Problem
Research questions and friction points this paper is trying to address.
Target-Stance Extraction
sequential pipeline
neural models
From Two Passes to One: Compact and Efficient Target-Stance Extraction
๐ค AI Summary
ๆฌๆๆๅบไธ็งๅๆฌก้่ฟ็่ๅๆถๆ๏ผไปฅๆด็ดงๅๅ้ซๆ็ๆนๅผ่งฃๅณ็ฎๆ ็ซๅบๆๅ้ฎ้ข๏ผ็ธๆฏ็ฐๆไธค้ถๆฎตๆนๆณๅๅฐ่ฟ50%ๅๆฐใ
๐ Abstract
Target-Stance Extraction (TSE) is the task of predicting both the target (or topic) of an author's writing and the author's stance toward it. Existing approaches to TSE use a sequential pipeline of two separate neural models: one to identify the target and another to determine the stance. We present a one-pass, joint architecture that predicts both in a single forward pass, reducing trainable parameters by nearly 50% with only a 4-7 F1 point tradeoff in performance. We further demonstrate that standard target-scrubbing practices artificially suppress target prediction accuracy. Retaining explicit target mentions, as in real-world deployments, improves F1 by at least 6 points across both target classification and target generation settings. These improvements allow for significantly easier integration of TSE in downstream applications such as public opinion tracking.
Innovation
Methods, ideas, or system contributions that make the work stand out.
one-pass
joint architecture
target-stance extraction
parameter reduction
performance tradeoff
๐ Similar Papers
No similar papers found.
