tgbully's Introduction

TGBully - Temporal Graph-based Cyberbullying Detection

Implementation of Temporal Graph-based Cyberbullying Detection Model [1].

The TGBully framework consists of three major components: (1) a semantic context modeling module that encodes each comment by considering both its textual content and user’s language behavior reflected from her/his historical comments, (2) a temporal graph interaction learning module that constructs a temporal graph and models the dynamic user interaction with a bully-featured GAT. The proposed GAT jointly captures topic coherence and temporal dynamics in the modeling process; and (3) a session classification module that attentively aggregates information from user interaction into a session representation, based on which it then classifies the session into a bullying/non-bullying session.

Code Usage

Download all the required data and word embedding files as instructed by the data README.
Run the script Instagram_code.ipynb and Vine_code.ipynb for experimental results on the two benchmark datasets.

Environment

python == 3.6.7
cudnn == 7.1.2
keras == 2.2.4
numpy == 1.16.4
tensorflow-gpu == 1.12.0

Reference

[1] Suyu Ge, Lu Cheng, and Huan Liu. Improving Cyberbullying Detection with User Interaction. The Web Conference (WWW), 2021. [paper]

tgbully's People

Stargazers

Watchers

tgbully's Issues

'List' object has no attribute 'keys'

First, thanks your nice work.
But I run Instagram_code.ipynb, in "Load user history data" part, line 24

# Load user history data
userhistory=[]
pattern = re.compile(r'^userID:.*\susername:.*\scommentText:.*$') 
for filename in os.listdir(normalUser_path):
    with open(normalUser_path+filename,'r',encoding="utf8") as f:
        lines=f.readlines()
        for line in lines:
            sub = pattern.search(line) 
            try:
                sub.group()
                userhistory.append(line)
            except:
                pass
for filename in os.listdir(commonUser_path):
    with open(commonUser_path+filename,'r',encoding="utf8") as f:
        lines=f.readlines()
        for line in lines:
            sub = pattern.search(line) 
            try:
                sub.group()
                userhistory.append(line)
            except:
                pass
for person in userhistory.keys():
    for i in range(len(userhistory[person])):
        userhistory[person][i]= word_tokenize(userhistory[person][i].lower())
        if userhistory[person][i][0]=='@':
            userhistory[person][i]=userhistory[person][i][2:]

and it throw the error look like

---> 24 for person in userhistory.keys():
     25     for i in range(len(userhistory[person])):
     26         userhistory[person][i]= word_tokenize(userhistory[person][i].lower())

AttributeError: 'list' object has no attribute 'keys'

in previous code, userhistory is a list object rather than a dict object.
is there something missing or something wrong in the code?

the dataset link is broken

Hi, thanks for sharing this great project.

The link of the datasets(both Ins. and Vine) are broken. Could you please provide a new one?

Recommend Projects

gesy17 / tgbully Goto Github PK

tgbully's Introduction

TGBully - Temporal Graph-based Cyberbullying Detection

Code Usage

Environment

Reference

tgbully's People

Stargazers

Watchers

Forkers

tgbully's Issues

'List' object has no attribute 'keys'

the dataset link is broken

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent