Investigate why LeskEvaluatorTest and MFSEvaluatorTest fail while parsing 'EnglishLS.train'

The LeskEvaluatorTest & MFSEvaluatorTest in the opennlp-wsd sandbox component both fail parsing the 'EnglishLS.train' file. The data is kept original, downloaded from https://web.eecs.umich.edu/~mihalcea/senseval/senseval3/data.html

Aims:

• Investigate what causes the xml parsing to fail
• Fix it and make both existing tests pass
• Optional: Improve the existing test code to be more strict.

Note:

The test setup to reproduce this is on a branch and to be merged into the main branch.