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: CaM-Gen:Causally-aware Metric-guided Text Generation

Abstract: Content is created for a well-defined purpose, often described by a metric or a signal represented in the form of structured information. The relationship between the metrics or the goal of a target content and the content itself are non-trivial. While large scale language models show promising text generation capabilities, guiding and informing the generated text with external metrics is challenging. These metrics and the content tend to have inherent relationships and not all of them may directly impact the content. We introduce a CaM-Gen: Causally-aware Generative Networks guided by user-defined input metrics incorporating the causal relationships between the metric and the content features. We leverage causal inference techniques to identify the causally significant aspects of text that leads to the target metric and then explicitly guide the generative model towards these by a feedback mechanism. We propose this mechanism for variational autoencoder-based and transformer-based generative models. The proposed models beat baselines in terms of the target metric accuracy while maintaining the fluency and the language quality of the generated text. To the best of our knowledge, this is one of the early attempts at incorporating a metric-guide using causal inference towards controlled generation.
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2010.12795 [cs.CL]
  (or arXiv:2010.12795v1 [cs.CL] for this version)

Submission history

From: Navita Goyal [view email]
[v1] Sat, 24 Oct 2020 06:17:35 GMT (707kb,D)

Link back to: arXiv, form interface, contact.