English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Learning disentangled representations of textual data is essential for manynatural language tasks such as fair classification, style transfer and sentencegeneration, among others. The existent dominant approaches in the context oftext data {either rely} on training an adversary (discriminator) that aims atmaking attribute values difficult to be inferred from the latent code {or relyon minimising variational bounds of the mutual information between latent codeand the value attribute}. {However, the available methods suffer of theimpossibility to provide a fine-grained control of the degree (or force) ofdisentanglement.} {In contrast to} {adversarial methods}, which are remarkablysimple, although the adversary seems to be performing perfectly well during thetraining phase, after it is completed a fair amount of information about theundesired attribute still remains. This paper introduces a novel variationalupper bound to the mutual information between an attribute and the latent codeof an encoder. Our bound aims at controlling the approximation error via theRenyi's divergence, leading to both better disentangled representations and inparticular, a precise control of the desirable degree of disentanglement {thanstate-of-the-art methods proposed for textual data}. Furthermore, it does notsuffer from the degeneracy of other losses in multi-class scenarios. We showthe superiority of this method on fair classification and on textual styletransfer tasks. Additionally, we provide new insights illustrating varioustrade-offs in style transfer when attempting to learn disentangledrepresentations and quality of the generated sentence.

A Novel Estimator of Mutual Information for Learning to Disentangle Textual Representations