TY - GEN
T1 - Neural network hate deletion
T2 - 5th International Conference on Internet Science, INSCI 2018
AU - Salminen, Joni
AU - Luotolahti, Juhani
AU - Almerekhi, Hind
AU - Jansen, Bernard J.
AU - Jung, Soon Gyo
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2018.
PY - 2018
Y1 - 2018
N2 - We propose a method for modifying hateful online comments to non-hateful comments without losing the understandability and original meaning of the comments. To accomplish this, we retrieve and classify 301,153 hateful and 1,041,490 non-hateful comments from Facebook and YouTube channels of a large international media organization that is a target of considerable online hate. We supplement this dataset by 10,000 Reddit comments manually labeled for hatefulness. Using these two datasets, we train a neural network to distinguish linguistic patterns. The model we develop, Neural Network Hate Deletion (NNHD), computes how hateful the sentences of a social media comment are and if they are above a given threshold, it deletes them using a language dependency tree. We evaluate the results by comparing crowd workers’ perceptions of hatefulness and understandability before and after transformation and find that our method reduces hatefulness without resulting in a significant loss of understandability. In some cases, removing hateful elements improves understandability by reducing the linguistic complexity of the comment. In addition, we find that NNHD can satisfactorily retain the original meaning on average but is not perfect in this regard. In terms of practical implications, NNHD could be used in social media platforms to suggest more neutral use of language to agitated online users.
AB - We propose a method for modifying hateful online comments to non-hateful comments without losing the understandability and original meaning of the comments. To accomplish this, we retrieve and classify 301,153 hateful and 1,041,490 non-hateful comments from Facebook and YouTube channels of a large international media organization that is a target of considerable online hate. We supplement this dataset by 10,000 Reddit comments manually labeled for hatefulness. Using these two datasets, we train a neural network to distinguish linguistic patterns. The model we develop, Neural Network Hate Deletion (NNHD), computes how hateful the sentences of a social media comment are and if they are above a given threshold, it deletes them using a language dependency tree. We evaluate the results by comparing crowd workers’ perceptions of hatefulness and understandability before and after transformation and find that our method reduces hatefulness without resulting in a significant loss of understandability. In some cases, removing hateful elements improves understandability by reducing the linguistic complexity of the comment. In addition, we find that NNHD can satisfactorily retain the original meaning on average but is not perfect in this regard. In terms of practical implications, NNHD could be used in social media platforms to suggest more neutral use of language to agitated online users.
KW - Hate deletion
KW - Neural networks
KW - Online hate
KW - Toxic comments
UR - http://www.scopus.com/inward/record.url?scp=85055785701&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-01437-7_3
DO - 10.1007/978-3-030-01437-7_3
M3 - Conference contribution
AN - SCOPUS:85055785701
SN - 9783030014360
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 25
EP - 39
BT - Internet Science - 5th International Conference, INSCI 2018, Proceedings
A2 - Bodrunova, Svetlana S.
PB - Springer Verlag
Y2 - 24 October 2018 through 26 October 2018
ER -