as the corresponding training corpora.
4. use trilearn.py for training the embeddings.