Expanding edge mask with new indices from tokenizer - numpy

Question

I have a boolean tensor of edge_mask of size N x N which tells me the connections between words and relations of a graph. In this example, edge_mask has size 17 x 17:
tensor([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
[0., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
...
[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 1., 1.]])
Because I have the example 17 nodes:
[(0, 'arrived'),
(1, ':nsubj'),
(2, 'Wollstonecraft'),
(3, ':pobj'),
(4, 'in Grenada'),
(5, ':prep'),
(6, 'on'),
(7, ':pcomp'),
(8, 'board'),
(9, ':dobj'),
(10, 'the ship Sydney'),
(11, ':prep'),
(12, 'on'),
(13, ':pcomp'),
(14, 'board'),
(15, ':dobj'),
(16, 'the ship Sydney')]
I also have the above list of nodes. I tokenised the sentence from which I extracted the graph and some words or expressions might be mapped to various tokens. The tokenised mapping is given by the following two lists:
[(1, ['ar', 'rived']),
(2, ['']),
(3, [':nsubj']),
(4, ['W', 'oll', 'stone', 'craft']),
(5, ['']),
(6, [':pobj']),
(7, ['in']),
(8, ['Gren', 'ada']),
(9, ['']),
(10, [':prep']),
(11, ['on']),
(12, ['']),
(13, [':pcomp']),
(14, ['board']),
(15, ['']),
(16, [':dobj']),
(17, ['the']),
(18, ['ship']),
(19, ['Sydney']),
(20, ['']),
(21, [':prep']),
(22, ['on']),
(23, ['']),
(24, [':pcomp']),
(25, ['board']),
...
(26, ['']),
(27, [':dobj']),
(28, ['the']),
(29, ['ship']),
(30, ['Sydney'])]
And:
[0,
1,
1,
2,
3,
4,
4,
4,
4,
5,
6,
7,
8,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
...
None,
None,
None,
None,
None]
The latter list has size of the expected output (M x M), and has the indices from the words in the former list (e.g., arrived is tokenized in the second and third tokens, because the latter list has 1s in the second and third positions and 1 maps to ["ar", "rived"]).
I wanted to update the last list with the entries on edge_mask according to the new mapping from the tokenizer. For example, having the second and third rows have the first row of the edge_mask because these rows regard the token arrived. Furthermore, this row should also be expanded because it is of size N and not M. What I want is a squared matrix where the files and ranks represent the subtokens, so the expansion should also be performed for the columns (for example, in the above mentioned case, there should be a new column 3 with the same value as 2). Do you have any suggestion on how to do this?