arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2101.08248cs.CLcs.LG
收藏

通过拼接最近邻的数据到文本生成

Data-to-text Generation by Splicing Together Nearest Neighbors

TTI-Chicago(芝加哥丰田技术研究所);Duke University(杜克大学);Microsoft Research(微软研究院);Rutgers University(罗格斯大学)

Sam Wiseman, Arturs Backurs, Karl Stratos

摘要:

我们建议直接拼接在一起检索的文本段从“邻居”源目标对来处理数据到文本生成任务。不像最近的工作,条件检索邻居,但生成文本令牌的令牌,从左到右,我们学习的政策,直接操纵段的邻居文本,通过插入或替换它们在部分构建的代。标准的技术训练这样的政策需要一个甲骨文推导每一代,我们证明,找到最短的这种推导可以减少到一个特定的加权上下文无关文法下的解析。我们发现,以这种方式学习的政策在自动和人工评估方面与强基线不相上下,但允许更可解释和可控的生成。

英文摘要:

We propose to tackle data-to-text generation tasks by directly splicing together retrieved segments of text from "neighbor" source-target pairs. Unlike recent work that conditions on retrieved neighbors but generates text token-by-token, left-to-right, we learn a policy that directly manipulates segments of neighbor text, by inserting or replacing them in partially constructed generations. Standard techniques for training such a policy require an oracle derivation for each generation, and we prove that finding the shortest such derivation can be reduced to parsing under a particular weighted context-free grammar. We find that policies learned in this way perform on par with strong baselines in terms of automatic and human evaluation, but allow for more interpretable and controllable generation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑