Adjust for medium word list

The small wordlist doesn't have word vectors and that gives a warning
with Similarity is used. So instead use the medium word list.

Author: rocky

pymathics/natlang/__main__.py

@@ -256,7 +256,7 @@ def _load_spacy(self, evaluation, options):
             if "SPACY_DATA" in os.environ:
                 instance = spacy.load(language_code, via=os.environ["SPACY_DATA"])
             else:
-                instance = spacy.load(f"{language_code}_core_web_sm")
+                instance = spacy.load(f"{language_code}_core_web_md")
 
             _SpacyBuiltin._spacy_instances[language_code] = instance
             return instance
@@ -567,7 +567,7 @@ class TextCases(_SpacyBuiltin):
      = {London}
 
     >> TextCases[Import["ExampleData/EinsteinSzilLetter.txt"], "Person", 3][[2;;3]]
-     = {E. Fermi, L. Szilard}
+     = {L. Szilard, Joliot}
     """
 
     def apply_string_form(self, text, form, evaluation, options):
@@ -683,13 +683,13 @@ class WordSimilarity(_SpacyBuiltin):
     </dl>
 
     >> NumberForm[WordSimilarity["car", "train"], 3]
-     = 0.731
+     = 0.439
 
     >> NumberForm[WordSimilarity["car", "hedgehog"], 3]
-     = 0.302
+     = 0.195
 
     >> NumberForm[WordSimilarity[{"An ocean full of water.", {2, 2}}, { "A desert full of sand.", {2, 5}}], 3]
-     = {0.731, 0.317}
+     = {0.505, 0.481}
     """
 
     messages = _merge_dictionaries(

setup.py

@@ -25,6 +25,8 @@ def read(*rnames):
 # Get/set VERSION and long_description from files
 long_description = read("README.rst") + "\n"
 
+__version__ = "0.0.0"  # overwritten by exec below
+
 # stores __version__ in the current namespace
 exec(compile(open("pymathics/natlang/version.py").read(), "version.py", "exec"))
 
@@ -34,7 +36,7 @@ def read(*rnames):
 # Environment variables "lang", "WORDLIST_SIZE", and "SPACY_DOWNLOAD" override defaults.
 
 # Full package name with two-letter language code, e.g. fr, zh
-lang = os.environ.get("lang", "en_core_web_sm")
+lang = os.environ.get("lang", "en_core_web_md")
 
 # Size of wordlist used
 # sm=small, lg=large, md=medium.