We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: AGGGEN: Ordering and Aggregating while Generating

Abstract: We present AGGGEN (pronounced 'again'), a data-to-text model which re-introduces two explicit sentence planning stages into neural data-to-text systems: input ordering and input aggregation. In contrast to previous work using sentence planning, our model is still end-to-end: AGGGEN performs sentence planning at the same time as generating text by learning latent alignments (via semantic facts) between input representation and target text. Experiments on the WebNLG and E2E challenge data show that by using fact-based alignments our approach is more interpretable, expressive, robust to noise, and easier to control, while retaining the advantages of end-to-end systems in terms of fluency. Our code is available at this https URL
Comments: Correct the first citation in the Zero-shot Few-shot scenarios paragraph in Section 7
Subjects: Computation and Language (cs.CL)
Journal reference: Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL2021)
Cite as: arXiv:2106.05580 [cs.CL]
  (or arXiv:2106.05580v2 [cs.CL] for this version)

Submission history

From: Xinnuo Xu [view email]
[v1] Thu, 10 Jun 2021 08:14:59 GMT (1717kb,D)
[v2] Thu, 17 Jun 2021 07:53:44 GMT (1718kb,D)

Link back to: arXiv, form interface, contact.