chunked/nltk_chunking/model_doc_van/chunk_12.txt

In this paper, we propose a novel large kernel attention (LKA) module to enable self-adaptive and long-range correlations in self-attention while avoiding the above issues.