import numpy as np

def text2vec(doc_tok, model, dim=300):
    doc_embedding = np.zeros(dim)
    valid_words = 0
    for word in doc_tok:
        if word in model:
            valid_words += 1
            doc_embedding += model.query(word)
        else:
            continue
    if valid_words > 0:
        return doc_embedding / valid_words
    else:
        return doc_embedding

def get_docs_embedding(docs_tok, model, dim=300):
    all_docs_embedding = []
    for doc in docs_tok:
        all_docs_embedding.append(text2vec(doc, model, dim))
    return np.array(all_docs_embedding)

headlines_embedding = get_docs_embedding(docs_tok=data['processed_headlines'], model=em_model)