Unisa at SemEval-2023 Task 3: A SHAP-based method for Propaganda Detection

被引：0

作者：

Bangerter, Micaela ^{[1
]}

Fenza, Giuseppe ^{[1
]}

Gallo, Mariacristina ^{[1
]}

Loia, Vincenzo ^{[1
]}

Volpe, Alberto ^{[1
]}

De Maio, Carmen ^{[2
]}

Stanzione, Claudio ^{[3
]}

机构：

[1] Univ Salerno, Dept Management & Innovat Syst, I-84084 Fisciano, SA, Italy

[2] Univ Salerno, Dept Informat Engn Elect Engn & Appl Math, I-84084 Fisciano, SA, Italy

[3] Ctr Higher Def Studies, Def Anal & Res Inst, I-00165 Rome, RM, Italy

来源：

17TH INTERNATIONAL WORKSHOP ON SEMANTIC EVALUATION, SEMEVAL-2023 | 2023年

关键词：

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

This paper presents proposed solutions for addressing two subtasks in SemEval-2023 Task 3: "Detecting the Genre, the Framing, and the Persuasion techniques in online news in a multilingual setup". In subtask 1, "News Genre Categorisation", the goal is to classify a news article as an opinion, a report, or a satire. In subtask 3, "Detection of Persuasion Technique", the system must reveal persuasion techniques used in each news article paragraph choosing among 23 defined methods. Solutions leverage the application of the eXplainable Artificial Intelligence (XAI) method, Shapley Additive Explanations (SHAP). In subtask 1, SHAP was used to understand what was driving the model to fail so that it could be improved accordingly. In contrast, in subtask 3, a re-calibration of the Attention Mechanism was realized by extracting critical tokens for each persuasion technique. The underlying idea is the exploitation of XAI for countering the overfitting of the resulting model and attempting to improve the performance when there are few samples in the training data. The achieved performance on English for subtask 1 ranked 6th with an F1-score of 58.6% (despite 78.4% of the 1st) and for subtask 3 ranked 12th with a micro-averaged F1-score of 29.8% (despite 37.6% of the 1st).

引用

页码：885 / 891

页数：7

共 17 条

[1] Alhindi Tariq, 2020, P 28 INT C COMP LING, P6139
[2] BERT Pre-processed Deep Learning Model for Sarcasm Detection
Bhardwaj, Saumya
Prusty, Manas Ranjan
[J]. NATIONAL ACADEMY SCIENCE LETTERS-INDIA, 2022, 45 (02): : 203 - 208
[3] Transforming Fake News: Robust Generalisable News Classification Using Transformers
Blackledge, Ciara
Atapour-Abarghouei, Amir
[J]. 2021 IEEE INTERNATIONAL CONFERENCE ON BIG DATA (BIG DATA), 2021, : 3960 - 3968
[4] Chernyavskiy Anton, 2020, arXiv
[5] Dimitrov D, 2021, PROCEEDINGS OF THE 15TH INTERNATIONAL WORKSHOP ON SEMANTIC EVALUATION, SEMEVAL-2021, P70
[6] Changing views: Persuasion modeling and argument extraction from online discussions
Dutta, Subhabrata
Das, Dipankar
Chakraborty, Tanmoy
[J]. INFORMATION PROCESSING & MANAGEMENT, 2020, 57 (02)
[7] An evaluation of feature selection methods for environmental data
Effrosynidis, Dimitrios
Arampatzis, Avi
[J]. ECOLOGICAL INFORMATICS, 2021, 61
[8] Hidey C, 2018, AAAI CONF ARTIF INTE, P5173
[9] Iyer Rahul R, 2017, CMNA@ ICAIL, P54
[10] Lundberg SM, 2017, ADV NEUR IN, V30

← 1 2 →