Anda di halaman 1dari 6

International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169

Volume: 4 Issue: 7 217 - 222


____________________________________________________________________________________________________________________
Optimal Advisor Search for Knowledge Sharing in Collaborative Environment
Vandana S.Lokhande1, Prof. Narendra Gawai2
1
M. Tech Student , Dept of CST, UMIT, S.N.D.T Womens University, Santa cruz, Mumbai.
2
Assistant Professor ,Dept of CST, UMIT, S.N.D.T Womens University, Santa cruz, Mumbai. .
Abstract:-In day todays life, many people are accessing the information on the web. In collaborative environments it may be possible that
numbers of web users may trying to access the similar information on the web for particular domain. Therefore knowledge sharing in
collaborative environment is major research problem for many research communities. For example, in an organization several departments may
successively need to buy anti-virus software and employees from these departments may have studied online about different anti-virus software
and their features independently. It will be productive to get them connected and share learned knowledge. The domain which is further divided
into sub domain is known as micro-aspects in web surfing data. Mining these micro-aspects is critical as it can provide a detailed description of
the knowledge gained by a person, which is the basis for advisor search. In this project, We investigate knowledge sharing system in
collaborative environments. We propose to analyze members Web surfing data to summarize the knowledge acquired by them. A two-step
framework is proposed for mining knowledge: (1) Web surfing data is clustered into tasks (Domain)by a k-means algorithm 2) Mine micro
aspects (Sub domain)in each task. Finally, search method is applied to the mined results to find proper advisor.

Keywords - Knowledge sharing , Collaborative environment, Advisor Search, Micro aspects , Task, Web surfing data, Clustering.
__________________________________________________*****_________________________________________________

Introduction: user that belong to the same task. Sessions are atomic units
In a collaborative environment, it is regular practice that in our analysis. A task can be further decomposed into
members are trying to acquire similar information on the micro-aspects(Sub Domain). A micro-aspect could be
web in order to gain specific information in one domain. We roughly defined as a significantly more cohesive subset of
present a new method how to identify, how to enable such sessions in a task. For example, the task Operating System
knowledge sharing . A two-step methodology is proposed might contain Applications of operating System and
for mining knowledge: (1) Web surfing data is clustered into Advantages Of Operating System. For example, User2
tasks (Domain)by a k-means algorithm 2) Mine micro starts to surf the web and searching for advantages , which
aspects (Sub domain)in each task. This method is different has already been studied by User3. In this case, it might be a
from the traditional expert search in that expert search aims good idea to consult User3, rather than studying by her self.
to find domain experts based on their associated documents Such recommendations are provided with this methodology
in an enterprise repository, while the goal of this proposed by analyzing surfing activities automatically.In this
work is to find proper advisors who are most likely example, not necessarily User3 is an expert in every aspect
possessing the desired piece of knowledge based on their of Operating system advantages; however, due to his
web surfing activities. In order to analyze the knowledge significant surfing activities in Operating system
acquired by web users, new method is proposed to log and advantages, it is reasonable to assume that he has gained
analyze users web surfing data. Users interactions with the enough knowledge in this area so that he can help User2. In
web can be segmented into different tasks, e.g., these case, resorting to a right person could be far more
Operating System and shopping. Textual contents of a efficient than studying by oneself, since people can provide
task are usually cohesive. This project defines a session as digested information, insights and live interactions,
an aggregation of consecutively browsed web contents of a compared to the Web.

Figure 1: Knowledge Sharing In Collaborative Environment

Most people in collaborative environments would be happy specific problems. However, finding a right person is
to share experiences with and give suggestions to others on challenging due to the variety of information needs.

217
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 7 217 - 222
____________________________________________________________________________________________________________________
Related Work 4) Discriminative models of integrating document evidence
1)KrisztianBalog and Group have presented paper on anddocument candidate associations for expert search, in
Formal models for expert finding in enterprise Corpora Proc.33rd Annu. Int. ACM SIGIR Conf. Res. Develop.
In this paper auther presented two general strategies to Inf.Retrieval, 2010,pp.
expert searching given a collection of document. The first In this paper ,expert identification(Who are the experts on
directly models an experts knowledge based on the topic X ? ) and expertise selection (What does expert Y
documents that they are associated with.Second locates know ?). In expert identification query is passed related to
documents on the queried topic and then finds the associated the topic X and the relevant documents are generated as
expert. [2]. output. In the output the relevance of each documentis found
and those documents who have the highest relevance willbe
2)KrisztianBalog,LeifAzzopardi ,Maarten de Rijke ranked high.From the highest rank document it will be easy
published a article on A language modeling framework for to recognize expert.
expert finding .
In which they introduce and detail language modeling 5)Ziyu Guan, Shengqi Yang, Huan Sun, MudhakarSrivatsa,
approaches that integrate representation ,association and and Xifeng Yan have proposed Fine-Grained Knowledge
search of experts using various textual data sources into a Sharing in Collaborative Environments .2015 :In this
generative probabilistic framework. They introduces two paper auther has proposed a method to find out Advisor
model in which first model finds out prominent topic in the using web surfing and browsing contents.
document and in the second model they identify important Web surfing data gives more accurate results than traditional
documents for a given topic and determine who is most documentbased method.Gaussian mixture model is used for
closely associated with these documents. [3]. clustering session in each task.In order to implement to mine
micro aspects in each task Hidden Markov Model is used.
3)Finding Experts in Community-Based Question- [1]
Answering Services. In this paper there is a network of self-
declared experts to answer other peoples questions. The Proposed System :
Wondir, is a free, publicly available, live question and We have developed the System for knowledge sharing in
answerengine that connects people with questions to people collaborative environment using k-means algorithm. The
with answers. People using such services are like a users of this system are users those are searching
community anyone can ask, anyone can answer, and information over Web. Input for this system is web surfing
everyone can share, since all of the questions and answers data in text format, which is collected using
are public and searchable immediately. The key is getting BrowsingHistoryView software . This system has following
the right question in front of the right person[4] processs

System Architecture

Figure 2: Architecture Of The System

This system has following Processs BrowsingHistoryView software is used for collection of
Collection Of Dataset we propose to log and analyze users data.
web surfing data, which reveals a users knowledge gaining
process. Web surfing data provides more comprehensive Session clustering module: Clustering is required for
information about the knowledge gaining activities of users. recovering tasks from sessions because the contents
generated for the same task may be textually similar while
218
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 7 217 - 222
____________________________________________________________________________________________________________________
for the different task may be different. Clustering is done by Algorithm for Clustering Process
using K-means algorithm. Input for session clustering is 1) Read Dataset
dataset and output is number of clusters i.e task. In this 2) Store it in string variable data
module We have divided browsed dataset into predefined 3) Convert in into lowercase
Task (clusters) i.e Java, Operating System , Data Structure 4) Split it by newline
etc. 5) Read Title,url and User Profile from dataset
6)Remove All stopwords from Title
Micro aspect module: Micro aspects in the task are already 7)Perform stimming process on Title
similar with another . If we model each component 8)Compare each word of Title with the domainWords of
independently it is likely that we mess up sessions from each Domain(Task) i.e c,java,OS,DS,RDBMS etc
different micro aspect. Therefore we should model different 9)If maximum match between domain words and Title
micro aspects in a task jointly, separating the common found assign that Title to that Domain,(Task)
content characteristics of the task from the distinctive
characteristics of each micro aspect. This module works Algorithm for Mining Micro Aspect
according to the fired query. It finds users profile who has 1)Take user input query in one string variable
searched data related to fired query. It gives url of the web 2)Identify Domain Of that query
pages related to fired query ,user profiles who has searched 3)Compare this string variable with Title Field of related
that web pages , number of hits by the users to that web Domain
pages in descending order. 3))Retrieve All those Titles , urls, Profiles , and count
matches to input query from specific domain
Advisor search module:-Advisor search is dedicated to
retrieving people who have desired piece of knowledge i.e Algorithm for Advisor Search
who has visited more number of web sites related to fire 1) For advisor search we first find the link. which most
query. Here for finding advisor search we will use web visited whose hit count is higher .
sessions instead of documents. Because sessions will give 2) i.e Maximum hits made by user by considering count
accessed information in micro aspect level. Input for this field.
module is micro-aspects and output is url of answer to the 3) Then we have to find user who visit this link maximum
query which is hitted maximum number of time by the user. number of time.
and contact details of the user who has visited more number 4) Display url and contact detail of advisor.
websites related to fired query.
Advantages Of The Proposed System
Implementation: 1) This system gives exactly url for fired Query.
Algorithm for Data Collection 2) It saves time as get exact answer.
1) Use BrowsingHistoryView for data collection. 3) As we are providing contact details of optimal advisor so
2) Allowed diff Users to search on different topics. it becomes easy to get advice from that advisor.
3) Web surffing data is stored in .txt file. 4) This system provides best results for micro-aspect
mining.

Results

Figure 3: Sample Dataset

219
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 7 217 - 222
____________________________________________________________________________________________________________________

Figure 4:Clustering

220
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 7 217 - 222
____________________________________________________________________________________________________________________

Figure 5:Mining Micro Aspect

Figure 6:Advisor Search

Conclusion And Future Scope by people from web surfing data. This system finds the url
We introduced the system Optimal Advisor Search For to related query which is hitted maximum number of time
Knowledge Sharing In Collaborative Environment , which and the advisor who has searched that url more number of
is desirable in practice. We have mined knowledge reflected time. Experiments on real Web surfing data shows good
221
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 7 217 - 222
____________________________________________________________________________________________________________________
results. For future scope software used for data collection finding.www.elsevier.com/locate/infoproman
not showing the time duration for which user spending on Information processing and management,June 2008.
specific web page.Can implement using Page Ranking and [4] X. Liu, W. B. Croft, and M. Koll. Finding experts in
HITS(hyperlink-induced topic search) community-based question-answering services. In
CIKM, pages 315316, 2005
[5] D. M. Blei, T. L. Griffiths, M. I. Jordan, and J. B.
References: Tenenbaum.Hierarchical topic models and the nested
[1] Ziyu Guan, Shengqi Yang, Huan Sun, MudhakarSrivatsa, chinese restaurant process. In NIPS, 2003
and Xifeng Yan Fine-Grained Knowledge Sharing in [6] P. R. Carlile. Working knowledge: how organizations
Collaborative Environments. IEEE transactions on managewhat they know. Human Resource Planning,
Knowledge and data Engineering, vol.27,No. 8,August 21(4):5860, 1998.
2015, pp.2163-2174 [7] H. Deng, I. King, and M. R. Lyu. Formal models for
[2] K. Balog, L. Azzopardi, and M. de Rijke, Formal expert finding on dblp bibliography data. In ICDM,
models for expert finding in enterprise corpora, in Proc. pages 163172, 2009.
th
29 Annu. Int. ACM SIGIR Conf. Res. Develop. Inf. [8] A. K. Jain. Data clustering: 50 years beyond k-means.
Retrieval, 2006, pp. 4350. Pattern Recognition Letters, 31(8):651666, 2010.
[3] KrisztianBalog, Leif Azzopardi ,Maarten de Rijke A [9] R. Jones and K. Klinkner. Beyond the session timeout:
language modeling framework for expert automatic hierarchical segmentation of search topics in
query logs.In CIKM, pages 699708, 2008.

222
IJRITCC | July 2016, Available @ http://www.ijritcc.org
_______________________________________________________________________________________

Anda mungkin juga menyukai