Multi-Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion-Cause Pair Extraction

Abstract

The Emotion-Cause Pair Extraction (ECPE) task aims to extract emotions and causes as pairs from documents. We observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typical ECPE dataset.Existing methods have set a fixed size window to capture relations between neighboring clauses. However, they neglect the effective semantic connections between distant clauses, leading to poor generalization ability towards positioninsensitive data. To alleviate the problem, we propose a novel Multi-Granularity Semantic Aware Graph model (MGSAG) to incorporate fine-grained and coarse-grained semantic features jointly, without regard to distance limitation. In particular, we first explore semantic dependencies between clauses and keywords extracted from the document that convey fine-grained semantic features, obtaining keywords enhanced clause representations. Besides, a clause graph is also established to model coarse-grained semantic relations between clauses. Experimental results indicate that MGSAG surpasses the existing state-ofthe-art ECPE models. Especially, MGSAG outperforms other models significantly in the condition of position-insensitive data.

From Coarse to Fine: A Distillation Method for Fine-Grained Emotion-Causal Span Pair Extraction in Conversation

“…Other works jointly extract emotions and causes in the ECPE task (Xia and Ding 2019) on document-level datasets (Gao et al 2017;Gui et al 2014), which was initially proposed as an utterance pairing task. MGSAG (Bao et al 2022) proposed to incorporate fine-grained and coarse-grained semantic features jointly without regard to distance limitation. UECA (Zheng et al 2022) designed prompt templates to predict is/isn't in a Question Answering format.…”

“…• The second set is constituted by methods achieved good performance in ECPE tasks. We select ECPE-2D (Ding, Xia, and Yu 2020), MGSAG (Bao et al 2022) and KEC (Li et al 2022). We make a simple modification to these models' final predictions, changing the utterancelevel classification task to a span extraction task.…”

“…These approachs required the construction of an emotion recognition module in a pipeline for real-world scenarios, which suffers from error propagation. (b) Joint identification of emotions and causes: such as KEC (Li et MGSAG (Bao et al 2022), UECA (Zheng et al 2022) and ECPE-2D (Ding, Xia, and Yu 2020), etc. These approachs performed coarse-grained Emotion-Cause Pair Extraction (ECPE), which have faced two major challenges.…”

Emotion Classification in Texts Over Graph Neural Networks: Semantic Representation is Better Than Syntactic

“…In recent years, semantically and syntactically aware models have gained popularity due to impressive performance in NLP problems [75], [76], [77], [78], [79], [80].…”

How to Enhance Causal Discrimination of Utterances: A Case on Affective Reasoning

“…Particularly in reasoning tasks (Uymaz and Metin, 2022;Feng et al, 2022), it is crucial for these methods to transcend the mere fitting of embeddings and possess the capacity to discriminate diverse causal relationships. (i.e., the ability of causal discrimination) (Bao et al, 2022;Shirai et al, 2023).…”