There are a couple of different ways. You can (1) improve the introduction (with providing more motivation and applications), (2) improve the proposed approaches with more examples, figures, and tables (3) adding more experiments and evaluating more benchmarks (4) providing insights on future research direction, problems with the current approach.

Before you do so be sure to verify the journal's policy on submissions. some require that the work is novel and its contribution be substantial.

Besides the answer that Mehdi Saeedi gave, an alternative in NLP is to provide an extension of a technique to another language. depending on the technique and the language, this can become pure original work which would be my main concern with some journals. Another approach would be to target a different domain (e.g. from medical terminology extend it to a more general class such as biology or a completely different domain).

The journal version needs to be rewritten and expanded.

In the expanded version, you need things that were not in the conference version. This may be one or more additional research questions, one or more forms of data collection, and/or an expanded "discussion" section. All three would ensure that it is not just the same content as you presented at the conference.

Extended version of conference paper should contains additional contribution or novelty in the proposed work. And also based on the journal policy the self plagiarism content should be maintained. Moreover, the results and discussion part should be projected more in journal when compared to conference paper.