-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathindex.html
More file actions
228 lines (178 loc) · 10.6 KB
/
Copy pathindex.html
File metadata and controls
228 lines (178 loc) · 10.6 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
<!-- HEADER-->
<!DOCTYPE html>
<html lang="pt-br">
<head>
<!-- Meta, title, CSS, favicons, etc. -->
<title>Violentometer</title>
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<meta name="keywords" content="violence, hate speech, agression, social media, artificial intelligence" />
<meta name="title" content="Violentometer - Monitoring violence levels on social media">
<meta name="author" content="Henrique Xavier">
<meta name="description" content="Este projeto monitora postagens na plataforma Twitter direcionadas a mulheres candidatas ao Legislativo para identificar discursos violentos.">
<link rel="shortcut icon" href="/favicon.ico">
<!-- FACEBOOK -->
<meta property="og:locale" content="pt_BR" />
<meta property="og:url" content="https://github.com/cewebbr/violentometro">
<meta property="og:site_name" content="Violentometer" />
<meta property="og:type" content="website" />
<meta property="og:title" content="Violentometer - Monitoring violence levels on social media" />
<meta property="og:description" content="This project is an example of how one can use AI to measure the evolution of violence levels on social media." />
<meta property="og:image" content="#" />
<!-- / FACEBOOK -->
<!-- TWITTER -->
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:site" content="https://github.com/cewebbr/violentometro">
<meta name="twitter:creator" content="Henrique Xavier">
<meta name="twitter:description" content="This project is an example of how one can use AI to measure the evolution of violence levels on social media.">
<meta name="twitter:image:src" content="#">
<meta property="twitter:title" content="Violentometer - Monitoring violence levels on social media" />
<!-- / TWITTER -->
<!-- STYLE -->
<link rel="stylesheet" href="assets/bootstrap/css/bootstrap.min.css">
<link rel="stylesheet" href="assets/fonts/fonts.css">
<link rel="stylesheet" href="assets/fontawesome/css/all.css">
<link rel="stylesheet" href="assets/css/theme.css?v=1.0">
<link href="assets/fonts/Barrio/css2.css?family=Barrio&display=swap" rel="stylesheet">
<link href="assets/fonts/Barrio/cssMontserrat.css?family=Montserrat" rel="stylesheet">
<!-- /STYLE -->
<!-- JS -->
<script src="assets/js/jquery-3.4.1.min.js"></script>
<script src="assets/bootstrap/js/bootstrap.min.js"></script>
<script src="assets/js/data.js"></script>
<script src="assets/js/main.js"></script>
<!-- JS -->
</head>
<body>
<!-- HEADER -->
<header>
<!-- acessibilidade - skip link -->
<a href="#conteudo" class="sr-only" title="Ir direto para o conteúdo">Ir para o conteúdo</a>
<!-- / acessibilidade - skip link -->
<!-- navbar -->
<nav id="navbar_main" class="mobile-offcanvas navbar navbar-expand-lg navbar-light sticky">
<!-- fechar menu mobile -->
<div class="offcanvas-header">
<button class="btn btn-close float-right">
<span class="fas fa-times fa-lg" aria-hidden="true">
<span class="sr-only">Fechar</span>
</span>
</button>
</div>
<!-- fechar menu mobile -->
<div class="container-fluid">
<!-- itens menu -->
<ul class="navbar-nav ml-auto align-items-center">
<li class="nav-item"><a href="#sobre" class="page-scroll nav-link">About</a></li>
<li class="nav-item"><a href="#graficos" class="page-scroll nav-link">Violence levels</a></li>
<li class="nav-item"><a href="#metodologia" class="page-scroll nav-link">Methodology</a></li>
</ul>
<!-- / itens menu -->
</div>
</nav>
<!-- / navbar -->
</header>
<!-- / HEADER -->
<!-- HEADER -->
<!-- CONTEÚDO -->
<div class="col-12 p-0">
<!-- BANNER-->
<div class="content_about" id="conteudo">
<div class="container" id="sobre">
<h1 class="title_site">VIOLENTOMETER</h1>
<p class="content_home">
The Violentometer is an experimental project for investigating the application of Artificial Intelligence (AI) to the measurement of violence levels on social media
(i.e. frequency of offensive comments and hate speech). Our research suggests that the levels measured by the AI follow the actual frequency of violent texts in a linear way,
in the same way a temperature scale is related to the average kinetic energy of the particles in a gas. Thus,
although an absolute measurement of violence is not attainable, it is possible to measure variations on violence levels through time for a fixed social group.
</p>
<p class="content_home">
As an application of this idea, this webpage presents daily measurements of the violence directed, on Twitter, to former female candidates (and current politicians) in the Brazilian 2022 elections.
</p>
</div>
</div> <!-- BANNER -->
<!-- GRAFICOS PILHA PIZZA -->
<section id="graficos-pilhas">
<div class="graficos mt-5">
<img src="assets/img/bg/bg-graficos-pilhas.png" class="bg_container_graficos" alt="Gráfico - lorem" aria-hidden="true" role="presentation">
<div class="container">
<h2 class="sub_title_site">Daily interactions and attacks</h2>
<!-- DATA DOS DADOS -->
<div class="col-12 p-0">
<h3 class="desc_box_number font-300">Last update on <span id='datadate'>19/09/2022</span> at <span id='datatime'>04:12:44</span>.</h3>
</div>
<!-- DATA DOS DADOS -->
<div class="graficos_pilhas">
<div class="row">
<div class="col">
<img src="assets/img/graficos/tweets_agressivos_por_dia_feminino.png" class="img-fluid" alt="Série temporal do número de ataques por dia, e a fração do total de tweets que eles representam.">
<div class="col-12 col-lg-6 col-md-6 col-sm-12">
<p class="desc_box_number">
The plot above shows, for each day: the total number of tweets directed to the candidates (top panel);
the estimated number of violent tweets directed to the candidates (middle panel);
and the fraction of tweets that were considered violent (bottom panel).
A table (CSV file) containing the plot's data can be found
<a href="assets/data/tweets_agressivos_por_dia_feminino.csv" title="Arquivo CSV com os dados do gráfico">here</a>.
</p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- METODOLOGIA -->
<section id="metodologia" class="mt-5">
<div class="container">
<div class="container_parceria">
<h3 class="title_parceria">Methodology</h3>
<p class="desc_box_number">Every three hours, an automated system selects at random a third of the 975 candidates' profiles provided as input and
captures the mentions and replies made to them within the last three hours, so that the mentions to the set of input profiles are monitored continuously,
in a statistical sense. Most of the mentions collected are responses to the candidates' tweets.
</p>
<p class="desc_box_number">To make sure that the mentions are directed to the candidates in question, we only select direct
responses to the candidates' tweets or tweets that do not mention any profile other than the candidate's. We also ignore
tweets with no text. The selected tweets are classified by the AI that calculates, from the text of the tweet, the probability
that each tweet contains hate speech or offensive comments.</p>
<p class="desc_box_number">The AI used in this project is a neural network of the “transformer” type, with BERT architecture,
pre-trained by
<a href="https://github.com/neuralmind-ai/portuguese-bert" title="Ir para página - BERTimbau" target="_blank">Souza et al. (2020)</a>
to guess following sentences and hidden words among 3.53 million documents retrieved from the Brazilian Web by
<a href="https://www.researchgate.net/publication/326303825_The_brWaC_Corpus_A_New_Open_Resource_for_Brazilian_Portuguese" title="Ir para página - Corpus brWaC" target="_blank">
Wagner et al. (2018)</a>.
This model was fine-tuned for the specific task of identifying violent texts from 926 examples and 4,713 counterexamples provided by the works of
<a href="https://github.com/rogersdepelle/OffComBR" title="Ir para página - Github de Pelle" target="_blank">Pelle & Moreira (2017)</a>
and <a href="https://github.com/paulafortuna/Portuguese-Hate-Speech-Dataset" title="Ir para página - Github de Fortuna" target="_blank">Fortuna et al. (2019)</a>.
<p class="desc_box_number">The number of daily attacks directed to the candidates is estimated by adding the
probabilities calculated by the AI and rescaling this value, obtained from the statistical samples, for all
975 candidates' profiles provided as input.</p>
</div>
</div>
</section>
<!-- METODOLOGIA -->
</div>
<!-- CONTEÚDO -->
<!-- FOTTER-->
<!-- FOOTER -->
<div class="logos-footer">
<div class="row">
<div class="col-lg-1 col-md-1 col-12">
<span class="title-footer">Credits</span>
</div>
<div class="col-lg-7 col-md-6 col-12">
<p class="content-iniciativa">This webpage's design was made by
<a href="https://institutovamosjuntas.org/" title="Ir para página - Instituto Vamos Juntas" target="_blank">Vamos Juntas</a>,
and the development was made by <a href="https://nic.br" title="Ir para página - NIC.br" target="_blank">NIC.br</a>. The AI model,
training data, code, analysis and results associated to this project are available on
<a href="https://github.com/cewebbr/violentometro" title="Ir para página - Violentometer's repository on Github" target="_blank">https://github.com/cewebbr/violentometro</a>.
</p>
<p class="content-iniciativa">The images in this webpage are free to use and were obtained from
<a href="https://nappy.co/" title="Ir para página - Nappy" target="_blank">Nappy</a>
and <a href="https://www.canva.com" title="Ir para página - Canva" target="_blank">Canva</a>.
</div>
</div>
</div>
<!-- FOOTER -->
</body>
</html><!-- FOTTER -->