Predictor values are always the name of the categories column

[amaiya]

Predictor values are always the name of the categories column

[Bidek56]

I believe you need to create 2 columns in your DF: ['postive', 'negative'] and set them to 0 or 1,

df['postive'] = df['labels'].apply(lambda x: 1 if x == 1 else 0)
df['negative'] = df['labels'].apply(lambda x: 1 if x == 0 else 0)

then call:

(x_train, y_train), (x_test, y_test), preproc = text.texts_from_df(
train_df=df, text_column='Source_NOFT', label_columns=['postive', 'negative'], max_features=MAX_FEATURES,
maxlen=MAX_LENGTH, ngram_range=3, preprocess_mode='standard')

then call:

predictor.predict(data)

[amaiya]

Thank you, @Bidek56

Yes, the labels are expected to be 1-hot-encoded. I have corrected the docstring for texts_from_csv and texts_from_df to make this clear.

In addition to @Bidek56 's answer, see also this example notebook for converting integer labels in a Pandas dataframe to be 1-hot-encoded.

[fcasadome]

Thank you very much for your quick responses @Bidek56 and @amaiya!
That makes sense I guess, I'm going to try right away

[fcasadome]

Works like a charm now!
Thank you very much again :)