Word of mouth (WOM) affects the buying behavior of information receivers stronger than advertisements. Opinion leaders further affect others in a specific domain through their new information, ideas and opinions. Identification of opinion leaders has become one of the most important tasks in the field of WOM mining. Existing work to find opinion leaders is based mainly on quantitative approaches, such as social network analysis and involvement . Opinion leaders often post knowledgeable and useful documents . Thus, the contents of WOM are useful to mine opinion leaders as well . This research proposes a text mining - based approach to evaluate features of expertise, novelty an d richness of information from contents of posts for identification of opinion leaders. According to experiments in a real - world bulletin board data set, this proposed approach demonstrate s highp otential ini dentify ingo pinionl eaders. 1
This research identifies opinion leaders using the technique of text mining , since the opinion leaders affect other members via word of mouth (WOM) ons ocialn etwork s. WOM definedb yA rndt (1967) is an oral person - to- person communication means between an inform ation receiver and a sender, who exchanget hee xperienceso fa b rand,a p roducto ra service based on a non - commercial purpose . Internet provides human beings with a new way of communication. Thus, WOM influences the consumers more quickly, broadly, widely , significantly and c onsumers are further influenced by other consumers without any geographic limitation (Flynn eta l.,1 996 ).
Nowadays, making buying decisions based on
WOM becomes one of c ollective decision - making
strategies. It is nature that a ll kind s of human
groups have opinion leaders , explicitly or
implicitly
As opinion leadership is relationships between membersi na s ociety,m anye xisting opinion leader identification tasks define opinion leaders by analyzing the entire opinion network in a specific domain, based on the technique of social network analysis (SNA) (Kim, 2007;; Kim and Han, 2009 ). This technique depends on relationship between initial publishers and followers. A member with the greatest value of n etwork centrality is considered as an opinion leader in this network (Kim,2 007 ).
However, a junk post does not present useful information. A WOM with new ideas is more interesting. A spam link usually wastes readers' time. A long post is generally more useful than a short one (Agarwal et al., 2008 ). A focused document is more significant than a vague one. That is, different documents may contain different influences onr eadersd uet o their quality of WOM. WOM d ocuments per se can also be a major indicator for recognizing opinion leaders. However, such quantitative approaches, i.e. number - based or SNA- based methods, ignore quality of WOM and only includeq uantitativec ontributionso fW OM.
Expertise, novelty, and richness of information are three important f eatures of opinion leaders , which are obtained from WOM documents (Kim and Han,2 009 ).T hus, t his researchp roposes a text mining- based approachi n order to identify opinion leadersi na r eal - worldb ulletinb oards ystem.
Besides this section, t his paper is organized as follows. Section 2 gives ano verviewo f
features of opinion leaders . Section 3 describes the proposed network hubs usually contain six aspects, which are ahead in adoption, connected, travelers, information- hungry, vocal, and exposed to media moret han others (Rosen, 200 2). Ahead ina doption means that network hubs may not be the first to adopt new products but they are usually ahead of the rest in the network. Connected means that network hubs play an i
nfluential role in a network, such as an information broke r among
various different groups. Travelerm eans that networkh ubs usually love to travel in order to obtain new ideas text mining approach to identify opinion leaders . from other groups. Information - hungry means that
data set, experiment design and results . Finally, a conclusion and further research worka reg iven inS ection 5. 2
The term “opinion leader” , proposed by Katz and Lazarsfeld (1957), comes from the concept of communication. Based on their research, the influence of an advertising campaign for political election is lesser than that of opinion leaders. This is similar to findings in product and service markets. Although advertising may increase recognition of products or services, word of mouth disseminated via personal relations in social networks has a greater influence on consumer decisions (Arndt, 1967;; Khammash and Griffiths, 2011). Thus, it is
important to identify the characteristicso fo pinionl eaders.
According to the work of Myers and Robertson (1972), opinion leaders may have the followin seven character istics. Firstly, opinion leadership in a specific topic is positively related to the quantity of output of the leader who talks, knows and is interested in the same topic. Secondly, people who influence others are themselves influenced by others in the same topic. Thirdly, opinion leaders usually have more innovative ideas in the topic. Fourthly and fifthly, opinion leadership is positively related to overall leadership and an individual’s social leadership. Sixthly, opinion leaders usually know more about demographic variables in the topic. Finally, opinion leaders are domain dependent. Thus, an opinion leader influences others in a specific topic in a social network. He or she knows more about this topic andp ublishes moren ew information.
Opinion leader s usually play a central role in a social network . T he characteristics of typical g network hubs are expected to provide answers to others in their group , so they pursue lo ts of facts. Vocal means that network hubs love to share their opinions with others and get responses from their audience. Exposed to media means that network hubs open themselves to more communication from mass media, and especially to print media. Thus, a network hub or an opinion leader is not only an influential node but also a novelty early adopter, generator or spreader. An opinion leader hasr ich expertisei na s pecifict opica ndl ovest ob e involved ing roupa ctivities.
As members in a social networ k influence each other, degree centrality of members and involvement in activities are useful to identify opinion leaders (Kim and Han , 2009). Inspired by theP ageRankt echnique, whichi sb asedo nt hel ink structure (Page et al., 1998 ), OpinionRank is proposedb y Zhou et al. (2009) to r ank members in a network. Jiang et al. (2013) proposed an extended version of PageRank based on the sentiment analysis and MapReduce. Agarwal et al. (2008) identified influential bloggers through four aspects, which are recog
nition, activity generation, novelty and eloquence. An influential blog is recognized by others when this blog hasa l ot ofi n links. The feature of activity generation is measured by how many comments a post receives and then umbero f posts it initiates. novel ideas, which may attract many in Noveltym eans - links from the blogs of
others. Finally, the feature of eloquence is evaluated by the length of post. A lengthyp osti st reateda sa ni nfluentialp ost.
Li and Du (2011) determined the expertise of authors and readers according between theirp ostsa ndt hep re to the similarity - built termo ntology. However b oth features of information novelty and influential position are dependent on link age relationships between blogs . We propose a novel text mining - based approach and compare it with several quantitativea pproaches. proper (http://www.google.com/trends) to obtain the first - search timet agf or
significant words in documents.
tag taken from the Google search specific time repository. For example, the first - search time tag for the search term, Nokia N81, is 2007 and for Nokia Windows Phone 8 is 2011. We define three degrees of novelty evaluated by the interval between the first search year of significant words and the collected year of our targeted document set, i.e. 2010. This significant word belongs to normal novelty if the interval is equal to two years. A significant word with an interval of less than two years belongs to high novelty and one with an interval greater than two years belongs to low novelty. We then summarize all novelty values based on significant words used by a member in a social network. The equationo fn oveltyf ora m ember iss hown in ( novi h e 0.66 em 0.33 el ,
eh em el where eh , em and el is the number of words that belong to the group novelty, respectively. s of high, normal and low 2). (2) (3) 3.4
In general, a long document suggests some useful information to the users (Agarwal et al., 2008 ). Thus, richness of information of posts can be used for the identification of opinion leaders. We use both textual information and multimedia information to represent the richness of informationa s( 3) . ric=d+ g , where d is the total number of signif icant words that the user uses in his or her posts and g is the total number of
multimedia objects that the user posts. (1)
Finally, we integrate expertise, novelty and richness of information from the content of posted documents. As each feature has its own distribution and rang e, we normalize each feature to a value between 0 and 1. Thus, the weights of opinion leaders based on the quality of posts becomet hea verageo ft heset hreef eaturesa s( 4). ITM Norm(nov) Norm (exp) Norm(ric) . (4) 3 number of documents that a member initiates plus the number of derivative documents by other membersi st reateda si nvolvement .
Thus, we have one qualit ative model, i.e. ITM, and four quantitative models, i.e. DEG, CLO , BET andI NV .W ep utt opt enr ankingsf rome ach model in a pool of potential opinion leaders. Duplicate members are removed and 25 members are left . We request2 0 human testers, which haveu sed and aref amiliarw ithM obile01 .
In our questionnaire, quantitative information is provided such as the number ofd ocuments that the potential opinion leaders initiate and the number of derivative documents t hat are posted by other members. F or the qualitative information, a maximum of three documents from each member are provided randomly to the testers. The top 10 rankings are also considered as opinion leaders basedo nh umanj udgment .
Due to lack of available benchmark data set, we crawl WOM documents from the Mobile01 bulletinb oards ystem( http://www.mobile01.com/), which is one of the most popular online discussion forums in Taiwan. This bulletin board system allows its members to contribute their opinions free of charge and its contents are available to the public. A bulletin board system generally has an organized structure of topics. This organized structure provides people who are interested in the same or similar topics with an online discussion forum that formsa socialn etwork. Finding opinion leaders on bulletin board s is important since they contain a lot of availably focused WOM. In our initial experiments, w e c ollected 1537 documents , which were initiated by 1064 members and attracted 9192 followers, who post ed 19611 opinions on th ose initial posts. In this data set, the totaln umbero fp articipantsi s9 460. We suppose that ten of 9460 members are considered as opinion leaders. We collect top 10 ranking members from each models and remove duplicates. We request 20 human testers to identify 10 opinion leaders from 25 potential opinion leaders obtained from five models. According to experiment results in Table 1, the proposed model outperforms others . This presents the significance of documents per se. Even INV is a very simple approach but it performs much better than social network analysis models, i.e. DEG, CL O andB ET . One possible reason is the sparse networks tructure. 4.2 Comparison Manys ubt opicsa rei nt heb ulletinb oards ystem s o theset opics f or ms everal isolateds ubn etworks. As we use real - world data, which has no gr ound tervuatlhu aatbioonu atp porpoianciohn sh oleualddeb res u, sead tuos ceorm cpeanrteet rheed Recall Precision meaFs-u re Accuracy difference between models (Kritikopoulos et al. , DEG 0.45 0.50 0.48 0.56 2006). In our research, there are 9460 members in CLO 0.36 0.40 0.38 0.48 this virtual community. We suppose that ten of BET 0.64 0.70 0.67 0.72 them have a high possibility of being opinion INV 0.73 0.80 0.76 0.80 leaders. ITM 0.82 0.90 0.86 0.88
As identification of opinion leaders is treated to be one of important task s of social network Table1 :R esultso f modelse valuatedb yr ecall, analysis (SNA), we compare the proposed model precision,F - measurea nda ccuracy (i.e. ITM) with three famous SNA approaches , which are degree central ity (DEG), closeness centrality (CLO), betweenness centrality (BET) . Involvement (INV) is an important characteristic of opinion leaders (Kim and Han, 2009 ). The 5
Word of mouth (WOM) has a powerful effect on con sumer behavior. Opinion leaders have stronger influence on other members in a society. How to find opinion leaders n opinion has been of interest to both practitioners and researchers.
This research considers this issue from the viewpoints of text mining. integrated text mining model by We propose an extracting three important features of opinion leaders regard ing novelty, expertise and richness of information, from documents.
Finally, w e compare this proposed text mining model with four quantitative approaches, i.e., involvement, degree centrality, closeness centrality and betweenness centrality , evaluated by h uman judgment . In our experiments, wef ound that the involvement approach is the best one among the quantitative approaches. The text mining approach outperforms its quantitative counterparts as the richness of document information provides a similar funct ion to the qualitative features of opinion leaders .
The proposed text mining approach further measures opinion leaders based on features of novelty and expertise.
In terms of possible future work, some integrated strategies of both qualitative and quantitative approaches should take advantages both approaches .
integrated strategy, which uses the text mining based approach in the first step , and uses the quantitative approach based on involvement in the second step, may achieve the b etter performance . Larger scale experiments including topics, the number of documents and testing , should be done furtheri no rdert o producem oreg eneralr esults. of
Agarwal, N., Liu, H. , Tang, L. and Yu, P. S. 2008. Identifying the Influentia l Bl oggers in a Community . Proceedingso fW SDM, 2 07 - 217. Arndt, J. 1967. Role of
Product- Related Conversations in the
Flynn, L. R., Goldsmith, R. E. andE astman, J. K. 1996.
Based on the Improved Particle Swarm Optimization Neural Networks. Proceedings of IEEE Cybernetics andI ntelligentS ystems,6 Jiang, L., Ge, B., Xiao, W. and Gao, M. 2013 . BBS Opinion Leader Mining Based on an
fluentials. The Free Khammash, M. and Griffiths, G. H. 2011. Arrivederci CIAO.com Buongiorno Bing.com - Electronic Word - of- Mouth( eWOM), Anteceden cesa ndC onsequences . International Journal of Information Management, Kim, D. K. 2007. Identifying Opinion Leaders by Using
sis of Opinion Leadership Data CollectionM e thods and Instruments . PhD Thesis, the Scripps College of Comm unication,
Kim, S. and Han, S. 2009. An Analytical Way to Find Influencers on Social Networks and Validate their
Kritikopoulos, A., Sideri, M. and Varlamis, I. 2006. BlogRank: Ranking Weblogs Based on Connectivity and Similarity Features. Proceedings of the 2nd International Workshop on Advanced Architectures
Li, F. and Du, T. C. 2011. Who Is Talking? An
Framework for Word - of- Mouth M arketing in Online Social Blogs . Decision Support Syste ms, 51,