Stance Detection and Open Research Avenues

2025-05-03 0 0 532.87KB 4 页 10玖币

侵权投诉

Dilek Küçük

dilek.kucuk@tubitak.gov.tr

TÜBİTAK Marmara Research Center

Ankara, Turkey

Fazli Can

canf@cs.bilkent.edu.tr

Bilkent University

Ankara, Turkey

ABSTRACT

This tutorial aims to cover the state-of-the-art on stance detection

and address open research avenues for interested researchers and

practitioners. Stance detection is a recent research topic where the

stance towards a given target or target set is determined based on

the given content and there are signicant application opportuni-

ties of stance detection in various domains. The tutorial comprises

two parts where the rst part outlines the fundamental concepts,

problems, approaches, and resources of stance detection, while the

second part covers open research avenues and application areas of

stance detection. The tutorial will be a useful guide for researchers

and practitioners of stance detection, social media analysis, infor-

mation retrieval, and natural language processing.

CCS CONCEPTS

•Computing methodologies →Natural language processing

;

Machine learning

;

Language resources

;

•Information sys-

tems →Information retrieval

;

Web and social media search

;

Sentiment analysis.

KEYWORDS

stance detection, aective computing, sentiment analysis, social

media analysis, data streams, stance quantication

1 INTRODUCTION

Stance detection is a research problem focusing on people’s posi-

tions towards specic targets, in natural language texts [

–

]. It

can be considered as a subproblem of aective computing, along

with related research topics such as sentiment analysis. Stance clas-

sication, stance analysis, and stance extraction are also used to

refer to the problem of stance detection in the related literature.

One of the important milestones of stance detection research is

the stance detection shared task on English tweets performed in

2016 [

]. Within the course of this competition, a stance-annotated

tweet dataset is created and publicly shared [

]. This shared task

is followed by other similar competitions on texts in Chinese [

Spanish and Catalan [26], Italian [7], and Basque [1].

Signicant subproblems of stance detection and closely-related

problems have been previously discussed in details in [

–

Based on the related gure in [

], a revised schematic represen-

tation demonstrating stance detection, its subproblems, and re-

lated problems is given in Figure 1. The newly-added problems are

contextual stance detection [

], intent detection [

], and stance

quantication [13], which are shown in blue in the gure.

Contextual stance detection does not only use the input text, but

makes use of contextual information (social media user proles,

interactions between posts, etc.) as well during the stance detection

procedure. Hence, contextual stance detection is a subproblem of

Figure 1: Stance Detection, Its Subproblems, and Related Re-

search Problems (Revised Version of the Corresponding Fig-

ure in [14]).

generic stance detection. Intent detection (together with slot lling),

on the other hand, is a research problem of dialogue systems and

natural language understanding, where the goals of the users are

extracted from their utterances [

]. Stance quantication aims the

determine the percentages of the textual items belonging to distinct

stance classes, instead of labeling each item with its stance label

[

]. We aim to cover all of the research problems in Figure 1 during

our tutorial.

Most of the recent work on stance detection employs deep learn-

ing approaches [

]. Yet, ensemble methods are also commonly

observed in stance detection research [6].

It has been previously reported that stance detection studies were

published on several languages including English, Chinese, Spanish,

Catalan, Italian, Japanese, Turkish, Czech, Russian, and Arabic [

Recent work also reveal that there are related studies performed on

Basque [1], German [18], Portuguese [27], and Persian [22], too.

This stance detection tutorial will consist of two main parts

where the rst part will be devoted to presentation of basic con-

cepts, related research problems, stance detection competitions,

machine learning and deep learning based approaches, and stance

detection resources like the datasets, with particular emphasis on

publicly shared datasets. The second part of the tutorial will mostly

cover open research topics related to stance detection. Basically

the following open research topics will be considered in the second

part:

(1)

Stance Detection in Data Streams: Data streams constitute a

signicant research area [

] and stance detection can

also be applied to large volumes of particularly social media

arXiv:2210.12383v1 [cs.CL] 22 Oct 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

StanceDetectionandOpenResearchAvenuesDilekKüçükdilek.kucuk@tubitak.gov.trTÜBİTAKMarmaraResearchCenterAnkara,TurkeyFazliCancanf@cs.bilkent.edu.trBilkentUniversityAnkara,TurkeyABSTRACTThistutorialaimstocoverthestate-of-the-artonstancedetectionandaddressopenresearchavenuesforinterestedresearchersandpra...

展开>> 收起<<

Stance Detection and Open Research Avenues.pdf

共4页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Stance Detection and Open Research Avenues

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: