Merge pull request #474 from allenai/052_upgrade

Update to the latest UMLS version

Author: dakinggg

.flake8

@@ -2,21 +2,16 @@
 max-line-length = 115
 
 ignore =
-    # these rules don't play well with black
-    E203  # whitespace before :
-    W503  # line break before binary operator
-    W504  # line break after binary operator
+    E203
+    W503
+    W504
 
 exclude =
     build/**
     docs/**
 
 per-file-ignores =
-    # __init__.py files are allowed to have unused imports and lines-too-long
     scispacy/__init__.py:F401
     scispacy/**/__init__.py:F401,E501
 
-    # scripts don't have to respect
-    #  E501: line length
-    #  E402: imports not at top of file (because we mess with sys.path)
     scripts/**:E501,E402

evaluation/evaluate_linker.py

@@ -0,0 +1,97 @@
+import spacy
+from scispacy.linking import EntityLinker
+from scispacy.data_util import read_full_med_mentions
+import os
+from tqdm import tqdm
+
+EVALUATION_FOLDER_PATH = os.path.dirname(os.path.abspath(__file__))
+
+
+def main():
+    nlp = spacy.load("en_core_sci_sm")
+    nlp.add_pipe(
+        "scispacy_linker", config={"resolve_abbreviations": True, "linker_name": "umls"}
+    )
+    linker = nlp.get_pipe("scispacy_linker")
+
+    med_mentions = read_full_med_mentions(
+        os.path.join(EVALUATION_FOLDER_PATH, os.pardir, "data", "med_mentions"),
+        use_umls_ids=True,
+    )
+
+    test_data = med_mentions[2]
+
+    total_entities = 0
+    correct_at_1 = 0
+    correct_at_2 = 0
+    correct_at_10 = 0
+    correct_at_40 = 0
+    correct_at_60 = 0
+    correct_at_80 = 0
+    correct_at_100 = 0
+    for text_doc, entities in tqdm(test_data):
+        for start, end, label in entities["entities"]:
+            text_span = text_doc[start:end]
+            candidates = linker.candidate_generator([text_span], 40)[0]
+            sorted_candidates = sorted(
+                candidates, reverse=True, key=lambda x: max(x.similarities)
+            )
+            candidate_ids = [c.concept_id for c in sorted_candidates]
+            if label in candidate_ids[:1]:
+                correct_at_1 += 1
+            if label in candidate_ids[:2]:
+                correct_at_2 += 1
+            if label in candidate_ids[:10]:
+                correct_at_10 += 1
+            if label in candidate_ids[:40]:
+                correct_at_40 += 1
+            # if label in candidate_ids[:60]:
+            #     correct_at_60 += 1
+            # if label in candidate_ids[:80]:
+            #     correct_at_80 += 1
+            # if label in candidate_ids[:100]:
+            #     correct_at_100 += 1
+
+            total_entities += 1
+
+    print("Total entities: ", total_entities)
+    print(
+        "Correct at 1: ", correct_at_1, "Recall at 1: ", correct_at_1 / total_entities
+    )
+    print(
+        "Correct at 2: ", correct_at_2, "Recall at 2: ", correct_at_2 / total_entities
+    )
+    print(
+        "Correct at 10: ",
+        correct_at_10,
+        "Recall at 10: ",
+        correct_at_10 / total_entities,
+    )
+    print(
+        "Correct at 40: ",
+        correct_at_40,
+        "Recall at 40: ",
+        correct_at_40 / total_entities,
+    )
+    # print(
+    #     "Correct at 60: ",
+    #     correct_at_60,
+    #     "Recall at 60: ",
+    #     correct_at_60 / total_entities,
+    # )
+    # print(
+    #     "Correct at 80: ",
+    #     correct_at_80,
+    #     "Recall at 80: ",
+    #     correct_at_80 / total_entities,
+    # )
+    # print(
+    #     "Correct at 100: ",
+    #     correct_at_100,
+    #     "Recall at 100: ",
+    #     correct_at_100 / total_entities,
+    # )
+
+
+if __name__ == "__main__":
+    main()

project.yml

@@ -2,7 +2,7 @@ title: "scispaCy pipeline"
 description: "All the steps needed in the scispaCy pipeline"
 
 vars:
-  version_string: "0.5.1"
+  version_string: "0.5.2"
   gpu_id: 0
   freqs_loc_s3: "s3://ai2-s2-scispacy/data/gorc_subset.freqs"
   freqs_loc_local: "assets/gorc_subset.freqs"

requirements.in

@@ -18,6 +18,8 @@ flake8
 black
 mypy
 types-requests
+types-setuptools
+types-tabulate
 
 # Required for releases.
 twine

scispacy/abbreviation.py

@@ -82,6 +82,20 @@ def find_abbreviation(
     return short_form_candidate, long_form_candidate[starting_index:]
 
 
+def span_contains_unbalanced_parentheses(span: Span) -> bool:
+    stack_counter = 0
+    for token in span:
+        if token.text == "(":
+            stack_counter += 1
+        elif token.text == ")":
+            if stack_counter > 0:
+                stack_counter -= 1
+            else:
+                return True
+
+    return stack_counter != 0
+
+
 def filter_matches(
     matcher_output: List[Tuple[int, int, int]], doc: Doc
 ) -> List[Tuple[Span, Span]]:
@@ -100,6 +114,10 @@ def filter_matches(
             # Take one word before.
             short_form_candidate = doc[start - 2 : start - 1]
             long_form_candidate = doc[start:end]
+
+            # make sure any parentheses inside long form are balanced
+            if span_contains_unbalanced_parentheses(long_form_candidate):
+                continue
         else:
             # Normal case.
             # Short form is inside the parens.
@@ -190,7 +208,7 @@ def __call__(self, doc: Doc) -> Doc:
         filtered = filter_matches(matches_no_brackets, doc)
         occurences = self.find_matches_for(filtered, doc)
 
-        for (long_form, short_forms) in occurences:
+        for long_form, short_forms in occurences:
             for short in short_forms:
                 short._.long_form = long_form
                 doc._.abbreviations.append(short)
@@ -209,7 +227,7 @@ def find_matches_for(
         all_occurences: Dict[Span, Set[Span]] = defaultdict(set)
         already_seen_long: Set[str] = set()
         already_seen_short: Set[str] = set()
-        for (long_candidate, short_candidate) in filtered:
+        for long_candidate, short_candidate in filtered:
             short, long = find_abbreviation(long_candidate, short_candidate)
             # We need the long and short form definitions to be unique, because we need
             # to store them so we can look them up later. This is a bit of a

scispacy/candidate_generation.py

@@ -1,4 +1,4 @@
-from typing import List, Dict, Tuple, NamedTuple, Type
+from typing import Optional, List, Dict, Tuple, NamedTuple, Type
 import json
 import datetime
 from collections import defaultdict
@@ -41,38 +41,38 @@ class LinkerPaths(NamedTuple):
 
 
 UmlsLinkerPaths = LinkerPaths(
-    ann_index="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2020-10-09/umls/nmslib_index.bin",  # noqa
-    tfidf_vectorizer="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2020-10-09/umls/tfidf_vectorizer.joblib",  # noqa
-    tfidf_vectors="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2020-10-09/umls/tfidf_vectors_sparse.npz",  # noqa
-    concept_aliases_list="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2020-10-09/umls/concept_aliases.json",  # noqa
+    ann_index="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2023-04-23/umls/nmslib_index.bin",  # noqa
+    tfidf_vectorizer="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2023-04-23/umls/tfidf_vectorizer.joblib",  # noqa
+    tfidf_vectors="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2023-04-23/umls/tfidf_vectors_sparse.npz",  # noqa
+    concept_aliases_list="https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/data/linkers/2023-04-23/umls/concept_aliases.json",  # noqa
 )
 
 MeshLinkerPaths = LinkerPaths(
-    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/mesh/nmslib_index.bin",  # noqa
-    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/mesh/tfidf_vectorizer.joblib",  # noqa
-    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/mesh/tfidf_vectors_sparse.npz",  # noqa
-    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/mesh/concept_aliases.json",  # noqa
+    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/mesh/nmslib_index.bin",  # noqa
+    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/mesh/tfidf_vectorizer.joblib",  # noqa
+    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/mesh/tfidf_vectors_sparse.npz",  # noqa
+    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/mesh/concept_aliases.json",  # noqa
 )
 
 GeneOntologyLinkerPaths = LinkerPaths(
-    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/go/nmslib_index.bin",  # noqa
-    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/go/tfidf_vectorizer.joblib",  # noqa
-    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/go/tfidf_vectors_sparse.npz",  # noqa
-    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/go/concept_aliases.json",  # noqa
+    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/go/nmslib_index.bin",  # noqa
+    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/go/tfidf_vectorizer.joblib",  # noqa
+    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/go/tfidf_vectors_sparse.npz",  # noqa
+    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/go/concept_aliases.json",  # noqa
 )
 
 HumanPhenotypeOntologyLinkerPaths = LinkerPaths(
-    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/hpo/nmslib_index.bin",  # noqa
-    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/hpo/tfidf_vectorizer.joblib",  # noqa
-    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/hpo/tfidf_vectors_sparse.npz",  # noqa
-    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/hpo/concept_aliases.json",  # noqa
+    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/hpo/nmslib_index.bin",  # noqa
+    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/hpo/tfidf_vectorizer.joblib",  # noqa
+    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/hpo/tfidf_vectors_sparse.npz",  # noqa
+    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/hpo/concept_aliases.json",  # noqa
 )
 
 RxNormLinkerPaths = LinkerPaths(
-    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/rxnorm/nmslib_index.bin",  # noqa
-    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/rxnorm/tfidf_vectorizer.joblib",  # noqa
-    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/rxnorm/tfidf_vectors_sparse.npz",  # noqa
-    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2020-10-09/rxnorm/concept_aliases.json",  # noqa
+    ann_index="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/rxnorm/nmslib_index.bin",  # noqa
+    tfidf_vectorizer="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/rxnorm/tfidf_vectorizer.joblib",  # noqa
+    tfidf_vectors="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/rxnorm/tfidf_vectors_sparse.npz",  # noqa
+    concept_aliases_list="https://ai2-s2-scispacy.s3-us-west-2.amazonaws.com/data/linkers/2023-04-23/rxnorm/concept_aliases.json",  # noqa
 )
 
 
@@ -196,15 +196,14 @@ class CandidateGenerator:
 
     def __init__(
         self,
-        ann_index: FloatIndex = None,
-        tfidf_vectorizer: TfidfVectorizer = None,
-        ann_concept_aliases_list: List[str] = None,
-        kb: KnowledgeBase = None,
+        ann_index: Optional[FloatIndex] = None,
+        tfidf_vectorizer: Optional[TfidfVectorizer] = None,
+        ann_concept_aliases_list: Optional[List[str]] = None,
+        kb: Optional[KnowledgeBase] = None,
         verbose: bool = False,
         ef_search: int = 200,
-        name: str = None,
+        name: Optional[str] = None,
     ) -> None:
-
         if name is not None and any(
             [ann_index, tfidf_vectorizer, ann_concept_aliases_list, kb]
         ):
@@ -363,7 +362,7 @@ def __call__(
 
 
 def create_tfidf_ann_index(
-    out_path: str, kb: KnowledgeBase = None
+    out_path: str, kb: Optional[KnowledgeBase] = None
 ) -> Tuple[List[str], TfidfVectorizer, FloatIndex]:
     """
     Build tfidf vectorizer and ann index.

scispacy/data_util.py

@@ -1,4 +1,4 @@
-from typing import NamedTuple, List, Iterator, Dict, Tuple
+from typing import Optional, NamedTuple, List, Iterator, Dict, Tuple
 import tarfile
 import atexit
 import os
@@ -148,9 +148,10 @@ def remove_overlapping_entities(
 
 def read_full_med_mentions(
     directory_path: str,
-    label_mapping: Dict[str, str] = None,
+    label_mapping: Optional[Dict[str, str]] = None,
     span_only: bool = False,
     spacy_format: bool = True,
+    use_umls_ids: bool = False,
 ):
     def _cleanup_dir(dir_path: str):
         if os.path.exists(dir_path):
@@ -209,7 +210,12 @@ def label_function(label):
 
     for example in examples:
         spacy_format_entities = [
-            (x.start, x.end, label_function(x.mention_type)) for x in example.entities
+            (
+                x.start,
+                x.end,
+                label_function(x.mention_type) if not use_umls_ids else x.umls_id,
+            )
+            for x in example.entities
         ]
         spacy_format_entities = remove_overlapping_entities(
             sorted(spacy_format_entities, key=lambda x: x[0])

scispacy/file_cache.py

@@ -8,7 +8,7 @@
 import json
 from urllib.parse import urlparse
 from pathlib import Path
-from typing import Tuple, Union, IO
+from typing import Optional, Tuple, Union, IO
 from hashlib import sha256
 
 import requests
@@ -17,7 +17,9 @@
 DATASET_CACHE = str(CACHE_ROOT / "datasets")
 
 
-def cached_path(url_or_filename: Union[str, Path], cache_dir: str = None) -> str:
+def cached_path(
+    url_or_filename: Union[str, Path], cache_dir: Optional[str] = None
+) -> str:
     """
     Given something that might be a URL (or might be a local path),
     determine which. If it's a URL, download the file and cache it, and
@@ -47,7 +49,7 @@ def cached_path(url_or_filename: Union[str, Path], cache_dir: str = None) -> str
         )
 
 
-def url_to_filename(url: str, etag: str = None) -> str:
+def url_to_filename(url: str, etag: Optional[str] = None) -> str:
     """
     Convert `url` into a hashed filename in a repeatable way.
     If `etag` is specified, append its hash to the url's, delimited
@@ -68,7 +70,7 @@ def url_to_filename(url: str, etag: str = None) -> str:
     return filename
 
 
-def filename_to_url(filename: str, cache_dir: str = None) -> Tuple[str, str]:
+def filename_to_url(filename: str, cache_dir: Optional[str] = None) -> Tuple[str, str]:
     """
     Return the url and etag (which may be ``None``) stored for `filename`.
     Raise ``FileNotFoundError`` if `filename` or its stored metadata do not exist.
@@ -99,7 +101,7 @@ def http_get(url: str, temp_file: IO) -> None:
             temp_file.write(chunk)
 
 
-def get_from_cache(url: str, cache_dir: str = None) -> str:
+def get_from_cache(url: str, cache_dir: Optional[str] = None) -> str:
     """
     Given a URL, look for the corresponding dataset in the local cache.
     If it's not there, download it. Then return the path to the cached file.

scispacy/hyponym_detector.py

@@ -36,7 +36,6 @@ class HyponymDetector:
     def __init__(
         self, nlp: Language, name: str = "hyponym_detector", extended: bool = False
     ):
-
         self.nlp = nlp
 
         self.patterns = BASE_PATTERNS
@@ -91,7 +90,6 @@ def expand_to_noun_compound(self, token: Token, doc: Doc):
         return doc[start:end]
 
     def find_noun_compound_head(self, token: Token):
-
         while token.head.pos_ in {"PROPN", "NOUN", "PRON"} and token.dep_ == "compound":
             token = token.head
         return token
@@ -135,7 +133,6 @@ def __call__(self, doc: Doc):
             )
 
             for token in hyponym.conjuncts:
-
                 token_extended = self.expand_to_noun_compound(token, doc)
                 if token != hypernym and token is not None:
                     doc._.hearst_patterns.append(