fix entity linker

spacy/errors.py

@@ -113,6 +113,8 @@ class Warnings(object):
             "ignored during training.")
 
     # TODO: fix numbering after merging develop into master
+    W093 = ("Could not find any data to train the {name} on. Is your "
+            "input data correctly formatted ?")
     W094 = ("Model '{model}' ({model_version}) specifies an under-constrained "
             "spaCy version requirement: {version}. This can lead to compatibility "
             "problems with older versions, or as new spaCy versions are "
@@ -556,9 +558,6 @@ class Errors(object):
             "{obj}.{attr}\nAttribute '{attr}' does not exist on {obj}.")
     E186 = ("'{tok_a}' and '{tok_b}' are different texts.")
     E187 = ("Only unicode strings are supported as labels.")
-    E188 = ("Could not match the gold entity links to entities in the doc - "
-            "make sure the gold EL data refers to valid results of the "
-            "named entity recognizer in the `nlp` pipeline.")
     E189 = ("Each argument to `get_doc` should be of equal length.")
     E190 = ("Token head out of range in `Doc.from_array()` for token index "
             "'{index}' with value '{value}' (equivalent to relative head "

spacy/gold/example.pyx

@@ -109,7 +109,7 @@ cdef class Example:
                     prev_j = -1
                 prev_value = value
 
-        if field in ["ENT_IOB", "ENT_TYPE"]:
+        if field in ["ENT_IOB", "ENT_TYPE", "ENT_KB_ID"]:
             # Assign one-to-many NER tags
             for j, cand_j in enumerate(gold_to_cand):
                 if cand_j is None:
@@ -196,7 +196,7 @@ def _annot2array(vocab, tok_annot, doc_annot):
                 entities = doc_annot.get("entities", {})
                 if value and not entities:
                     raise ValueError(Errors.E981)
-                ent_kb_ids = _parse_links(vocab, words, value, entities)
+                ent_kb_ids = _parse_links(vocab, tok_annot["ORTH"], value, entities)
                 tok_annot["ENT_KB_ID"] = ent_kb_ids
         elif key == "cats":
             pass
@@ -308,7 +308,6 @@ def _parse_ner_tags(biluo_or_offsets, vocab, words, spaces):
 
 def _parse_links(vocab, words, links, entities):
     reference = Doc(vocab, words=words)
-
     starts = {token.idx: token.i for token in reference}
     ends = {token.idx + len(token): token.i for token in reference}
     ent_kb_ids = ["" for _ in reference]

spacy/pipeline/pipes.pyx

@@ -2,7 +2,6 @@
 import numpy
 import srsly
 import random
-from ast import literal_eval
 
 from thinc.api import CosineDistance, to_categorical, get_array_module
 from thinc.api import set_dropout_rate, SequenceCategoricalCrossentropy
@@ -92,15 +91,6 @@ class Pipe(object):
         """Modify a batch of documents, using pre-computed scores."""
         raise NotImplementedError
 
-    def update(self, docs, set_annotations=False, drop=0.0, sgd=None, losses=None):
-        """Learn from a batch of documents and gold-standard information,
-        updating the pipe's model.
-
-        Delegates to predict() and get_loss().
-        """
-        if set_annotations:
-            docs = list(self.pipe(docs))
-
     def rehearse(self, examples, sgd=None, losses=None, **config):
         pass
 
@@ -1094,31 +1084,19 @@ class EntityLinker(Pipe):
             predictions = self.model.predict(docs)
 
         for eg in examples:
-            doc = eg.predicted
+            kb_ids = eg.get_aligned("ENT_KB_ID", as_string=True)
-            ents_by_offset = dict()
+            for ent in eg.doc.ents:
-            for ent in doc.ents:
+                kb_id = kb_ids[ent.start]  # KB ID of the first token is the same as the whole span
-                ents_by_offset[(ent.start_char, ent.end_char)] = ent
+                if kb_id:
-            links = self._get_links_from_doc(eg.reference)
-            for entity, kb_dict in links.items():
-                if isinstance(entity, str):
-                    entity = literal_eval(entity)
-                start, end = entity
-                mention = doc.text[start:end]
-
-                # the gold annotations should link to proper entities - if this fails, the dataset is likely corrupt
-                if not (start, end) in ents_by_offset:
-                    raise RuntimeError(Errors.E188)
-                ent = ents_by_offset[(start, end)]
-
-                for kb_id, value in kb_dict.items():
-                    # Currently only training on the positive instances - we assume there is at least 1 per doc/gold
-                    if value:
                     try:
                         sentence_docs.append(ent.sent.as_doc())
                     except AttributeError:
                         # Catch the exception when ent.sent is None and provide a user-friendly warning
                         raise RuntimeError(Errors.E030)
         set_dropout_rate(self.model, drop)
+        if not sentence_docs:
+            warnings.warn(Warnings.W093.format(name="Entity Linker"))
+            return 0.0
         sentence_encodings, bp_context = self.model.begin_update(sentence_docs)
         loss, d_scores = self.get_similarity_loss(
             scores=sentence_encodings,
@@ -1137,11 +1115,10 @@ class EntityLinker(Pipe):
     def get_similarity_loss(self, examples, scores):
         entity_encodings = []
         for eg in examples:
-            links = self._get_links_from_doc(eg.reference)
+            kb_ids = eg.get_aligned("ENT_KB_ID", as_string=True)
-            for entity, kb_dict in links.items():
+            for ent in eg.doc.ents:
-                for kb_id, value in kb_dict.items():
+                kb_id = kb_ids[ent.start]
-                    # this loss function assumes we're only using positive examples
+                if kb_id:
-                    if value:
                     entity_encoding = self.kb.get_vector(kb_id)
                     entity_encodings.append(entity_encoding)
 
@@ -1158,10 +1135,11 @@ class EntityLinker(Pipe):
     def get_loss(self, examples, scores):
         cats = []
         for eg in examples:
-            links = self._get_links_from_doc(eg.reference)
+            kb_ids = eg.get_aligned("ENT_KB_ID", as_string=True)
-            for entity, kb_dict in links.items():
+            for ent in eg.doc.ents:
-                for kb_id, value in kb_dict.items():
+                kb_id = kb_ids[ent.start]
-                    cats.append([value])
+                if kb_id:
+                    cats.append([1.0])
 
         cats = self.model.ops.asarray(cats, dtype="float32")
         if len(scores) != len(cats):
@@ -1172,9 +1150,6 @@ class EntityLinker(Pipe):
         loss = loss / len(cats)
         return loss, d_scores
 
-    def _get_links_from_doc(self, doc):
-        return {}
-
     def __call__(self, doc):
         kb_ids, tensors = self.predict([doc])
         self.set_annotations([doc], kb_ids, tensors=tensors)

spacy/tests/pipeline/test_entity_linker.py

@@ -252,10 +252,18 @@ def test_preserving_links_ents_2(nlp):
 
 # fmt: off
 TRAIN_DATA = [
-    ("Russ Cochran captured his first major title with his son as caddie.", {"links": {(0, 12): {"Q7381115": 0.0, "Q2146908": 1.0}}}),
+    ("Russ Cochran captured his first major title with his son as caddie.",
-    ("Russ Cochran his reprints include EC Comics.", {"links": {(0, 12): {"Q7381115": 1.0, "Q2146908": 0.0}}}),
+        {"links": {(0, 12): {"Q7381115": 0.0, "Q2146908": 1.0}},
-    ("Russ Cochran has been publishing comic art.", {"links": {(0, 12): {"Q7381115": 1.0, "Q2146908": 0.0}}}),
+         "entities": [(0, 12, "PERSON")]}),
-    ("Russ Cochran was a member of University of Kentucky's golf team.", {"links": {(0, 12): {"Q7381115": 0.0, "Q2146908": 1.0}}}),
+    ("Russ Cochran his reprints include EC Comics.",
+        {"links": {(0, 12): {"Q7381115": 1.0, "Q2146908": 0.0}},
+         "entities": [(0, 12, "PERSON")]}),
+    ("Russ Cochran has been publishing comic art.",
+        {"links": {(0, 12): {"Q7381115": 1.0, "Q2146908": 0.0}},
+         "entities": [(0, 12, "PERSON")]}),
+    ("Russ Cochran was a member of University of Kentucky's golf team.",
+        {"links": {(0, 12): {"Q7381115": 0.0, "Q2146908": 1.0}},
+         "entities": [(0, 12, "PERSON")]}),
 ]
 GOLD_entities = ["Q2146908", "Q7381115", "Q7381115", "Q2146908"]
 # fmt: on

spacy/tests/test_gold.py

@@ -176,10 +176,12 @@ def test_gold_biluo_different_tokenization(en_vocab, en_tokenizer):
     spaces = [True, True, True, False, False]
     doc = Doc(en_vocab, words=words, spaces=spaces)
     entities = [(len("I flew to "), len("I flew to San Francisco Valley"), "LOC")]
+    links = {(len("I flew to "), len("I flew to San Francisco Valley")): {"Q816843": 1.0}}
     gold_words = ["I", "flew to", "San", "Francisco Valley", "."]
-    example = Example.from_dict(doc, {"words": gold_words, "entities": entities})
+    example = Example.from_dict(doc, {"words": gold_words, "entities": entities, "links": links})
     assert example.get_aligned("ENT_IOB") == [2, 2, 3, 1, 2]
     assert example.get_aligned("ENT_TYPE", as_string=True) == ["", "", "LOC", "LOC", ""]
+    assert example.get_aligned("ENT_KB_ID", as_string=True) == ["", "", "Q816843", "Q816843", ""]
 
     # additional whitespace tokens in GoldParse words
     words, spaces = get_words_and_spaces(