Nous proposons une étude sur l’analyse automatique d’arguments via des techniques d’apprentissage supervisé exploitant le paradigme de l’apprentissage multi-tâches. Nous définissons pour cela une approche multi-tâches à base d’apprentissage profond que nous évaluons sur un cas d’étude spécifique portant sur l’extraction d’arguments dans un corpus de dissertations. Les résultats obtenus permettent de discuter l’intérêt de définir un modèle multi-tâches unique – optimisé sur différents critères en tirant parti de la diversité des tâches d’apprentissage auxquelles il est confronté – par rapport à un ensemble de classifieurs entraînés de manière indépendante et spécifique. Nous montrons en particulier l’impact de l’ajout de tâches auxiliaires de bas niveau, telles que l’étiquetage morpho-syntaxique et l’analyse de dépendances grammaticales, pour l’obtention de classifieurs multi-tâches performants. Nous observons aussi que l’apprentissage multi-tâches permet l’obtention de modèles efficaces de performances semblables à l’état de l’art pour le cas d’étude traité.
We present a method performing automatic extraction and analysis of arguments from raw texts in a multi-task learning framework. The results obtained show that training a single model on different tasks can lead to good performances. We explore the impact of adding low-level auxiliary tasks, such as Part-Of-Speech tagging and dependency parsing, on a model’s ability to handle more complex tasks. Our experiments show that multi-task learning can lead to competitive results when performing automatic argument mining.
Révisé le :
Accepté le :
Publié le :
Mots clés : Natural Language Processing, Argument mining, Multi-Task Learning
CC-BY 4.0
@article{ROIA_2022__3_3-4_201_0,
author = {Jean-Christophe Mensonides and S\'ebastien Harispe and Jacky Montmain and V\'eronique Thireau},
title = {Analyse automatique d{\textquoteright}arguments et apprentissage multi-t\^aches~~: un cas d{\textquoteright}\'etude},
journal = {Revue Ouverte d'Intelligence Artificielle},
pages = {201--222},
publisher = {Association pour la diffusion de la recherche francophone en intelligence artificielle},
volume = {3},
number = {3-4},
year = {2022},
doi = {10.5802/roia.29},
language = {fr},
url = {https://roia.centre-mersenne.org/articles/10.5802/roia.29/}
}
TY - JOUR AU - Jean-Christophe Mensonides AU - Sébastien Harispe AU - Jacky Montmain AU - Véronique Thireau TI - Analyse automatique d’arguments et apprentissage multi-tâches : un cas d’étude JO - Revue Ouverte d'Intelligence Artificielle PY - 2022 SP - 201 EP - 222 VL - 3 IS - 3-4 PB - Association pour la diffusion de la recherche francophone en intelligence artificielle UR - https://roia.centre-mersenne.org/articles/10.5802/roia.29/ DO - 10.5802/roia.29 LA - fr ID - ROIA_2022__3_3-4_201_0 ER -
%0 Journal Article %A Jean-Christophe Mensonides %A Sébastien Harispe %A Jacky Montmain %A Véronique Thireau %T Analyse automatique d’arguments et apprentissage multi-tâches : un cas d’étude %J Revue Ouverte d'Intelligence Artificielle %D 2022 %P 201-222 %V 3 %N 3-4 %I Association pour la diffusion de la recherche francophone en intelligence artificielle %U https://roia.centre-mersenne.org/articles/10.5802/roia.29/ %R 10.5802/roia.29 %G fr %F ROIA_2022__3_3-4_201_0
Jean-Christophe Mensonides; Sébastien Harispe; Jacky Montmain; Véronique Thireau. Analyse automatique d’arguments et apprentissage multi-tâches : un cas d’étude. Revue Ouverte d'Intelligence Artificielle, Volume 3 (2022) no. 3-4, pp. 201-222. doi : 10.5802/roia.29. https://roia.centre-mersenne.org/articles/10.5802/roia.29/
[1] Unit segmentation of argumentative texts, Proceedings of the 4th Workshop on Argument Mining (2017), pp. 118-128 | DOI
[2] When is multitask learning effective ? Semantic sequence prediction under varying data conditions (2016) (https://arxiv.org/abs/1612.02251)
[3] Using arguments for making and explaining decisions, Artificial Intelligence, Volume 173 (2009) no. 3-4, pp. 413-436 | DOI | MR | Zbl
[4] Can AI models capture natural language argumentation ?, International Journal of Cognitive Informatics and Natural Intelligence (IJCINI), Volume 6 (2012) no. 3, pp. 19-32 | DOI
[5] Identifying beneficial task relations for multi-task learning in deep neural networks (2017) (https://arxiv.org/abs/1702.08303)
[6] Enriching word vectors with subword information, Transactions of the Association for Computational Linguistics, Volume 5 (2017), pp. 135-146 | DOI
[7] A natural language bipolar argumentation approach to support users in online debate interactions, Argument & Computation, Volume 4 (2013) no. 3, pp. 209-230 | DOI
[8] Five Years of Argument Mining : a Data-driven Analysis, IJCAI (2018), pp. 5427-5433 | DOI
[9] Multitask learning, Machine learning, Volume 28 (1997) no. 1, pp. 41-75 | DOI
[10] Learning phrase representations using RNN encoder-decoder for statistical machine translation (2014) (https://arxiv.org/abs/1406.1078)
[11] Bert : Pre-training of deep bidirectional transformers for language understanding (2018) (https://arxiv.org/abs/1810.04805)
[12] On the acceptability of arguments and its fundamental role in nonmonotonic reasoning, logic programming and n-person games, Artificial intelligence, Volume 77 (1995) no. 2, pp. 321-357 | DOI | MR | Zbl
[13] On the role of discourse markers for discriminating claims and premises in argumentative discourse, Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (2015), pp. 2236-2242 | DOI
[14] Neural end-to-end learning for computational argumentation mining (2017) (https://arxiv.org/abs/1704.06104)
[15] Measuring nominal scale agreement among many raters, Psychological bulletin, Volume 76 (1971) no. 5, pp. 378-382 | DOI
[16] Argumentative link prediction using residual networks and multi-objective learning, Proceedings of the 5th Workshop on Argument Mining (2018), pp. 1-10 | DOI
[17] Argument extraction from news, blogs, and social media, Hellenic Conference on Artificial Intelligence (Lecture Notes in Computer Science), Volume 8445 (2014), pp. 287-299 | DOI
[18] A joint many-task model : Growing a neural network for multiple nlp tasks (2016) (https://arxiv.org/abs/1611.01587)
[19] Delving deep into rectifiers : Surpassing human-level performance on imagenet classification, Proceedings of the IEEE international conference on computer vision (2015), pp. 1026-1034 | DOI
[20] Long short-term memory, Neural computation, Volume 9 (1997) no. 8, pp. 1735-1780 | DOI
[21] Encoding Position Improves Recurrent Neural Text Summarizers, Proceedings of the 3rd International Conference on Natural Language and Speech Processing (2019), pp. 142-150
[22] Adam : A method for stochastic optimization (2014) (https://arxiv.org/abs/1412.6980)
[23] The measurement of observer agreement for categorical data, biometrics (1977), pp. 159-174 | DOI | Zbl
[24] Argument Mining : A Survey, Computational Linguistics, Volume 45 (2020) no. 4, pp. 765-818 | DOI
[25] Context dependent claim detection, Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics : Technical Papers (2014), pp. 1489-1500
[26] Argumentation mining : State of the art and emerging trends, ACM Transactions on Internet Technology (TOIT), Volume 16 (2016) no. 2, p. 10 | DOI
[27] Identifying high-level organizational elements in argumentative discourse, Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics : Human Language Technologies (2012), pp. 20-28
[28] Distributed representations of words and phrases and their compositionality, Advances in neural information processing systems (2013), pp. 3111-3119
[29] How transferable are neural networks in nlp applications ? (2016) (https://arxiv.org/abs/1603.06111)
[30] Rectified linear units improve restricted boltzmann machines, Proceedings of the 27th international conference on machine learning (ICML-10) (2010), pp. 807-814
[31] Argument mining with structured SVMs and RNNs (2017) (https://arxiv.org/abs/1704.06869)
[32] Identifying appropriate support for propositions in online user comments, Proceedings of the first workshop on argumentation mining (2014), pp. 29-38 | DOI
[33] On the difficulty of training recurrent neural networks, International conference on machine learning (2013), pp. 1310-1318
[34] Joint prediction in MST-style discourse parsing for argumentation mining, Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (2015), pp. 938-948 | DOI
[35] Glove : Global vectors for word representation, Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP) (2014), pp. 1532-1543 | DOI
[36] End-to-end argumentation mining in student essays, Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics : Human Language Technologies (2016), pp. 1384-1394 | DOI
[37] Deep contextualized word representations (2018) (https://arxiv.org/abs/1802.05365)
[38] Here’s my point : Joint pointer architecture for argument mining (2016) (https://arxiv.org/abs/1612.08994)
[39] Language models are unsupervised multitask learners, OpenAI Blog, Volume 1 (2019) no. 8
[40] Text chunking using transformation-based learning, Natural language processing using very large corpora, Springer, 1999, pp. 157-176 | DOI
[41] An overview of gradient descent optimization algorithms (2016) (https://arxiv.org/abs/1609.04747)
[42] An overview of multi-task learning in deep neural networks (2017) (https://arxiv.org/abs/1706.05098)
[43] Introduction to the CoNLL-2000 shared task : Chunking (2000) (https://arxiv.org/abs/cs/0009008)
[44] Argument extraction from news, Proceedings of the 2nd Workshop on Argumentation Mining (2015), pp. 56-66 | DOI
[45] Exact solutions to the nonlinear dynamics of learning in deep linear neural networks (2013) (https://arxiv.org/abs/1312.6120)
[46] Multi-task learning for argumentation mining in low-resource settings (2018) (https://arxiv.org/abs/1804.04083)
[47] A Gold Standard Dependency Corpus for English., LREC (2014), pp. 2897-2904
[48] Deep multi-task learning with low level tasks supervised at lower layers, Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2 : Short Papers) (2016), pp. 231-235 | DOI
[49] Hierarchical multi-task learning for organization evaluation of argumentative student essays, Proceedings of the 29th International Joint Conference on Artificial Intelligence (2020), pp. 3875-3881 | DOI
[50] Dropout : a simple way to prevent neural networks from overfitting, The journal of machine learning research, Volume 15 (2014) no. 1, pp. 1929-1958 | MR | Zbl
[51] Parsing argumentation structures in persuasive essays, Computational Linguistics, Volume 43 (2017) no. 3, pp. 619-659 | DOI | MR
[52] Multi-task cross-lingual sequence tagging from scratch (2016) (https://arxiv.org/abs/1603.06270)
Cité par Sources :