Huertas97 commited on
Commit
81ab8b5
1 Parent(s): a262acc

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +32 -9
app.py CHANGED
@@ -132,11 +132,14 @@ st.markdown('<p class="big-font">Welcome to <font color="#4B8BBE">Leet</font><fo
132
 
133
  with st.expander("Project Description", expanded=False):
134
  st.write("""
135
- Developed in Applied Intelligence and Data Analysis ([AI+DA](http://aida.etsisi.upm.es/)) group at Polytech University of Madrid (UPM).
136
-
137
- This tool uses a Spacy-Transformer Name Entity Recognition model to detect the presence of words camouflaged. Word camouflage is currently used to evade content moderation in Social Media. Therefore, the aim of this tool is to counter new ways of misinformation that emerge in social media platforms. This version is a initial step focused on detecting word camouflage in news, reports or talks since the models have been fine-tuned over TED talks, OPUS News Commentaries and Wikipedia.
138
-
139
- Currently, two languages are supported: English and Spanish. Additionally, you can select whether the detected entities are broken down into the three types of camouflaged words: Canonical Leetspeak, Punctuation Camouflaged, Inversion Camouflaged. Finally, you can select between "accuracy" or "efficiency" performance.
 
 
 
140
  """)
141
 
142
  with st.expander("Try any of these examples", expanded=False):
@@ -145,17 +148,37 @@ with st.expander("Try any of these examples", expanded=False):
145
  - Desperately dominated by fam1ly sitüatløns, he leaves her.
146
  - You might as well come out to investigate a strang3 n'o?i+se or something.
147
  - But one other thing that we have to re;think is the way that we dy£ our #c!l.o|th?£+s.
148
- - And he wanted Baltimore to get that same kind of att£ntløn from the outside, but )i)n)t)r)o)s)p)e)c)t)i)o)n from the inside about what was going on with us.
149
-
150
 
151
  SPANISH
152
  - _d+i%o"s mío!
153
  - La C0v!d es un 3ng@ño de los G0b!3rno$
154
- - @#plan#demia, pl@πd€m1∆ instead of “pandemia” (pandemic)
155
  - se asocian con el m13;d0 y el d'o'lor. g£rønlmo solía decir
156
  - Con las nuevas tecnologías digitales, los agrlcultør£s pueden manejar mejor el uso de sus tierras, su energía y su agua, y prepararse para el mal clima.
157
  - En el tiempo transcurrido entre mi período de escuela %s%3%c%_%n%d%a%r%1%a y el mo'm3n'to de empezar a enseñar vimos surgir el fenómeno de in't£r'net
158
- - Las pre0c_pac1on3s van desde inquietudes por las ramificaciones desestabilizadoras de una estrategia de salida de la FC, hasta aprehensión por pérdidas de capital en la rápidamente creciente cartera de valores de la Fed (actualmente de $3 billones y en camino a los $4 billones para finales de este año).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
159
  """)
160
 
161
  # - Why do all these _r_e_p_o_r_t_e_r_s, who get praise and money for doing what Assange has done, maintain a cow;ardly silence (at best) while a fellow publisher faces threats of extradition, banning, and espionage charges (which can incur the death penalty), not to mention calls for his as'sa'ss1nat'i'on?
 
132
 
133
  with st.expander("Project Description", expanded=False):
134
  st.write("""
135
+ Developed by the Applied Intelligence and Data Analysis (AI+DA) group at the Polytechnic University of Madrid (UPM), this tool employs a Spacy-Transformer Named Entity Recognition (NER) model to detect camouflaged words.
136
+ Word camouflage is a technique used to evade content moderation on social media platforms, contributing to the spread of misinformation. This tool aims to counter such emerging threats by focusing on detecting camouflaged words in news articles, reports, and talks, as the models have been fine-tuned using TED talks, OPUS News Commentaries, and Wikipedia data.
137
+
138
+ The multilingual model currently supports over 20 languages, including Arabic, Azerbaijani, Danish, German, Greek, English, Spanish, Finnish, French, Hungarian, Indonesian, Italian, Kazakh, Norwegian Bokmål, Nepali, Dutch, Portuguese, Romanian, Russian, Slovenian, Swedish, Tajik, and Turkish.
139
+ The NER model has been tested in English, Spanish, French, Italian, and
140
+ German.
141
+ Additionally, users can choose to have detected entities broken down into three types of camouflaged words: Canonical Leetspeak, Punctuation Camouflaged, and Inversion Camouflaged.
142
+
143
  """)
144
 
145
  with st.expander("Try any of these examples", expanded=False):
 
148
  - Desperately dominated by fam1ly sitüatløns, he leaves her.
149
  - You might as well come out to investigate a strang3 n'o?i+se or something.
150
  - But one other thing that we have to re;think is the way that we dy£ our #c!l.o|th?£+s.
151
+ - And he wanted Baltimore city to get that same kind of att£ntløn from the outside, but )i)n)t)r)o)s)p)e)c)t)i)o)n from the inside about what was going on with us.
152
+ - Why do all these _r_e_p_o_r_t_e_r_s, who get praise and money for doing what Assange has done, maintain a cow;ardly silence (at best) while a fellow publisher faces threats of extradition, banning, and espionage charges (which can incur the death penalty), not to mention calls for his as'sa'ss1nat'i'on?
153
 
154
  SPANISH
155
  - _d+i%o"s mío!
156
  - La C0v!d es un 3ng@ño de los G0b!3rno$
157
+ - pl@πd€m1∆ instead of “pandemia” (pandemic)
158
  - se asocian con el m13;d0 y el d'o'lor. g£rønlmo solía decir
159
  - Con las nuevas tecnologías digitales, los agrlcultør£s pueden manejar mejor el uso de sus tierras, su energía y su agua, y prepararse para el mal clima.
160
  - En el tiempo transcurrido entre mi período de escuela %s%3%c%_%n%d%a%r%1%a y el mo'm3n'to de empezar a enseñar vimos surgir el fenómeno de in't£r'net
161
+ - Las pre0c_pac1on3s van desde inquietudes por las ramificaciones desestabilizadoras de una estrategia de salida de la FC, hasta aprehensión por pérdidas de capital en la rápidamente creciente cartera de valores de la Fed (actualmente de 3 billones y en camino a los 4 billones para finales de este año).
162
+
163
+ FRENCH
164
+ - Des poem£s. Je suis obligé d'écrire des poem3s.
165
+ - Placez-le ensuite dans un endroit f'r'ais jusqu'au moment de ser;vir.
166
+ - Cependant, parfois ils rencontrent des sc£narløs _i_n_c_o_n_n_u_s comme indiqué ci-dessus circonstance et font face à une grande quantité de perte de données.
167
+ - Il semble préférable qu'une telle permanence soit organisée de manière uniforme par l'Ordre des barr£a*x francophones et germanophone et l'Ordre des barr£a*x flamands.
168
+ - Elle doivent être mentvrai .fl%a~g'r]a+n/t=e:s avant que nous y prêtions attention.
169
+
170
+ ITALIAN
171
+ - Tornato a londra Händel produsse Ezio, un di'sa'stro sostoco.
172
+ - Vi devo dire cosè successo una 5eI2a quan;do tutti i 5£¬sø₹l hanno smesso di funzionare.
173
+ - 1877 New _h_a_m_p_s_h_1_r_3 diventa l'ultimo stato a concedere i pari diritti politici agli ebr31
174
+ - Per nostra esperienza non possiamo sbarazzarci della sen2a7_i[]ne che la non chiarezza delle cøm'p£'t£n'ze in questi casi gli fa più che comodo.
175
+ - In alternativa, invia a 1llary (proprietario) una richiesta di informazioni tramite il modulo Contatta il proprietario/gestore se prima desideri ricevere un preventivo o maggiori informazioni.
176
+
177
+ GERMAN
178
+ - Die +m+e+i+s+t+e z;£lt fan;gen die le_t3 mit dem #s"t.r}3$1+t an.
179
+ - Würde ich Gleichgesinnte erst nach Einbruch der d_nk3l;heit treffen können?
180
+ - Es war nur eine mi'nimal'e Menge an externen komponent£n notwendig, um ein komplettes sys;tem zu bauen.
181
+ - Die kir;che wurde von dem Schutzpatron der p;£st im Zeichen des Bundes und der dankbark31t £rrichtet.
182
  """)
183
 
184
  # - Why do all these _r_e_p_o_r_t_e_r_s, who get praise and money for doing what Assange has done, maintain a cow;ardly silence (at best) while a fellow publisher faces threats of extradition, banning, and espionage charges (which can incur the death penalty), not to mention calls for his as'sa'ss1nat'i'on?