Вы находитесь на странице: 1из 11

Presentation

On
Filteration of Unwanted Text Messages from
Online Social Networks

Guided By:- Prepared By:-


Mr Amit Yerpude Garima Singh
Associate Professor Mtech CSE
TABLE OF CONTENT

Introduction
Literature Review
Problem Identification
Proposed Methodology
Expected Outcome
Conclusion and Future Scope
INTRODUCTION

Online Social Networking sites (OSNs) helps to connect people easily.


There are various online social sites that are available like Facebook,
Twitter etc. which brought world closer.
In OSN, user creates an account on social sites after which they are able to
perform various actions like adding friends, sharing videos and images.
OSN sites provide some space or area to post the status such a space is
called Wall.
But sometimes people post offensive messages on a particulars wall which
may cause a serious problem to users reputation.
To avoid such kind of serious problem we can apply the combination of
Content based filtering technique with short text based classification.
To avoid such kind of serious problem we can apply the combination of
Content based filtering technique with short text based classification.
Posts will be divided into various classes
Vulgar Message
Hate Message
Offensive Message and
Violence Message
Also how that how much percentage of the post belongs to above
mentioned class.
LITERATURE REVIEW

M.Vanetti proposed Filtered-Wall (FW) architecture which help to avoid


unpleasant messages from being posted on users walls. This system is
based on Machine learning (ML) strategy. ML is used to categorize the
content of a message. In some social networking sites there is no content
based preference for example Facebook. This site allows user to post
comments on another wall. Filtering rules are applied on content of the
messages to filter out the vulgar or offensive words. Blacklist (BL)
improves the efficiency of the Filtered Wall system in which user can
blacklist or block another who has tried to post vulgar content.
Barbara Carminati , describes a filtering rule over their own private wall to
avoid the unwanted messages. User can protect their wall from offensive or
unpleasant messages, to protect users wall from unwanted messages
automated filtering wall system is used which have a tendency to protect
the wall from unwanted messages. The system will block only undesired
messages sent by the other users. The drawback of this is that the miscreant
user will not be blocked. This also means that only the content posted by
the user will be blocked
Fabrizio Sebastiani suggests an approach to the problem of text
categorization based on machine learning. This consists of a general
inductive process that automatically builds a classifier by learning from a
set of pre-classified documents. Text categorization strategy provides
various benefits over knowledge engineering strategies which acquire
specifications of classifiers advised by experts like savings in employment
power and good portability to various domains.
Victoria Bobicev introduced Partial Matching Method(PMM). This method
is used to compress text properties and create a language model adapted for
a particular text. It shows that the method achieves a high accuracy on text
classification. The PPM compression algorithm is based on n-gram
statistical model.
PROBLEM IDENTIFICATION

Difficulty in classifying short text.


No choice is provided, that whether we want to block the
comment or leave it.
No blocking of person.
User
Logged Database
In

Post into Users Wall


User
Ask User
Private
for
Wall
Acceptance
Private
Message
Retrieve
Post Contains:
NO
Short Text Vulgar Words
Classification Violence Words YES
Offensive Block Post
Words If
Apply Hate Words
Filtering
Rule
Short Text Classification
Input:
A short text
A fixed set of classes (Violence, Vulgar, offensive, Hate)
Output:
A learned Classifier and classified short text.

Algorithm:
1: Input Message, OSN Wall Post of User.
2. Apply Preprocessing of the post, removing unwanted and non-sentiment
words.
2: Scan all the text from wall of the user.
3: Perform Text Categorization
- Text Categorization Based on Filtering Rules: Filtering Rules are
customizable by the user. User can have authority to decide what contents
should be blocked or displayed on his wall by using Filtering rules.
EXPECTED OUTCOME

Expected result is to provide a choice of acception and


rejection of blocking the user.
Not only the comment but person can also be blocked.
Short text based classification.
Post will be Classified as:-
Vulgar Message
Hate Message
Offensive Message and
Violence Message
CONCLUSION AND FUTURE SCOPE

This survey paper presents various approaches for designing of


a system to filter not required messages from OSN walls using
different classifiers.
Additionally the flexibility of a system can be increased
through filtering rules as well as blacklist management.
In this context, since the underlying domain is dynamic, the
collection of pre-classified data which are provided for
training purpose which is may not valid for longer time.
Future scope lies in the use of any of the classifiers or
hybridization of theses classifiers can get more accurate
results.

Вам также может понравиться