How to use the markovify.chain.BEGIN function in markovify

only from sentences that start with the specified word/phrase.

        If strict == False, then markovify will draw its initial inspiration
        from any sentence containing the specified word/phrase.

        **kwargs are passed to `self.make_sentence`
        """
        split = tuple(self.word_split(beginning))
        word_count = len(split)

        if word_count == self.state_size:
            init_states = [ split ]

        elif word_count > 0 and word_count < self.state_size:
            if strict:
                init_states = [ (BEGIN,) * (self.state_size - word_count) + split ]

            else:
                init_states = [ key for key in self.chain.model.keys()
                    # check for starting with begin as well ordered lists
                    if tuple(filter(lambda x: x != BEGIN, key))[:word_count] == split ]

                random.shuffle(init_states)
        else:
            err_msg = "`make_sentence_with_start` for this model requires a string containing 1 to {0} words. Yours has {1}: {2}".format(self.state_size, word_count, str(split))
            raise ParamError(err_msg)

        for init_state in init_states:
            output = self.make_sentence(init_state, **kwargs)
            if output is not None:
                return output