Syntax Dependency and Semantic Enhancement for Aspect-Based Sentiment Analysis

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F25%3A8QLNXWBP" target="_blank" >RIV/00216208:11320/25:8QLNXWBP - isvavai.cz</a>

Výsledek na webu

<a href="https://www.scopus.com/inward/record.uri?eid=2-s2.0-85205473466&doi=10.23919%2fCCC63176.2024.10661913&partnerID=40&md5=2beae47923a9e41111e5a454a42bcc6e" target="_blank" >https://www.scopus.com/inward/record.uri?eid=2-s2.0-85205473466&doi=10.23919%2fCCC63176.2024.10661913&partnerID=40&md5=2beae47923a9e41111e5a454a42bcc6e</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.23919/CCC63176.2024.10661913" target="_blank" >10.23919/CCC63176.2024.10661913</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Syntax Dependency and Semantic Enhancement for Aspect-Based Sentiment Analysis

Popis výsledku v původním jazyce

Aspect-based sentiment analysis is a fine-grained sentiment analysis task. In this domain, graph neural models based on dependency trees are widely used. However, effectively utilizing the semantic and syntactic structural information of dependency trees remains a challenging research problem. To address this issue, this paper proposes a Syntax Dependency and Semantics Enhancement (SDSE) model. This model aims to enhance the understanding of specific aspect-related syntactic dependency relations and sentence semantics. Specifically, the SDSE model integrates self-attention mechanisms and aspect-aware attention mechanisms to obtain the attention score matrix of the sentence. Then, leveraging graph convolutional networks on the attention score matrix, the model extracts semantic feature information of the sentence. This approach not only learns semantic associations related to aspects but also captures the overall semantic information of the sentence. Moreover, to mitigate dependency parsing errors, the SDSE model introduces aspect word merging. This involves parsing the sentence after merging aspect words to obtain the merged syntactic dependency graph, thereby enhancing the model's focus on opinion entities. Finally, the dependency graph and processed sentence encodings are fed into graph convolutional networks for training. Experimental results demonstrate that our proposed SDSE model outperforms the current state-of-the-art methods on benchmark datasets. © 2024 Technical Committee on Control Theory, Chinese Association of Automation.

Název v anglickém jazyce

Syntax Dependency and Semantic Enhancement for Aspect-Based Sentiment Analysis

Popis výsledku anglicky

Aspect-based sentiment analysis is a fine-grained sentiment analysis task. In this domain, graph neural models based on dependency trees are widely used. However, effectively utilizing the semantic and syntactic structural information of dependency trees remains a challenging research problem. To address this issue, this paper proposes a Syntax Dependency and Semantics Enhancement (SDSE) model. This model aims to enhance the understanding of specific aspect-related syntactic dependency relations and sentence semantics. Specifically, the SDSE model integrates self-attention mechanisms and aspect-aware attention mechanisms to obtain the attention score matrix of the sentence. Then, leveraging graph convolutional networks on the attention score matrix, the model extracts semantic feature information of the sentence. This approach not only learns semantic associations related to aspects but also captures the overall semantic information of the sentence. Moreover, to mitigate dependency parsing errors, the SDSE model introduces aspect word merging. This involves parsing the sentence after merging aspect words to obtain the merged syntactic dependency graph, thereby enhancing the model's focus on opinion entities. Finally, the dependency graph and processed sentence encodings are fed into graph convolutional networks for training. Experimental results demonstrate that our proposed SDSE model outperforms the current state-of-the-art methods on benchmark datasets. © 2024 Technical Committee on Control Theory, Chinese Association of Automation.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

—

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Chinese Control Conf., CCC

ISBN

978-988758158-1

ISSN

1934-1768

e-ISSN

—

Počet stran výsledku

6

Strana od-do

8224-8229

Název nakladatele

IEEE Computer Society

Místo vydání

—

Místo konání akce

Kunming

Datum konání akce

1. 1. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—