Adding a self build vocabulary in scikit-learn?

Question

In sklearn.feature_extraction.text.TfidfVectorizer, we can inject our own vocabulary using vocabulary parameter of the model. but in this case only my own selected words are used for the model.

I want to use automatically detected features with my custom vocabulary.

One way to solve this problem is to create the model and get the features using

vocab=vectorizer.get_feature_names()

appending my list on vocab

vocab + vocabulary

and again build the model.

Is there a way to perform this whole process in a single step?

Answer 1

I don't think there is a simpler way than that to achieve what you want. One thing you can do is to use the code of CountVectorizer used to create the vocabulary. I went through the source code and the method is

_count_vocab(self, raw_documents, fixed_vocab)

called with fixed_vocab=False.

So what I suggest is for you to adapt the following code (Source) to create the vocabulary before you run the TfidfVectorizer.

def _count_vocab(self, raw_documents, fixed_vocab):
        """Create sparse feature matrix, and vocabulary where fixed_vocab=False
        """
        if fixed_vocab:
            vocabulary = self.vocabulary_
        else:
            # Add a new value when a new vocabulary item is seen
            vocabulary = defaultdict()
            vocabulary.default_factory = vocabulary.__len__

        analyze = self.build_analyzer()
        j_indices = _make_int_array()
        indptr = _make_int_array()
        indptr.append(0)
        for doc in raw_documents:
            for feature in analyze(doc):
                try:
                    j_indices.append(vocabulary[feature])
                except KeyError:
                    # Ignore out-of-vocabulary items for fixed_vocab=True
                    continue
            indptr.append(len(j_indices))

        if not fixed_vocab:
            # disable defaultdict behaviour
            vocabulary = dict(vocabulary)
            if not vocabulary:
                raise ValueError("empty vocabulary; perhaps the documents only"
                                 " contain stop words")

        j_indices = frombuffer_empty(j_indices, dtype=np.intc)
        indptr = np.frombuffer(indptr, dtype=np.intc)
        values = np.ones(len(j_indices))

        X = sp.csr_matrix((values, j_indices, indptr),
                          shape=(len(indptr) - 1, len(vocabulary)),
                          dtype=self.dtype)
        X.sum_duplicates()
        return vocabulary, X