Python spacy issue when training :ValueError: Found array with 0 feature(s) (shape=(n, 0)) while a minimum of 1 is required by SVC

Question

I am running into a value error that my array is incorrect, which is extremely strange since I have confirmed that my array is not zero. I printed out the 'len' of each feature and training set. Found array with 0 feature(s) (shape=(7, 0)) while a minimum of 1 is required by SVC. I am using spacy 3.4.1 and python 3.8.10. What am I doing wrong?

import spacy
from sklearn import svm

nlp = spacy.load("en_core_web_trf")

train_x = [
        "good characters and plot progression", 
        "check out the book", 
        "good story. would recommend", 
        "novel recommendation", 
        "need to make a deposit to the bank", 
        "balance inquiry savings", 
        "save money"
        ]


train_y = [
            "BOOKS", 
            "BOOKS", 
            "BOOKS", 
            "BOOKS", 
            "BANK", 
            "BANK", 
            "BANK", 
            ]


docs = [nlp(text) for text in train_x]
train_x_vectors = [doc.vector for doc in docs]

print (len(train_x_vectors))
print (len(train_y))

clf_svm = svm.SVC(kernel='linear')
clf_svm.fit(train_x_vectors, train_y)

`Doc.vector` is going to be empty with trf pipelines in spaCy by default. Are you sure you're not pssing empty vectors? — polm23, Oct 17 '22 at 03:57

score 1 · Answer 1 · answered Dec 05 '22 at 10:18

1

The spacy pipeline package "en_core_web_trf" does not come with word vectors you need to use en_core_web_lg in order to utilize ".vector"

answered Dec 05 '22 at 10:18

Rolito

11
1

Python spacy issue when training :ValueError: Found array with 0 feature(s) (shape=(n, 0)) while a minimum of 1 is required by SVC

1 Answers1