Remove duplicate information from altLabels and texts

[ioggstream]

I expect

• Skill altLabels present in the json.gz summaries to not contain redundant definitions

Instead

They contain texts with little variations (e.g., plurals, ...)

Notes

• Use spacy to identify duplicate sentences based on lemmas
• pick one sentence for each equivalence class

These sentences could then be used e.g. to create spacy matchers based on lemmas.