Skip to main content



How Does the Network Structure Affect Q&A Activities in Quora

Link: https://pdfs.semanticscholar.org/3019/421d0b208aff9439c4adf9ee282ea14e483f.pdf

A basic function of the Internet is to provide us answers. We’ve been accustomed to searching on the Internet for what we want to know. But a problem seems to arise that, with the explosion of data, users get lost easily with the overwhelming information online. Thus, Q&A sites are created to help. There’re actually two types of Q&A sites. One, Yahoo Answers for example, aims to offer first-hand answers from domain experts, whom you might not know. But this kind of sites is declining. It has the same problem of the Internet — low-quality answers and redundant information. Another type, however, is growing rapidly. Quora is a typical instance of this kind. What differs this kind from the original one is that it not only simply maintains a Q&A system, but combines it with a user community, thus adding socialization to it. So, the linked thesis talked about how the internal structures, the different networks, helped Quora to provide valuable answers. It’s quite interesting to learn how heterogeneous nodes interact with each other by discussing the Quora networks. What’s more, we can apply what we’ve learned in class to the networks and see how the properties account for the phenomenon.

As we indicated before, Quora combines a typical Q&A site with a social community. Therefore, in order to understand how the network structures work, we have to figure out how the network look s like first. The following picture clearly explains how different types of nodes are connected in Quora.

Here we briefly explain what does the graph mean:

  • The questions are classified into different topics. Each question has at least one topic.
  • A user can follow the topics he likes, thus making a connection between the user node and the topic node.
  • A user can follow the user he is interest with, thus making a connection between the two users. Notice that as the user community is a “Twitter-like” one, which means a user could follow another user without his permission, the edge between two users is one-directed.
  • A user can view a question, answer a question or ask a question, thus making a connection between the user and the question.
  • Questions are connected based on the similarity. If Quora thinks two questions are similar, then they are connected.

As there are three types of nodes, the article analyzed the mechanism from different types of interactions — the user-topic graph, the social (user-user) graph and the related questions (question-question) graph. Based on such methodologies, the thesis reached some interesting conclusions as below.

  • Questions under well-followed topics generally draw more answers and views.
  • Users who contributed more and good answers tend to have more followers.
  • Users with more followers tend to get more votes from their followers.
  • High degree questions generally receive more answers and views compared to others.
  • The questions “super-users” interact with receive additional views and answers from their followers.

All these conclusions seem pretty straight-forward and intuitive. But the significance lies in revealing the fact that users are guided to focus on a really small part of the whole network, thus ignoring the trivial and useless answers. The “super-users” (those who have a lot of followers) and topics both play significant roles in this process.

What is more interesting is to see how Strong Triadic Closure Property works in this combinational graph. For example, in the Quora networks, the property can account for why super-users’ answers receive more upvotes. As super-users are usually those who offer valuable answers and have many followers, the ties between them and their followers are quite strong. Because followers think they are helpful and interesting. Besides, as they tend to give useful answers, the ties between super-users and questions are strong as well. So according to STCP, followers tend to have connections to the questions which super-users answer — view the questions and upvote them. Another example is the relation between questions and topics. It is observed that questions under well-followed topics generally draw more answers and views. As there’re ties between topics and users, topics and questions, it’s quite possible to have a tie between the user and the question. We can see that the property is actually an “adhesive” — it makes nodes cluster together, thus making users focus on valuable questions.

Though we’ve seen how the network structures influence the Q&A activities and the applications of Strong Triadic Closure Property, some problems still remain unsolved. One very important point is that how we measure a tie is strong or weak. We can say offering valuable answers makes the tie between user and question strong, or we can say just viewing a question is a weak relation. But we lack logical derivations and direct evidence. With more specific and deep-going works, we can have more meaningful results of the Quora network.

 

Comments

Leave a Reply

Blogging Calendar

September 2017
M T W T F S S
 123
45678910
11121314151617
18192021222324
252627282930  

Archives