dedcode

{
  "mentions": [
    {
      "name": "Robinson College",
      "type": "string",
      "start": "2898",
      "end": "2914",
      "context": "char=0,5253",
      "uri": "https://en.wikipedia.org/wiki/The_New_York_Times"
    },

dedcode

# coding: utf-8

# In[163]:

import codecs
from collections import defaultdict

tweets = defaultdict()
with io.open('/Users/xi/Downloads/NEEL2016-training.tsv', encoding='utf-8') as tw:

dedcode

NIL138
['Alan', 'Jasons', 'goonergang']

NIL143
['jfalconer', 'HopticalA', 'Alex']

NIL840
['BestProNews']

NIL18

dedcode

@torch.no_grad()
def buildSubgraph(self, head_index: Tensor, rel_type: Tensor, tail_index: Tensor) -> Tensor:
    k = self.neighbors_topk  # Number of top tails to find
    M = self.neighbors_size

    # Initialize a dictionary to store unique objects the triples
    unique_nodes = {(h.item(), r.item(), t.item()): set() 
                        for h, r, t in zip(head_index, rel_type, tail_index)}

    # local utility that uses the KGE to find the nearest nodes.