Fixing serialization error and span label issue

[makcedward]

This PR is going to fix two issues when using this library

• Unable to serialize the list of Tuple variables. Converting it to list of dict

File [~/.venv/lib/python3.9/site-packages/spacy/tokens/_serialize.py:154], in DocBin.get_docs(self, vocab)
    152 # backwards-compatibility: may be b'' or serialized empty list
    153 if self.span_groups[i] and self.span_groups[i] != SpanGroups._EMPTY_BYTES:
--> 154     doc.spans.from_bytes(self.span_groups[i])
    155 else:
    156     doc.spans.clear()

File [~/.venv/lib/python3.9/site-packages/spacy/tokens/_dict_proxies.py:97], in SpanGroups.from_bytes(self, bytes_data)
     95 else:
     96     for value_bytes, keys in msg.items():
---> 97         group = SpanGroup(doc).from_bytes(value_bytes)
     98         # Deserialize `SpanGroup`s as copies because it's possible for two
     99         # different `SpanGroup`s (pre-serialization) to have the same bytes
    100         # (since they can have the same `.name`).
    101         self[keys[0]] = group

File [~/.venv/lib/python3.9/site-packages/spacy/tokens/span_group.pyx:238], in spacy.tokens.span_group.SpanGroup.from_bytes()

File [~/.venv/lib/python3.9/site-packages/srsly/_msgpack_api.py:27], in msgpack_loads(data, use_list)
     25 # msgpack-python docs suggest disabling gc before unpacking large messages
     26 gc.disable()
---> 27 msg = msgpack.loads(data, raw=False, use_list=use_list)
     28 gc.enable()
     29 return msg

File [~/.venv/lib/python3.9/site-packages/srsly/msgpack/__init__.py:79], in unpackb(packed, **kwargs)
     77         object_hook = functools.partial(decoder, chain=object_hook)
     78     kwargs["object_hook"] = object_hook
---> 79 return _unpackb(packed, **kwargs)

File [~/.venv/lib/python3.9/site-packages/srsly/msgpack/_unpacker.pyx:191], in srsly.msgpack._unpacker.unpackb()

TypeError: unhashable type: 'list'

• Unable to recongize spaCy's span label. Using span.label instead of span.label_

File [~/.venv/lib/python3.9/site-packages/skweak/aggregation.py:126], in AbstractAggregator.fit(self, docs, **kwargs)
    121 """Fits the parameters of the aggregator model based on a collection
    122 of documents. The method extracts a dataframe of observations for
    123 each document and calls the _fit method"""
    125 obs_generator = (self.get_observation_df(doc) for doc in docs)
--> 126 self._fit(obs_generator, **kwargs)

File [~/.venv/lib/python3.9/site-packages/skweak/generative.py:98], in GenerativeModelMixin._fit(self, all_obs, cutoff, n_iter, tol)
     95 self._reset_counts(sources)
     97 # And add the counts from majority voter
---> 98 self._add_mv_counts(all_obs)
    100 # Finally, we postprocess the counts and get probabilities
    101 self._do_mstep_latent()

File [~/.venv/lib/python3.9/site-packages/skweak/generative.py:419], in NaiveBayes._add_mv_counts(self, all_obs)
    416 mv.label_groups = self.label_groups
    417 for obs in all_obs:
    418     # And aggregate the results
--> 419     agg_array = mv.aggregate(obs).values
    421     if len(agg_array)==0:
    422         continue

File [~/.venv/lib/python3.9/site-packages/skweak/voting.py:51], in MajorityVoterMixin.aggregate(self, obs)
     48 def count_fun(x):
     49     return np.bincount(x[x>=0], weights=weights[x>=0], 
     50                        minlength=len(self.observed_labels)) 
---> 51 label_votes = np.apply_along_axis(count_fun, 1, obs.values).astype(np.float32)
     53 # For token-level sequence labelling, we need to normalise the number 
     54 # of "O" occurrences, since they both indicate the absence of 
     55 # prediction, but are also a possible output
     56 if self.observed_labels[0]=="O":

File <__array_function__ internals>:180, in apply_along_axis(*args, **kwargs)

File [~/.venv/lib/python3.9/site-packages/numpy/lib/shape_base.py:376], in apply_along_axis(func1d, axis, arr, *args, **kwargs)
    374     ind0 = next(inds)
    375 except StopIteration as e:
--> 376     raise ValueError(
    377         'Cannot apply_along_axis when any iteration dimensions are 0'
    378     ) from None
    379 res = asanyarray(func1d(inarr_view[ind0], *args, **kwargs))
    381 # build a buffer for storing evaluations of func1d.
    382 # remove the requested axis, and add the new ones on the end.
    383 # laid out so that each write is contiguous.
    384 # for a tuple index inds, buff[inds] = func1d(inarr_view[inds])

ValueError: Cannot apply_along_axis when any iteration dimensions are 0