shortest path dependency between three entities using spacy

Question

I'm trying to build ternary intra sentence relationships. One of the methods I'm considering is shortest path dependency algorithm with pos tag sequence , shortest path dependency sequence which will be used as features to a kernel based SVM. I'm not sure on how to formulate these features.

txt='Domestic revenues increased 14% to $680.8 million and were 77% of total revenues for the year ended December 31, 2015.'

doc = nlp(txt)
for token in doc:
    print((token.head.text, token.text, token.dep_,token.pos_))
edges = []
for token in doc:
    for child in token.children:
        edges.append(('{0}'.format(token.lower_),
                      '{0}'.format(child.lower_)))

graph = nx.Graph(edges)

the shortest path between second token of domestic revenues and 2015 looks like this

shortest path length :7
shortest path: ['revenues', 'increased', 'were', 'for', 'year', 'ended', 'december', '2015']

1. How do I use this dependency graph as a feature sequence for ternary relationship ? ( Audit-nsubj-increased-quantmod-million--conj-were )

2. How do i use generalized pos tags for these entity relationships (Audit-verb-num-num).

Since the entities in question are compound Im ok for the model to classify last tokens of the entities as a ternany relationship like this: (revenues,million,2015)--> (Audit,value,data)