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

Addressing the issues of who saying what to whom in multi-party conversations(MPCs) has recently attracted a lot of research attention. However, existingmethods on MPC understanding typically embed interlocutors and utterances intosequential information flows, or utilize only the superficial of inherent graphstructures in MPCs. To this end, we present a plug-and-play and lightweightmethod named graph-induced fine-tuning (GIFT) which can adapt variousTransformer-based pre-trained language models (PLMs) for universal MPCunderstanding. In detail, the full and equivalent connections among utterancesin regular Transformer ignore the sparse but distinctive dependency of anutterance on another in MPCs. To distinguish different relationships betweenutterances, four types of edges are designed to integrate graph-induced signalsinto attention mechanisms to refine PLMs originally designed for processingsequential texts. We evaluate GIFT by implementing it into three PLMs, and testthe performance on three downstream tasks including addressee recognition,speaker identification and response selection. Experimental results show thatGIFT can significantly improve the performance of three PLMs on threedownstream tasks and two benchmarks with only 4 additional parameters perencoding layer, achieving new state-of-the-art performance on MPCunderstanding.

GIFT: Graph-Induced Fine-Tuning for Multi-Party Conversation Understanding