diff --git a/scispacy/candidate_generation.py b/scispacy/candidate_generation.py
index d4b42c0..d42877a 100644
--- a/scispacy/candidate_generation.py
+++ b/scispacy/candidate_generation.py
@@ -244,7 +244,7 @@ def __init__(
             [ann_index, tfidf_vectorizer, ann_concept_aliases_list, kb]
         ):
             raise ValueError(
-                "You cannot pass both a name argument and other constuctor arguments."
+                "You cannot pass both a name argument and other constructor arguments."
             )
 
         # Set the name to the default, after we have checked
@@ -466,11 +466,11 @@ def create_tfidf_ann_index(
     empty_tfidfs_boolean_flags = numpy.array(
         concept_alias_tfidfs.sum(axis=1) != 0
     ).reshape(-1)
-    number_of_non_empty_tfidfs = sum(empty_tfidfs_boolean_flags == False)  # noqa: E712
+    number_of_empty_tfidfs = sum(empty_tfidfs_boolean_flags == False)  # noqa: E712
     total_number_of_tfidfs = numpy.size(concept_alias_tfidfs, 0)
 
     print(
-        f"Deleting {number_of_non_empty_tfidfs}/{total_number_of_tfidfs} aliases because their tfidf is empty"
+        f"Deleting {number_of_empty_tfidfs}/{total_number_of_tfidfs} aliases because their tfidf is empty"
     )
     # remove empty tfidf vectors, otherwise nmslib will crash
     concept_aliases = [
diff --git a/scispacy/data_util.py b/scispacy/data_util.py
index 87cd206..6e6f563 100644
--- a/scispacy/data_util.py
+++ b/scispacy/data_util.py
@@ -183,18 +183,12 @@ def _cleanup_dir(dir_path: str):
     corpus = os.path.join(resolved_directory_path, expected_names[0])
     examples = med_mentions_example_iterator(corpus)
 
-    train_ids = {
-        x.strip()
-        for x in open(os.path.join(resolved_directory_path, expected_names[4]))
-    }
-    dev_ids = {
-        x.strip()
-        for x in open(os.path.join(resolved_directory_path, expected_names[2]))
-    }
-    test_ids = {
-        x.strip()
-        for x in open(os.path.join(resolved_directory_path, expected_names[3]))
-    }
+    with open(os.path.join(resolved_directory_path, expected_names[4])) as train_file:
+        train_ids = {x.strip() for x in train_file}
+    with open(os.path.join(resolved_directory_path, expected_names[2])) as dev_file:
+        dev_ids = {x.strip() for x in dev_file}
+    with open(os.path.join(resolved_directory_path, expected_names[3])) as test_file:
+        test_ids = {x.strip() for x in test_file}
 
     train_examples = []
     dev_examples = []
@@ -299,19 +293,20 @@ def read_ner_from_tsv(filename: str) -> List[SpacyNerExample]:
     """
     spacy_format_data = []
     examples: List[Tuple[str, str]] = []
-    for line in open(cached_path(filename)):
-        line = line.strip()
-        if line.startswith("-DOCSTART-"):
-            continue
-        # We have reached the end of a sentence.
-        if not line:
-            if not examples:
+    with open(cached_path(filename)) as tsv_file:
+        for line in tsv_file:
+            line = line.strip()
+            if line.startswith("-DOCSTART-"):
                 continue
-            spacy_format_data.append(_handle_sentence(examples))
-            examples = []
-        else:
-            word, entity = line.split("\t")
-            examples.append((word, entity))
+            # We have reached the end of a sentence.
+            if not line:
+                if not examples:
+                    continue
+                spacy_format_data.append(_handle_sentence(examples))
+                examples = []
+            else:
+                word, entity = line.split("\t")
+                examples.append((word, entity))
     if examples:
         spacy_format_data.append(_handle_sentence(examples))
 
diff --git a/scispacy/file_cache.py b/scispacy/file_cache.py
index 9ff9918..4239164 100644
--- a/scispacy/file_cache.py
+++ b/scispacy/file_cache.py
@@ -60,12 +60,12 @@ def url_to_filename(url: str, etag: Optional[str] = None) -> str:
     last_part = url.split("/")[-1]
     url_bytes = url.encode("utf-8")
     url_hash = sha256(url_bytes)
-    filename = url_hash.hexdigest()
+    filename = url_hash.hexdigest()[:16]
 
     if etag:
         etag_bytes = etag.encode("utf-8")
         etag_hash = sha256(etag_bytes)
-        filename += "." + etag_hash.hexdigest()
+        filename += "." + etag_hash.hexdigest()[:16]
 
     filename += "." + last_part
     return filename
diff --git a/scispacy/hearst_patterns.py b/scispacy/hearst_patterns.py
index 992b7a4..c28b16a 100644
--- a/scispacy/hearst_patterns.py
+++ b/scispacy/hearst_patterns.py
@@ -492,7 +492,7 @@
             {"LEMMA": "which"},
             {"LEMMA": "look"},
             {"LEMMA": "like"},
-            hyponym,
+            hypernym,
         ],
         "position": "last",
     },
diff --git a/scispacy/per_class_scorer.py b/scispacy/per_class_scorer.py
index 57993b0..ade6af8 100644
--- a/scispacy/per_class_scorer.py
+++ b/scispacy/per_class_scorer.py
@@ -18,9 +18,9 @@ def __call__(
         gold_spans = copy.copy(gold_spans)
         predicted_spans = copy.copy(predicted_spans)
         untyped_gold_spans = {(x[0], x[1]) for x in gold_spans}
-        untyped_predicted_spans = {(x[0], x[1]) for x in predicted_spans}
 
-        for untyped_span, span in zip(untyped_predicted_spans, predicted_spans):
+        for span in predicted_spans:
+            untyped_span = (span[0], span[1])
             if span in gold_spans:
                 self._true_positives[span[2]] += 1
                 gold_spans.remove(span)
diff --git a/scispacy/train_utils.py b/scispacy/train_utils.py
index f02fd18..5ee826d 100644
--- a/scispacy/train_utils.py
+++ b/scispacy/train_utils.py
@@ -26,7 +26,8 @@ def evaluate_ner(
 
     metrics = scorer.get_metric()
     if dump_path is not None:
-        json.dump(metrics, open(dump_path, "a+"))
+        with open(dump_path, "a+") as metrics_file:
+            json.dump(metrics, metrics_file)
     for name, metric in metrics.items():
         if "overall" in name or "untyped" in name or verbose:
             print(f"{name}: \t\t {metric}")