Writing a simple baseline summarizer

[anspas]

Go from parser output to multi-sentence generator input From 20 sentences, experiment with baselines:

• Choose first sentence, generate this as summary
• Choose 2 sentences, transform into multi-sentence AMR, then generate summary from this (the Penman library we saw earlier in class should help with this)
• Choose random 4 sentences, 6, etc.
• Play with different ways of choosing these sentences (e.g. sentences with highest number of nodes / most semantic concept)
• Extractive vs. abstractive summarization

[anspas]

Up next:

• debug pseudo-summarizer.py
• convert summary AMRs (selected sentences) to multi-sentence AMR