=Paper=
{{Paper
|id=Vol-2517/T3-21
|storemode=property
|title=DEEP at HASOC2019: A Machine Learning Framework for Hate Speech and Offensive Language Detection
|pdfUrl=https://ceur-ws.org/Vol-2517/T3-21.pdf
|volume=Vol-2517
|authors=Hamada A Nayel,Shashirekha H L
|dblpUrl=https://dblp.org/rec/conf/fire/NayelL19
}}
==DEEP at HASOC2019: A Machine Learning Framework for Hate Speech and Offensive Language Detection==
https://ceur-ws.org/Vol-2517/T3-21.pdf
DEEP at HASOC2019 : A Machine Learning
Framework for Hate Speech and Offensive
Language Detection
Hamada A. Nayel1[0000−0002−2768−4639] and Shashirekha H. L.2
1
Department of Computer Science
Faculty of Computers and Artificial Intelligence
Benha University, Egypt
hamada.ali@fci.bu.edu.eg
2
Department of Computer Science
Mangalore University, India
hlsrekha@gmail.com
Abstract. In this paper, we describe the system submitted by our team
for Hate Speech and Offensive Content Identification in Indo-European
Languages (HASOC) shared task held at FIRE 2019. Hate speech and
offensive language detection have become an important task due to the
overwhelming usage of social media platforms in our daily life. This
task has been applied for three languages namely, English, Germany and
Hindi. The proposed model uses classical machine learning approaches
to create classifiers that are used to classify the given post according to
different subtasks.
Keywords: Multi-task Classification · Multi-lingual Text Analysis · Hate
Speech and Offensive Detection 3
1 Introduction
Wide spread use of internet is giving more freedom to people to express their
thoughts freely and anonymously in different forms such as blogs, business net-
works, forums and social media such as Twitter and Facebook. Internet has
enabled the interaction among people from different culture, race, religion, ori-
gin, gender and nationality. It has also paved the way for the online information
to reach the mass within a matter of seconds. Social media is generating lot of
content as it has opened up a new galaxy of opportunities for people to express
their opinions online resulting in the exchange of ideas in a positive way as well as
contributing to the propagation of hate speech in a negative way. The anonymity
of users provided by the internet is also impacting the society in generating the
web contents in a negative way by the usage of profane words, hate speech,
3
Copyright c 2019 for this paper by its authors. Use permitted under Creative Com-
mons License Attribution 4.0 International (CC BY 4.0). FIRE 2019, 12-15 December
2019, Kolkata, India.
Nayel and Shashirekha
derogatory terms, racists slurs, toxic, abusive and offensive language. This nega-
tive content may be aggressive and potentially harmful lowering the self-esteem
of people leading to mental illness and suicidal attempts and has also forced
people to deactivate their social media accounts. The increase of cyberbullying
and cyberterrorism, and the use of hate speech content on the Internet, make
the identification of hate speech an essential ingredient for anti-bullying policies
of social media [5]. Many terrorist activities, which are related to hate speech,
are gaining huge attention in social media in terms of posts and suggestions [1].
Hate speech and other offensive and objectionable content online are posing
threats and challenges to the society. A lot of countries prohibit hate speech
in social media subject to the condition that it should not target any group or
trigger any crime. As hate speech acts as a opinion builder, many online forums
such as YouTube, Face book and Twitter, have their own policies to remove hate
speech content or anything which is impacting the society in a negative way.
It is therefore important to take preventive measures to cope up with hate
speech and objectionable content or remove such content. Detecting hate speech
is a difficult task as there is no clear definition for hate speech and can vary
according to the context. For example, speech which contain subtle and nuance
sentences are difficult to detect as hate speech. Further, the usage of slang terms
may be found neutral, but may trigger hate crime or offense in some context.
Although there is no universal definition for hate speech, the most accepted
definition is provided by Nockleby [12]: “any communication that disparages a
target group of people based on some characteristic such as race, colour, ethnic-
ity, gender, sexual orientation, nationality, religion, or other characteristic” [5].
Detection and removal of hate speech and objectionable content manually is
a tedious task due to the massiveness of the web and the increasing number of
online users. Further, the task becomes difficult due to the anonymity of users
on the internet. Hence, there is a great demand for tools and techniques that
automatically detect the hate speech and objectionable content quickly on the
web to reduce the spread of hate speech.
In this paper, we describe the system submitted by our team for the shared
task of Hate Speech and Offensive Content Identification in Indo-European Lan-
guages (HASOC) track at FIRE 2019 [7]. Rest of the paper is organized as
follows: In Section 2 we review the related work on hate speech and offensive
language detection. Description of the task and a summary of dataset is given
in Section 3. Our proposed model is discussed in Section 4 followed by the ex-
periments and results in Section 5. Finally, we conclude the paper in Section
6.
2 Related work
Though, defining and understanding hate speech is difficult, several online fo-
rums, IT industries and researchers have explored many algorithms to detect
hate speech, offensive and abusive content on the web [5, 2, 6, 3, 15, 4]. Thomas
et.al. [2] trained a multi-class classifier to classify tweets into one of three cate-
A ML approach for HASOC
gories, namely, hate speech, offensive but not hate speech, neither offensive and
nor hate speech. They performed experiments using logistic regression, Naı̈ve
Bayes, decision trees, random forests, and linear SVM classifiers with 5-fold
cross validation and obtained an overall precision 0.91, recall of 0.90, and F1
score of 0.90 for the best performing model. Gibert et.al. [5], constructed a man-
ually labeled hate speech dataset composed of thousands of sentences obtained
from Stormfront, a white supremacist online forum which contains hateful and no
hateful sentences. Their model was built using Support Vector Machines (SVM),
Convolutional Neural Networks (CNN), Recurrent Neural Networks with Long
Short term Memories to annotate the test data.
Sean MacAvaney et.al. [6], examines the challenges faced by online auto-
matic approaches for hate speech detection in text in addition to the existing
approaches and various datasets for hate speech detection. They also have pro-
posed a multi-view SVM approach that achieves near state-of-the-art perfor-
mance, while being simpler and producing more easily interpretable decisions
than neural methods. The experiments were conducted on HatebaseTwitter,
Stanfront and TRAC datasets. Avishek Garain and Arpan Basu [3], presents a
description of their system to detect offensive language in Twitter submitted to
“SemEval-2019 Task 6 [14]” which uses bidirectional LSTM, a neural network
based model to capture information from both the past and future context. Ziqi
and Lei [15], discusses about the typical features responsible for classification
between hate speech and no hate content. In their work they proposed Deep
Neural Network structures serving as feature extractors that are particularly
effective for capturing the semantics of hate speech. Their methods were evalu-
ated on the largest collection of hate speech datasets based on Twitter, and are
shown to be able to outperform the best performing method by up to 5 percent-
age points in macro-average F1, or 8 percentage points in the more challenging
case of identifying hateful content. Aditya et.al., [4] proposes an approach to
automatically classify tweets on Twitter into three classes: hateful, offensive and
clean by considering features like n Grams and TF/IDF values. They performed
a comparative analysis of the models considering several values of n in n-grams
and TF/IDF normalization methods on Twitter dataset and achieved 95.6% ac-
curacy. Areej Al-Hassan and Hmood Al-Dossari [1] in their survey paper, gives
detailed information about hate speech in terms of insufficient dataset, complex-
ities and challenges and the complete knowledge about text mining and Natural
Language Processing techniques to find and classify hateful and no-hateful con-
tents.
3 Task Description and Corpus
Hate Speech and Offensive Content Identification in Indo-European Languages
(HASOC) shared task aims at identifying the hate speech and offensive content in
three Indo-European languages namely English, Germany and code-mixed Hindi.
The shared task is divided into three sub-tasks to identify the hate speech and
offensive content posts and classify them into various predefined categories. The
Nayel and Shashirekha
first sub-task, Sub-task A is a coarse-grained binary classification that focuses
on the identification of hate speech and offensive language in the given posts
and is offered for English, German and Hindi language posts. Sub-task B is a
fine-grained multi-class classification that classifies the hate speech and offensive
language posts into one of the three predefined categories, namely hate speech,
offensive and profane. This subtask is offered for English, German and Hindi
language posts. The third sub-task, Sub-task C aims at checking the type of
offense as targeted insult or un-targeted insult and is offered for English and
Hindi. The corpus consists of posts collected from Twitter tweets and Facebook
comments in English (EN), German (GR) and Hindi (HI) languages. Each post
contains three labels for Sub-task A, Sub-task B and Sub-task C respectively.
Table 1 gives a glimpse of the shared task and the associated categories in each
sub-task.
Table 1. A glimpse of shared task and the associated categories in each sub-task
Sub-task Class Labels Description
NOT Contains neither hate speech nor offensive content
A
HOF Contains hate, offensive and profane content
HATE Contains hate speech content
B OFFN Contains offensive language content
PRFN Contains profane words
TIN Contains an insult to an individual, group, or others
C
UNT Contains non targeted profanity
4 Methodology
A detailed description of our model and the classifiers used are given in this
section.
4.1 Problem Formulation
Given a set of posts P = {p1 , p2 , ..., pn }, where each post is composed of a set of
words pi = {w1 , w2 , ..., wk }, in each language L = {EN, GR, HI} and the classes
A = {N OT, HOF }, B = {HAT E, OF F N, P RF N } and C = {T IN, U N T } for
the sub-tasks A, B and C respectively, the shared task is to formulate a multi-
label classification problem, where an unlabeled post/instance is assigned with
multiple class labels one from each class A, B and C. ie., given a unlabeled post
pk , multi-label classification problem will assign the triple < a, b, c > such that,
a ∈ A, b ∈ B and c ∈ C.
A ML approach for HASOC
4.2 Model
The general structure of the proposed model is given in Fig. 1. This model creates
sub-models for each subtask and then assigns multiple class labels, one for each
subtask, by combining the output of each sub-model. The model consists of the
following phases:
Fig. 1. The general structure of our model.
I. Preprocessing Preprocessing also known as corpus cleaning is a crucial and
first step in building any classifiers. Each post pk has been tokenized into a set
of words to get n-gram (n = 2) bag of words. In this step, all un-informative
tokens such as urls, digits and special characters have been removed from all the
posts.
II. Feature Extraction In this phase, a TF/IDF vector has been computed
for all the posts in the training set. This vector will be used as an input for
training the classifier. TF/IDF has been calculated as described in [10].
III. Training the Classifier In this phase, the features that have been ex-
tracted in previous phase are used as input for training the classifiers. Three
Nayel and Shashirekha
classification algorithms namely, Linear classifier, SVM and Multilayer Percep-
tron (MLP) have been used separately to train the proposed model. Linear
classifier is a simple classifier that uses a set of linear discriminant functions to
distinguish between different classes [13]. SVM is a kind of linear classifier which
uses the samples close to the classes’s boundaries for training and these samples
are known as support vectors [11]. SVM has been used effectively in different
NLP tasks [9, 8]. MLP is a deep learning approach that uses back propagation
for training the neural network [10]. It is characterized by several layers of input
nodes connected as a directed graph between the input and output layers.
Three different classifiers namely, Linear classifier, SVM and MLP are created
for each subtask. Classifiers for Sub-task A will classify the unlabeled instance
into one of the two predefined categories as mentioned in Table-1. While the en-
tire data is used to construct classifiers for Sub-task A, a sub-corpus (only posts
labelled as HOF in Sub-task A) that contain only the hate speech and offensive
language posts is used as training corpus to create classifiers for Sub-tasks B and
C. Classifiers for Sub-task B will classify the unlabeled instance into one of the
three predefined categories and classifier for Sub-task C will classify the same
instance into one of the two predefined categories as mentioned in Table 1.
5 Experiments and Results
In our proposed model, the Stochastic Gradient Descent (SGD) optimization
algorithm has been used for optimizing the parameters of linear classifier, while
SVM uses Lagrange multipliers to solve the optimization problem [13]. The loss
function used in linear classifier was ”Hinge” loss function. Linear kernel has
been used for SVM classifier. In MLP classifier the logistic function has been
used as activation function using 20 neurons in the hidden layer.
We have used cross-validation approach with five folds to train different sub-
models for each subtask and the outputs of each sub-task have been combined to
generate the final output. The organizers of the shared task used Macro-averaged
F1-score (M-f1) and weighted F1-score (W-f1) [10] as performance evaluation
metrics for all sub-tasks. The dataset comprises of a set of tweets and Facebook
comments in the three languages labelled with different categories. Statistics of
the dataset is shown in Table 2.
Table 3 shows the M-f1 and W-f1 of our submissions for all sub-tasks over
English, Germany and Hindi language posts. From the table, it is clear that SVM
outperforms other classifiers for all subtasks for English. For German language
posts, MLP outperforms other classifiers for Sub-tasks A and B. It is worth to
note that, Sub-task C is not applied for Germany. For Hindi language posts,
MLP outperforms other classifiers for Subtask A, while SVM reported better
results than MLP and linear classifier for subtask B.
There is a gab between M-f1 and W-f1 for all subtasks except for the subtask
A for Hindi. This is due to the fact that the performance of the subtasks B and
C depends on the performance of the subtask A. If the prediction of subtask A
is wrong then by default subtasks B and C will also have wrong prediction.
A ML approach for HASOC
Table 2. Statistics of the Training Corpus
Sub-task Class Labels English German Hindi
NOT 3591 3412 2196
A
HOF 2261 407 2469
HATE 1143 111 556
B OFFN 451 210 676
PRFN 667 86 1237
TIN 2041 - 1545
C
UNT 220 - 924
Total Samples 5852 3819 4665
Table 3. M-f1 and W-f1 for all classifiers and subtasks for English, Germany and Hindi
English Germany Hindi
Subtask Classifier
M-f1 W-f1 M-f1 W-f1 M-f1 W-f1
SVM 66.12% 73.02% 45.62% 76.64% 75.49% 75.52%
A Linear Classifier 44.32% 65.13% 45.65% 76.70% 74.13% 74.11%
MLP 62.61% 69.05% 46.37% 76.92% 75.94% 75.98%
SVM 42.36% 67.69% 22.81% 76.64% 47.20% 59.00%
B Linear Classifier 23.02% 64.94% 22.83% 76.70% 44.16% 56.05%
MLP 34.50% 62.12% 23.46% 76.92% 46.63% 58.37%
SVM 42.23% 69.89% - - 51.72% 69.67%
C Linear Classifier 29.44% 64.95% - - 47.86% 68.63%
MLP 39.75% 65.93% - - 52.38% 68.03%
6 Conclusion
In this paper, a machine learning approaches have been used for creating a
model for detecting hate speech and offensive language content. Proposed model
achieved good results compared to its simplicity. Extension of our work includes
using deep learning approach to build the classifier and test it on much bigger
dataset.
References
1. Al-Hassan, A., Al-Dossari, H.: Detection of Hate Speech in Social Networks: A
Survey on Multilingual Corpus. In: Proceedings of 6th International Conference on
Computer Science and Information Technology (CoSIT 2019). pp. 83–100. Dubai,
UAE (February 23-24 2019)
2. Davidson, T., Warmsley, D., Macy, M.W., Weber, I.: Automated Hate Speech De-
tection and the Problem of Offensive Language. In: Proceedings of the Eleventh In-
ternational Conference on Web and Social Media, ICWSM 2017, Montréal, Québec,
Canada, May 15-18, 2017. pp. 512–515. AAAI Press (2017)
Nayel and Shashirekha
3. Garain, A., Basu, A.: The Titans at SemEval-2019 Task 6: Offensive Language
Identification, Categorization and Target Identification. In: May, J., Shutova, E.,
Herbelot, A., Zhu, X., Apidianaki, M., Mohammad, S.M. (eds.) Proceedings of
the 13th International Workshop on Semantic Evaluation, SemEval@NAACL-HLT
2019, Minneapolis, MN, USA, June 6-7, 2019. pp. 759–762. Association for Com-
putational Linguistics (2019), https://www.aclweb.org/anthology/S19-2133/
4. Gaydhani, A., Doma, V., Kendre, S., Bhagwat, L.: Detecting Hate Speech and
Offensive Language on Twitter using Machine Learning: An N-gram and TFIDF
based Approach. CoRR (2018), http://arxiv.org/abs/1809.08651
5. de Gibert, O., Perez, N., Garcı́a-Pablos, A., Cuadros, M.: Hate Speech Dataset
from a White Supremacy Forum. In: Proceedings of the 2nd Workshop on Abusive
Language Online (ALW2). pp. 11–20. Association for Computational Linguistics,
Brussels, Belgium (Oct 2018). https://doi.org/10.18653/v1/W18-5102
6. MacAvaney, S., Yao, H.R., Yang, E., Russell, K., Goharian, N., Frieder, O.: Hate
speech detection: Challenges and solutions. PLoS ONE (2019)
7. Modha, S., Mandl, T., Majumder, P., Patel, D.: Overview of the HASOC track at
FIRE 2019: Hate Speech and Offensive Content Identification in Indo-European
Languages. In: Proceedings of the 11th annual meeting of the Forum for Informa-
tion Retrieval Evaluation (December 2019)
8. Nayel, H., Shashirekha, H.L.: Improving NER for Clinical Texts by En-
semble Approach using Segment Representations. In: Proceedings of the
14th International Conference on Natural Language Processing (ICON-2017).
pp. 197–204. NLP Association of India, Kolkata, India (December 2017),
http://www.aclweb.org/anthology/W/W17/W17-7525
9. Nayel, H.A.: NAYEL@APDA: Machine Learning Approach for Author Profiling
and Deception Detection in Arabic Texts. In: Proceedings of the 11th annual meet-
ing of the Forum for Information Retrieval Evaluation (December 2019)
10. Nayel, H.A., Shashirekha, H.L.: Mangalore University INLI@FIRE2018: Artificial
Neural Network and Ensemble based Models for INLI. In: Mehta, P., Rosso, P.,
Majumder, P., Mitra, M. (eds.) Working Notes of FIRE 2018 - Forum for In-
formation Retrieval Evaluation, Gandhinagar, India, December 6-9, 2018. CEUR
Workshop Proceedings, vol. 2266, pp. 110–118. CEUR-WS.org (2018), http://ceur-
ws.org/Vol-2266/T2-10.pdf
11. Nayel, H.A., Shashirekha, H.L., Shindo, H., Matsumoto, Y.: Improving Multi-
Word Entity Recognition for Biomedical Texts. CoRR abs/1908.05691 (2019),
http://arxiv.org/abs/1908.05691
12. Nockleby, J.T.: Hate Speech. In: Encyclopedia of the American Constitution. pp.
1277–1279 (2000)
13. Theodoridis, S., Koutroumbas, K.: Chapter 3 - Linear Classifiers. In: Pattern
Recognition (Fourth Edition), pp. 91 – 150. Academic Press, Boston, fourth edition
edn. (2009)
14. Zampieri, M., Malmasi, S., Nakov, P., Rosenthal, S., Farra, N., Kumar, R.:
SemEval-2019 Task 6: Identifying and Categorizing Offensive Language in Social
Media (OffensEval). In: Proceedings of the 13th International Workshop on Se-
mantic Evaluation. pp. 75–86 (2019)
15. Zhang, Z., Luo, L.: Hate Speech Detection: A Solved Problem? The Chal-
lenging Case of Long Tail on Twitter. CoRR abs/1803.03662 (2018),
http://arxiv.org/abs/1803.03662