Cache results from spacy using pickle

[aoldoni]

• Attempt to optimise performance by caching result trees.
• Add parameter to force uncached processing

[aoldoni]

Seems not possible: https://github.com/spacy-io/spaCy/search?q=pickle&type=Issues&utf8=%E2%9C%93 Seeing same error as: https://github.com/spacy-io/spaCy/issues/377

[aoldoni]

One way to resolve this is:

• Parse text using spacy structure.
• Transform spacy tree into custom Tree structure that supports edge labels (NLTK doesn't).
• Change code all over to use this custom structure.
• If, in the future, if extra data from spacy is needed, then add to this custom structure as part of the conversion method.
• Pickle that structure and reload that structure instead of the spacy one.

Downside is basically reimplementing the spacy structure, but in a "picklable" manner.

[aoldoni]

Attempted in https://github.com/aoldoni/comp9596/commit/0f23de30f552f51ac3530e4485c6f32b744805ad

[aoldoni]

Fixed in https://github.com/aoldoni/comp9596/commit/a32b85e4de3570537312df1f637b27274448b371 by implementing TreeNode, a new tree structure that is serializable and mimics Spacy tree structure.