import nltk
lemma = nltk.stem.WordNetLemmatizer()
w_tokenize = nltk.tokenize.WhitespaceTokenizer()

def converted_text(text):
    return[nltk.stem.WordNetLemmatizer().lemmatize(w, pos = 'n') for w in w_tokenize.tokenize(text)]

data['purpose_lemmatized'] = data['purpose'].apply(converted_text)