OSN for neural network and the coomunication network
- 1. Introduction Today the most popular, interactive medium to communicate, share and disseminate a considerable amount of human life information are On-line Social Networks (OSNs). Daily and continuous communications imply the exchange of several types of content, including free text, image, audio and video data.
- 2. Information filtering can therefore give users the ability to automatically control the messages written on their own walls, by filtering out unwanted messages.
- 3. Unwanted messages are an inherent part of today’s spamming world. User needs a stringent prevention and a good firewall against such spamming messages. The aim of the present work is to propose an automated system, called Filtered Wall (FW), able to filter unwanted messages from OSN user walls. Motivation
- 4. E-commerce Social Networking sites Web Forums Discussion Panels Applications:
- 5. The present work defines the construction of a system which supports content-based message filtering for Social Networking Platforms (SNP), depending on Machine Learning techniques. Proposed system has relationships with the state of the art in content-based filtering, and with the field of policy-based personalization for SNPs and, generally in web contents. Literature Survey
- 6. Documents refined using content-based filtering are mostly text documents and thus content-based filtering comes nearer to text classification. The process of filtering can be modeled as a case of single label, binary classification, dividing incoming documents into related and non-related types. Multi-label text categorization which tags messages is used by more complicated filtering systems. Content-based filtering
- 7. Proposed filtering policy language allows the setting of FRs conferring to a range of benchmarks that do not scrutinize only the results of the classification process but also the relationships of the wall owner with other SNP users as well as information on the user profile. In the field of SNPs, the majority of access control models proposed so far enforce topology-based access control, conferring to which access control necessity are articulated in terms of relationships that the requester should establish with the source proprietor. Policy-based personalization of SNP contents
- 8. Proposed system utilizes similar concept to identify the users to which a FR applies. Proposed filtering policy language enhances the languages recommended for access control policy specification in SNPs to cope with the extended requirements of the filtering domain. As we are dealing with filtering of unwanted contents, one of the important factors of our system is the availability of a description for the message contents to be accomplished by the filtering mechanism
- 9. Proposed method depends on 3-tier architecture which supports SNP services. Goal of first layer is to deliver basic SNP functionalities. First layer is called as Social Network Manager (SNM). Second layer is known as Social Network Applications (SNAs). Third layer is called as Graphical User Interfaces (GUIs) which is additional layer to support some needed SNAs. System Development
- 10. User collaborate with system by using GUI for the purpose of setting and managing their FRs/BLs. GUI provides the functionality of Fire Walls (FWs), on which only certified messages are displayed according to their FRs/BLs rules. The basic parts of our implemented system are Content-Based Messages Filtering (CBMF) and the Short Text Classifier (STC). Goal of these parts is to organize messages in to set of groups depends on their nature. With the help of STC module, first part accomplishes message separation.
- 11. Architecture which supports to SNP services depends on 3-tier architecture as shown in above figure.
- 12. 1) After arriving into the private wall of one of the contacts in frendlist, the user tries to post a message, which is intercepted by FW. 2) Role of ML-based text classifier is to abstract metadata from the content of the message. 3) This abstracted data by classifier is further used by FW along with social graph and users profiles, to enforce the filtering and BL rules. 4) According to the generated outcome of step 3, either message will be published on wall or filtered by FW.
- 13. Proposed method approach the assignment by defining a hierarchical two-level strategy assuming that it is better to classify and eliminate neutral sentences and then sort non neutral sentences by the class of interest instead of doing everything in one step. This choice is stimulated by related work showing advantages in classifying text and short texts using a hierarchical approach. A set of distinguish and discriminate features allowing the demonstration of fundamental concepts and the collection of a complete and consistent set of instances. SHORT TEXT CLASSIFIER
- 14. The process of extracting a proper group of properties which describes texts of given document is critical, which can also harmful for the performance of overall classification technique. Some strategies were invented for text categorization procedure but accurate or more proper feature set and feature representation has not yet been investigated. Depending on these, we had taken into accounts three different properties as Bag of Words (BoW), Document properties (DP) and Contextual Features (CF). Text Representation
- 15. A message is therefore available only if it is not blocked by any of the filtering rules that apply to the message creator. A filtering rule FR is a tuple (author, creatorSpec, contentSpec, action), where author is the user who specifies the rule. If user profile does not enclose a value for the attributes referred by a FR, the wall owner can make a decision whether to block or notify messages. Filtering Rules-
- 16. Definition. (Filtering rule). A filtering rule (FR) is a tuple consisting (author, creatorSpec, contentSpec, action), where: • author stands for the user who describes the filtering rules; • creatorSpec is a creator specification implicitly denotes a set of SNP users; • contentSpec is a Boolean expression defined on content constraints of the form (C; ml), where • C is a class of the first or second level and • ml is the minimum membership level threshold required for class C to make the constraint satisfied; • action є {block; notify} denotes the action to be performed by the system on the messages matching contentSpec and created by users identified by creatorSpec. FR Rule
- 17. The concept of Blacklist Management is used to bypass messages from unwanted peoples, irrespective of what they exactly consists of. Through BL rules, wall holder is capable to block unknown persons, peoples with which wall holder have only indirect relationships or peoples about whom wall holder have some cheap opinion. BLs is directly supervised by the system which should be capable to establish who are the users to be placed in the BL and decide when user’s retention in the BL is finished. Blacklist
- 18. We use two measures based on user’s bad behavior as: 1) if user has been injected into blacklist for more times than some defined threshold, then that user will remain into blacklist unless user’s behavior is not improved. 2) Relative Frequency (RF) is used to catch bad behaviors of users. The task of RF is to find out those users whose messages always try to break down the filtering rules.
- 19. Definition (BL rule). A BL rule is a tuple consists of (author, creatorSpec, creatorBehavior, T), where: • author is the SNP user who specifies the rule, i.e., the holder of wall; • creatorSpec is a creator specification; • creatorBehavior holds two components as RFBlocked and minBanned. • RFBlocked = (RF, mode, window) is defined such that: where #tMessages is the total number of messages that each SNP user identified by creatorSpec BL Rule:
- 20. #bMessages is the number of messages out of which messages in #tMessages have been blocked. minBanned = (min, mode, window), where min is the minimum number of times in the time interval specified in window that SNP users identified by creatorSpec have to be inserted into the BL due to BL rules specified by author wall. T denotes the time period the users identified by creatorSpec and creatorBehavior have to be banned from author wall.
- 21. Experimental Analysis: Fig: a message filtered by the wall’s owner FRs
- 22. The set of classes considered in our experiments is Neutral; Violence; Vulgar; Offensive; Hate; Sex. All messages posted by users are checked against these set of classes. 1266 messages from publicly accessible groups have been selected and extracted by means of an automated procedure that removes undesired spam messages. For each message, stores the message body and the name of the group from which it originates. The messages come from the group’s web page section, where any registered user can post a new message or reply to messages already posted by other users.
- 23. The system uses Machine Lerning soft classifier to implement FRs and BL to boost filtering preference. FRs should allow users to state constraints on message creators. Proposed system allows user to decide to describe BL list and to decide who has to be banned from their walls and for how long. Therefore, a user might be banned from a wall, by, at the same time, being able to post in other walls. Conclusion-
- 24. By analyzing the user’s behavior in the past, learning methods applied for content-based filtering in proposed system find out the proper and relevant documents. This technique yields to restrain to user to prepare documents similar to those already seen. So, the approach is recognized as over- specialization problem.
- 25. M. Vanetti, E. Binaghi, B. Carminati, M. Carullo, and E. Ferrari, “Content-based filtering in Social Networking Platforms,” in Proceedings of ECML/PKDD Workshop on Privacy and Security issues in Data Mining and Machine Learning (PSDML 2010), 2010. M. Chau and H. Chen, “A Machine Learning approach to web page filtering using content and structure analysis,” Decision Support Systems, vol. 44, no. 2, pp. 482–494, 2008. R. J. Mooney and L. Roy, “Content-based book recommending using learning for text categorization,” in Proceedings of the Fifth ACM Conference on Digital Libraries. New York: ACM Press, 2000, pp. 195–204. N. J. Belkin and W. B. Croft, “Information filtering and information retrieval: Two sides of the same coin?” Communications of the ACM, vol. 35, no. 12, pp. 29–38, 1992. References
- 26. S. Pollock, “A rule-based message filtering system,” ACM Trans- actions on Office Information Systems, vol. 6, no. 3, pp. 232– 254, 1988. P. W. Foltz and S. T. Dumais, “Personalized information delivery: An analysis of information filtering methods,” Communications of the ACM, vol. 35, no. 12, pp. 51–60, 1992. P. E. Baclace, “Competitive agents for information filtering,” Communications of the ACM, vol. 35, no. 12, p. 50, 1992. A. Adomavicius, G.and Tuzhilin, “Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions,” IEEE Transaction on Knowledge and Data Engineering, vol. 17, no. 6, pp. 734–749, 2005.