Getting an error when training a new text classifier - gold = GoldParse(doc, **gold)

I am trying to train a custom dataset from Kaggle after successfully stepping through the spacy training example for the IMDB dataset for a text classifier. https://raw.githubusercontent.com/explosion/spaCy/master/examples/training/train_textcat.py

I am however getting an error when performing the training. My custom training dataset can be downloaded from here: https://www.kaggle.com/nicapotato/womens-ecommerce-clothing-reviews/data

Here is the code that I altered.

#train_data, _ = thinc.extra.datasets.imdb()
#Custom dataset
df = pd.read_csv("~/notebooks/data/ecom_reviews.csv")
train_data = df.apply(lambda row: (row['Review Text'],row['Recommended IND']), axis=1).tolist()
#End custom dataset

It seems to me that the train_data is exactly the same shape and yet I am getting an error when switching to my custom dataset compared to the imdb dataset.

Here is the full error.

Training the model...
LOSS 	  P  	  R  	  F  
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-19-2bef4fe1e613> in <module>()
     12                 texts, annotations = zip(*batch)
     13                 nlp.update(texts, annotations, sgd=optimizer, drop=0.2,
---> 14                            losses=losses)
     15             with textcat.model.use_params(optimizer.averages):
     16                 # evaluate on the dev data split off in load_data()

~/anaconda2/envs/ipykernel_py3/lib/python3.6/site-packages/spacy/language.py in update(self, docs, golds, drop, sgd, losses)
    397                 doc = self.make_doc(doc)
    398             if not isinstance(gold, GoldParse):
--> 399                 gold = GoldParse(doc, **gold)
    400             doc_objs.append(doc)
    401             gold_objs.append(gold)

gold.pyx in spacy.gold.GoldParse.__init__()

TypeError: 'float' object is not iterable
============================================================================

Your Environment

Info about spaCy

• spaCy version: 2.0.9
• Platform: Linux-4.4.0-21-generic-x86_64-with-debian-stretch-sid
• Python version: 3.6.3
• Models: en

[sublimotion]

My bad the data needed to be cleansed.

After dropping the null values I did not run into any issues.

I am pretty happy with the results. This model classifies the review if it has been recommended or not. It can then be used in turn to predict which review is most likely to be recommended.

Training the model… LOSS P R F
52.900 0.839 0.991 0.909 34.806 0.870 0.964 0.914 23.225 0.895 0.946 0.919 15.268 0.905 0.943 0.923 10.861 0.898 0.931 0.914 8.093 0.905 0.943 0.923 7.451 0.906 0.934 0.920 4.859 0.906 0.934 0.920 4.395 0.906 0.934 0.920 3.798 0.913 0.919 0.916 2.978 0.911 0.928 0.919 2.894 0.912 0.937 0.924 2.514 0.906 0.925 0.915 2.595 0.901 0.931 0.916 1.843 0.901 0.931 0.916 2.508 0.902 0.919 0.910 1.950 0.905 0.916 0.910 1.544 0.905 0.916 0.910 1.742 0.904 0.934 0.919 1.895 0.893 0.934 0.913

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.