Lectures on Computational Advertising

Video lectures on Computational Advertisement by

Andrei Broder, Fellow and Vice President for Computational Advertising in Yahoo! Research

Computational Social Science promises to provide novel solutions to social and management sciences problems. Computer science studies has been emerged from traditional Information systems to novel complex problems. Computational advertisement is one of the most important field of study for all top IT organizations, including Microsoft, Yahoo, IBM and Google.


http://bangalore.yahoo.com/bigthinkers/video.php?eid=4&vid=16

http://bangalore.yahoo.com/bigthinkers/video.php?eid=4&vid=17

http://bangalore.yahoo.com/bigthinkers/video.php?eid=4&vid=18

http://bangalore.yahoo.com/bigthinkers/video.php?eid=4&vid=19

The On-Click Behavior for Digital Marketing

Identification and modelling  of human behavior is always an interesting and important study since long time. Astrologers, palmists make their try to predict how an individual will behave and what move a person will take in future. Human physical aspects also considered important about the prediction of expected actions of individuals. These Bio-metric attributes consider shape of ears, forehead structures, walking style, gaze style and of course hand writing patterns also been studied a lot.

With the rapid growth of internet users till Dec 2009 ,1,802,330,457 people are using WWW i.e. 26% of the world population. The 53.5% of european uses internet and the growth rate of internet users is 305.1% in 2000-2001. However, in Asia People using WWW pages and leaving trajectories. The web pages on WWW  forum provide back links and forward links to move on. Through these links a user move on and back in pursuit of his

'interest'. Once a surfer found his/her interested content, image, video or discussion on a page he get stick with it.

Clicking on back/forward links depends upon certain liking and disliking of individuals. Therefore an individual made 'clicks' and out reach his desire material. 

A visualization of the network structure of the Internet by Hal Burch and Bill Cheswick, courtesy of Lumeta Corporation

An individual living in Manchester City is a fan of Manchester United, therefore, its quite simple to comprehend that why he visits so many times to view Facebook page of Manchester United. Secondly, he is expected to purchase tee shirts, shoes or other MU logo based sports wear. Moreover, not only that particular individual like MU he will also influence his (social networking Facebook, Twitter) friends. It also depends upon his own position in his social context i.e. lets consider following is the friends network over twitter (who follow whom, here we have few nodes followed by a great number of followers. 

So if that MU fan is one of the most centric persons over twitter or face book, then its highly predictable that he is expected to influence others about his own favorite club.

Therefore, people having similar interests are found to click similar kinds of web sites, secondly, patterns could be found from browsing history of people having common interests.

On click and browsing patterns helps us to understand common behavior of individuals having common interests. Social network analysis plays significant role in identifying communities, important individuals and cliques which reflect common attitude and behavior. This outcome can be used to develop marketing strategies and the development of computational advertisement basis.

Computational Advertisement

Marketing products through advertising is in very beginning,. The advancement of technology further speeded up the pace and frequency of adds for marketing products. Web based advertising is one of the advanced way of marketing products, however, advertiser are concern about the accuracy of the add and its target market sufficient to grab the attention for web based users. Aside from advertisers interest behaviorists and social scientists are also curious about what actually influence a person to click certain adverstisment on the virtual world of the Web. The social scientists are now entering into the new world of computer social sciences which will not only enable them to quantify the behavior but may also assist them to get more knowledge about the various actions elicited by an individual or groups while using a web. 

In future advertising agencies will be more interested to outreach their target audience on www forums. TV advertisements will soon be shifted towards particular web pages of common interest. Multi million WWW users are well knitted together and a lot of social networks have been emerged from such informal contacts. These networks are spread across the languages, countries and even continents. An individual in a social network get influenced by its neighbors and also influence others more effectively and quickly as compared to our social contacts in our locality.

Micro blogging also ignition the quick spread of ideas over www.

Consisely, a web based advertisement and the reason for clicking the web based advertisement may also be identified through computational social sciences ideas. Computer science has to contribute in this new field i.e. Computational Advertisement. Text mining approaches, classification of relevant web pages (http://www.cse.lehigh.edu/~brian/pubs/2007/classification-survey/LU-CSE-07-010.pdf), traffic monitoring on web sites, identifying important centric web sites and many other problems still waiting for computer science community to work upon. (Good resource on Computational Advertisement http://www.stanford.edu/class/msande239/)

At the end Social Network Analysis (Steve Borgatti) tools with other computer science techniques and marketing, advertisement frameworks have to contribute a lot under the new discipline i.e. Computational Social Science.

COMPUTATIONAL SOCIAL SCIENCE FOR CULTURAL HERITAGE

Computational social science is contributing in many different fields. My presentation abstract has been accepted in a prestigious conference

The Fifth SEEDI Conference: Digitization of cultural and scientific heritage      http://www.digi.ba/?jezik=bos&n=9

Importance of preservation of cultural heritage is growing quickly with the current pace of cultural diffusions. Old cultural habits, ethics, moral values and attitudes are getting shape of new definitions. I will discuss the topic in detail in later posts. My abstract is presented here:

MAASHRA: A Social Network Analysis based Text Mining Framework for Tracing Patterns in Literature

The progress of recent technologies in the different fields of science has made it possible to comprehend the evolutionary trends of cultures in different regions. Today literature can help us to explore the evolution of culture in the age of paper (i.e before digital era). In this paper we discuss a framework which helps to discover patterns in story telling, folklores, myths, and poetry of a region and locating there origins and tracing its spread in that era. Text mining techniques are discussed to extract information from literary content which further uses machine learning techniques for classification under separate concepts. Communities influence other communities which imprints their trends in literature particularly. The proposed framework uses social network analysis tools which incorporate graph theoretic approaches to trace groups of common interest, influential individuals, and spread of concepts in different geo-locations.  

Social network analysis tools use for finally co-relating authors, geo-locations and conceptual evolution.

THE MOVERS, THE SHAKERS , THE PLAYERS

Following is the friendship network of Dobro's company. In the following friendship network nodes represents employees and edges represent friendship relations within the employees accross the department.. The friendship network can be identified by frequency of emails, phone calls or by asking individual employee. The friendship network tells us that instead to convincing every individual in the organization  we do identify very fewer 'highly centric' individuals that could be potential 'message' or 'information' carriers. These 'highly centric' individuals are circled in the following figure. 

These 'highly centric' individuals are the potential carriers of knowledge in the organization, these individuals spread and diffuse information throughout the organization very quickly across the departments and in any formal hierarchy. The 'quad' of these highly centric individuals has been noted as the most important interaction sub network within the organization. This quad can play either positive or negative role in an organization depending upon their attitude. One can exploit this network to spread whatever you want in the organization, otherwise, they will exploit their contacts in other way round. 

We can quantify the flow of information through these individuals over the network. Several social network analysis tools exists to study spread of information, importance of individuals, importance of sub networks and others. I will discuss it soon.

Change Management in Network Science

Change Management and Key Integrators with Social Network Analysis

Consider a recent scenario, a big tobacco group purchased a local tobacco company named 'Dobro'. Dobro's employees strength is 2000 with 20 managers and directors The famous brand company wants to align the work environment at the local chapter with their international standards. The working culture at 'Dobro' has to be re visited with ethical standards. The parent company arranged series of meetings with all the key formal figures of the local chapter. The parent company took every initiative to minimize the distance between their managers and directors with 'Dobro' company. Surprisingly, instead of bringing any change in the 'Dobro' company employees (other than the managers/directors or formal key people) they found negative impression for the integrators.

Case 1:

Does all the 2000 employees in the company are not worthy and not open to bring any change in the organization? is that the real problem?

Case 2:

Does 'Dobro's' managers not capable enough to understand what parent company want from them?

Case 3:

Does parent company missing the key players in the organization? the invisibles.

Bringing compliance during change management is always challenging in most of the cases and generally fails to maximize the performance of the two integrators.

Infact, organiazations often launch initiatives without understanding the inner working of formal and informal networks involved, relying on the philosophy that the most influential person is the one who has the key position in the formal hierarchy. However, Social Network Analysis SNA, reveals that the key players are not worthy enough in formal networks, informal networks, rather, evolved wherever people cooperating on the basis of different patterns of thinking and value systems, displaying styles in communication.

SNA define methods which exploit network structures to find the hidden structures, sub networks, communication pathways, trust networks, decision networks, support networks and

potential key integrators. Social Network experts then uses this information to quantify significance and behaviors of individuals and groups. These findings are in the form of smaller networks with different contexts as discussed earlier. Discussion network tells different story and trust network spells out 'who influence whom' and so on.

In any part of the world its not possible that 'all the employees are not interested in working' or 'managers could not understand the strategy of the new parent company'. However, the parent company failed to identify the real 'key players' who influences the organization and does not hold any key position.

Here comes the art of SNA needed to come up with strategies to implement requirements in the child company by influencing the key players rather than the prominent figures in the formal hierarchy.

How a leader influence his follower?

Traditional Approach: An Individualistic Perspective on Leadership vs Network Perspective on Leadership

Traditionally a leadership is assumed as a born quality, however here we make a comparison of traditional aspect of leadership with the network perspective.

All the great leaders have had one characteristic in common: it was the willingness to confront unequivocally the major anxiety of their people in their time. This, and not much else, is the essence of leadership. John Galbraith(1979)

The art of leadership is saying no, not yes. Tony Blair

Leaders need to be optimists. Their vision is beyond the present. Rudolph Giuliani

Innovation distinguishes between a leader and a follower. Steve Jobs

Some of the best business and nonprofit CEOs I’ve worked with over a sixty-five-year consulting career were not stereotypical leaders. They were all over the map in terms of their personalities, attitudes, values, strengths, and weaknesses. Peter Drucker

In view of recent prominent leaders a leader should have an certain traits critical thinker, optimistic, ambitious, innovator, and on the behavioral side “ability to face challenges”, “must not give up hope”, “ability to say, no”, or “willingness to confront people anxiety”. In view of management a leader should have several qualities like openness, humility, charisma, assertiveness, etc.

In reality, “Leadership is intrinsically subjective, and an attitude, strength or value that makes someone a leader might not work for someone else” Galo.

In future, personal qualities will not be enough to grade someone as a good leader rather how he influence his followers.

Who could be a good Leader

In the previous post three different groups were given to identify leaders independently and their role on their followers. Following is map of interaction of people, where nodes are humans and edges are ‘communicate’ and it can be emerged from several contexts:

· Management science students playing football shown in group 1, students of same department playing soccer in group 2, and students from different department playing both games in group 2.

· Researchers working in an institute on similar topics in group 1, researchers working on different topics at same institute in group 2 and researchers working in another institute having research on different topics group 3.

· Friendship networks of employees with similar departments in group 1 and group 2 and their friendship with another department in group 3.

For instance, consider students sports group case. Students are interested in electing some leader: A leader which will be influential in his own group with have enough influence to other groups to convey and propagate adequately. Each student have its own characteristics to be a leader, can we use the friendship or communication network to choose the leader?

Suppose if we choose Graham as a leader, would James be a good candidate?. He has only one friend or communicating with one student in group 2 and have no connection with any member of group 2 and group 3. If Graham would have to persuade most of the fellows in his point then he has very little leverage to do so as he don’t know much of others. If there were an important problem that student of group 3 would want to address, Graham might not know that unless it affects him too. Indeed, Graham would not be a good leader for his community. Notably how Graham could have all the above mentioned traits of the best leader but still not be a good representative due to lack of means to influence the rest.

Then what about Julia, she has the highest degree (the measure of the number of links) among all students. She is in a better position to identify the problems of the group as she communicates or friend with a lot of students. She should also be able to use her ability to influence, persuade and follow her neighbors in the link graph. Is Julia then the best leader for all students group? She might. But there are some reasons to be noticed:

Julia might suffer an information/friendship overload. Lot of people communicate with her, lots of students wants support from her which make it difficulty to exert influence and she become unable to process all the information.

Julia does not communicate with anybody in group 2, and she has only very few connections in group 1. She will find it very challenging to exert her influence with those students in group 2. Therefore, she might be unaware of the issues concerned with group 2.

Julia communicates with a lot of people, but she might not have any friend among them. So what we can call this situation that a marginal role in a “deeper” network.

Then who should be the best leader here? see u soon

Ins Network Advisor

Note: Influenced by Edoardo Gallo in “A Network Perspective on Leadership”

Network Perspective of Leaders (A review)

What a leader should know about the network

Leadership has been taken into account in an individualistic perspective since many years. Qualities of leadership have been considered in a micro-oriented manner. Available underlying assumptions that put the leader in the spot light by leaving all the rest outside will not proceed further due to the interconnectedness nature of growing society. How an individual ( i.e. a leader) will be able to influence the mass or which individual will have a tendency to influence the mass are totally different perspectives in the identification of leaders.

In future a leader should be equipped with the information of “network perspective” of the organization, or society. The network perspective of an organization tells us who influence whom? Who supports whom? Who trust on whom? Friends in the set up? Who conflicts with whom? Performance and viability of individuals and groups, information flow pathways, bottlenecks of responsibilities and many more.

Identification and strengthening communication ties with influential individuals is important for a leader to develop two way ‘trafic’ (information). An individual could be important as acting like a bridge between communities and can be tagged as ‘key integrator’, and on the other hand an individual could be a crucial such that if we will pick out the group will be divided on multiple groups in terms of ‘trust’.

Consider the following cases: We have three different topologies of interactions within different groups, who is the most important individual in each group? Does it make any sense or not?

Comment on the above 3 Figs 1a, 1b, and 1c: which individual is most important in each group?

Is there any individual leader/influential person present in each group who really influence the others?

See u again soon

Ins Network Advisor

Social Network Analysis: A tool for interaction and behavior modeling

We are living in a combination of closely knitted networks, just like a spider web.We exit a network only to enter into another one. Relationships, interactions, and likeness and dis likeness can be represented with the help of graphs or networks. At home, our interactions with wife, parents, children, neighbors and relatives can be represented as nodes with edges in a graph.

At business, like in a supply chain, a supplier provides supplies to other entity that could also be a supplier to another entity. These interactions can be represented through graphs as well. Moreover, our trust, discussion, friendship attitudes at our offices can also be represented through networks and graphs.

These graphs or networks are just like our friendship interactions at face book and are named social networks as they are based on our social relationships. Identification of influential individuals/groups, group of people having common interest, people acting as bridges between two communities all can be done through social network analysis (SNA). SNA tools enable us to identify structures within graphs, important nodes, paths (communication, responsibilities, information flow) and many more.

For e.g.

Consider a protein-protein interaction(PPI) network, in a bacteria a PPI consists on about 4000 protiens. Which proteins are the most important in all others? which set of proteins are most connected (max interaction with each other making clique)? which protein causing series of actions by acting as a bridge between two communities? and many other issues can be solved through SNA tools.

Figure. In the C. elegansinteractome, single proteins (circles) partner up (indicated by lines) to form an interconnected network that controls cell function. The inset details a small part of the network. (Image courtesy of Marc Vidal)[1]

We will discuss Complex Networks and SNA in next posts.

see u again

Ins Network Advisor

[1]http://proteinfunction.net/index.php/About_PPI_Maps

Computational Social Science and Social Network Analysis Resource

In this blog I would like to share some important links that could be a great resource for a novice in Computational Social Science:

Prominent Groups and Research Institutes on Social Network Analysis and Computational Social Sciences:

George Mason University

Santa fe Institute

Institute for Quantitative Social Sciences, Harvard University

Human Complex Systems, UCLA

MIT Reality Mining

About Social Network Analysis

University of Essex

International Network on Social Network Analysis

University of Greenwich

Fas Research

ERCIM Graph and Network Analysis Cluster

IBM Watson Research Center

IRE Social Network Analysis

Try these links, valuable resource will be helpful to you as a novice.

see u again

Ins Network Advisor

Computational Social Science (CSS) an Introduction

We are living in the age of technology and information, our each and every action is being recorded in the form of data and advancement in technology has allowed us to transmit that data very quickly on long distances. If we are driving our car and take exit on a city our car number get recorded in the visual camera and our exit information saved in our movement, after eating pizza or pratha on some restaurant in other city when we pay the bills through the

credit or debt card our transaction will be the signature of our place, time and activity, similarly if we are on airport and checking our black berry or email on net café then our activity is being recorded and can be followed. All of these forms of data could be helpful to connect people of common interests, common objectives or facing similar problems. Analysis patterns of group behaviors sharing common attributes has become a new challenge which is very requires different from currently existing mathematical, computational, sociological, and psychological tools.
Establishing relationships between people, through technology, on the basis of common interests, behaviors, movements, or profession has been termed as development of social networks. These social networks can be based on hobbies like stamp collectors, supporters of same football team like Barcelona, following advancements in science and technology like Bioinformatics, or discussing social issues like marriage and divorce growth rate. The study of such interactions is evolved to a new field of science i.e. Computational Social Sciences.
Computational social science helps us to study such massive amount of data from different view points. Such as computer scientists and mathematicians are always interested in finding new patterns, structures and information present in graphs, irrespective to their meaning. However, a consumer marketing analyst is more interested in finding buying pattern of people sharing common interest on internet etc.

Computational social science is an interdisciplinary filed of mathematics, computers, management, and psychology. Computational social science deals with the data related to human relationships like trust in an organization for team building or friendship in a society for community development. Connections or relationships between people in an organization or society in terms of formal or informal is represented with the help of graphs. In such graphs nodes represent people and edges represent their relationships.

Applications of Computational Social Sciences (CSS)

1. Behavioral patterns can be identified.
2. Supply chain networks can be studied by several ways i.e. who supplies whom? Who can continue supplies if one or two of his suppliers failed? Etc
3. Human resource management systems like who is the key integrator in different groups within an organization? or who is the most influential person in the organization (can be extracted from informal networks)
4. Trend analysis by relating crimes, behavior and locations.
Many other domains can be covered with CSS. In this blog I will try to discuss such problems, their solution methodologies, results, with theoretical and experimental foundations.

see u again
Ins Network Advisor