fix: Sentence bounderies are now considered when check negations and …

…intensifiers

asent/about.py

@@ -1,5 +1,5 @@
 __title__ = "asent"
 
-__version__ = "0.4.3"  # the ONLY source of version ID
+__version__ = "0.5.4"  # the ONLY source of version ID
 __download_url__ = "https://github.com/kennethenevoldsen/asent"
 __documentation__ = "https://kennethenevoldsen.github.io/asent"

asent/getters.py

@@ -223,14 +223,16 @@ def make_is_negated_getter(
 
     def is_negated_getter(token: Token) -> bool:
         """Determine if token is negated."""
-        for t in token.doc[token.i - lookback : token.i]:
+        # don't look back before the sentence start
+        min_token_idx = max(token.sent.start, token.i - lookback)
+        for t in token.doc[min_token_idx : token.i]:
             if t._.is_negation:
                 return t
 
     return is_negated_getter
 
 
-def make_token_polarity_getter(
+def make_token_polarity_getter(  # noqa: C901
     valence_getter: Optional[Callable[[Token], float]] = None,
     is_negated_getter: Optional[Callable[[Token], Union[bool, Optional[Token]]]] = None,
     intensifier_getter: Optional[Callable[[Token], float]] = None,
@@ -306,8 +308,14 @@ def token_polarity_getter(
                 # dampen the scalar modifier of preceding words and emoticons
                 # (excluding the ones that immediately preceed the item) based
                 # on their distance from the current item.
+
                 if token.i > start_i:
-                    prev_token = token.doc[token.i - start_i]
+                    tok_id = token.i - start_i
+                    # stop if before sentence start
+                    if tok_id < token.sent.start:
+                        break
+
+                    prev_token = token.doc[tok_id]
                     b = prev_token._.intensifier
                     if b != 0:
                         intensifiers.append(prev_token)
@@ -409,7 +417,7 @@ def sift_sentiment_scores(sentiments: Iterable[float]) -> Tuple[float, float, in
     return pos_sum, neg_sum, neu_count
 
 
-def make_span_polarity_getter(
+def make_span_polarity_getter(  # noqa: C901
     polarity_getter: Optional[Callable[[Token], float]],
     contrastive_conj_getter: Optional[Callable[[Token], bool]],
 ) -> SpanPolarityOutput:

docs/news.rst

@@ -1,6 +1,11 @@
 News and Changelog
 ==============================
 
+* 0.5.3 (26/05/22)
+
+  - Fixed bug where negation and intensifiers were considered outside of the sentences boundaries. Adressing `58 <https://github.com/KennethEnevoldsen/asent/issues/58>`__.
+  - Improvements to the documentation. Thanks to @tomaarsen for the pull request.
+
 * 0.4.2 (28/05/22)
 
   - Added new Danish Dictionary from `AFINN <https://github.com/fnielsen/afinn>`__

tests/test_bugs.py

@@ -0,0 +1,26 @@
+"""Test specifically targeted an bugs."""
+
+import spacy
+
+import asent  # noqa
+
+
+def test_no_negations_and_intensifiers_out_of_sentence():
+    """Test that no negations are not found outside the sentence span.
+
+    https://github.com/KennethEnevoldsen/asent/issues/58
+    """
+
+    # create spacy pipeline
+    nlp = spacy.blank("en")
+    nlp.add_pipe("sentencizer")
+
+    nlp.add_pipe("asent_en_v1")
+
+    text = "Would you do that? I would not. Very stupid is what that is."
+    doc = nlp(text)
+    assert doc[10]._.is_negated is None
+
+    text = "Would you do that? I would not very. Stupid is what that is."
+    doc = nlp(text)
+    assert doc[10]._.valence == doc[10]._.polarity.polarity