Contextual Social Search and Conversations

Document Sample
Contextual Social Search and Conversations Powered By Docstoc
					    Healthbuzz: Contextual Social Search and Conversations

                      Saurav Sahay                                     Stephanie Ahn                             Szu-Chia Lu
                         Georgia Tech                                      Georgia Tech                           Georgia Tech
                  Brian Sherwell                                  Anushree Venkatesh                           Ashwin Ram
                         Georgia Tech                                      Georgia Tech                           Georgia Tech

ABSTRACT                                                                            provide one-stop solution reducing this gap for health specific in-
                                                                                    formation access. Healthbuzz supports combined functionality of
Healthbuzz is a health specific social search and conversation sys-
                                                                                    web search, group collaboration, real time conversation and con-
tem being developed for Georgia Tech (GT) student community. It
                                                                                    textual recommendations. The problem we are addressing is that
allows users to search health issues or start a new health specific
                                                                                    of quickly getting answers to or finding relevant local health related
conversation and provides three types of contextual results: related
                                                                                    information for GT student community.
conversations to join, community people to talk to with similar pro-
files, and Web search results based on users’ search criteria or on-
going conversation. We found out that people connections were a
                                                                                                      Table 1: Traits of existing systems
crucial aspect of the Healthbuzz system that engaged the partici-
pants and provided greater user satisfaction. To our surprise, users
                                                                                                                  Web Search       Health Portal   IM
did not show an active interest in the web based recommendations                                                      √                 √
during a conversation, unless pointed to explicitly by one of the                             Search
                                                                                                                                        √          √
participants. We briefly present the system and discuss the findings                      Group Collaboration
of the usability study of the system.                                                   Network Conversation

Categories and Subject Descriptors
H.4 [Information Systems Applications]: Miscellaneous; H.5.3
[Group and Organization Interfaces]: Evaluation/methodology
                                                                                    2     System Description
Social Search, Conversational Recommendations, Usability Study

1     Introduction
Information seeking is mostly a solitary activity on the web today.
Socially enabled online information access is a new phenomenon
facilitated by recent Web technologies. This collaborative social
search involves finding together specific people or resources that
can help you with your task at hand. However, traditional search
engines, content portals, or health forums are not very useful for
personalized health information access for local informational need
in a community setting such as Georgia Tech student community.
Existing systems do not provide one stop solution tools and aids for
accessing contextually relevant resources, or communicating with
similar people.
   There is a wide gap between systems that effectively use so-
cial network for community support and communication; and sys-
tems that provide search and content aggregation for effective in-
formation access as depicted in Table 1. Healthbuzz is designed to

Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to                           Figure 1: System Prototype
republish, to post on servers or to redistribute to lists, requires prior specific      Healthbuzz is an adaptation of our system called Cobot (Com-
permission and/or a fee.
SSM 2010, New York                                                                  munity bot) The key functions of Healthbuzz
Copyright 200X ACM X-XXXXX-XX-X/XX/XX ...$10.00.                                    are connecting GT community and providing real time contextual
recommendations regarding health conversations. Also, unlike fo-         with other students. They are familiar to finding general informa-
rums where users need to post a question and wait for others’ re-        tion by using a search engine (e.g., Google) or a health portal (e.g.,
sponses, Healthbuzz enables users to have instant interactions by        WebMD). However, it remained a challenge to find specific infor-
providing a real time conversation by connecting people. It pro-         mation that exactly fitted their particular issues (eg. How many
vides combined experiences of group collaboration, social network        cases of H1N1 have been reported in the Northside dorms of grad-
connection, Web search, and real time conversations. Users can           uate housing?’). For specific health related issues they wanted to
search in natural language, join existing conversations or create        discuss, they found that Web forums are useful since they can post
new ones for their concerns.                                             their specific questions and interact with other people. We noted
                                                                         two major shortcomings with Web forums. Since the forum is usu-
3    Evaluation and Discussion                                           ally available for anyone, there are trust issues. Another problem
We recruited four subjects who were graduate students of Georgia         comes from its asynchronous characteristic. Users have to wait
Tech. We conducted surveys, interviews and a usability study of the      until other people post their responses to their questions. When
Healthbuzz system. We created two different scenario-based tasks         they have a follow up question or want to interact with the person,
and asked the subjects to perform the tasks on Healthbuzz and Web        they need to post further questions or comments and wait for the
Search engine. Since students were concerned about H1N1 and              responses again. It is difficult to expect even near real time inter-
related questions, we created our tasks accordingly.                     actions due to the asynchronous nature of the medium. Generic IM
   Scenario 1: You are a GT student living in the dorms. You be-         chat provides very limited friend list based on one’s personal social
lieve your roommate is showing some H1N1 symptoms including:             network. Also, it is difficult to find a person who would be knowl-
coughing, having a runny nose, and fever. How can your room-             edgeable about their health related issues among their IM friends
mate find out if he has been infected with H1N1? What services            list. Also, when it comes to discussing specific and personal health
does GT offer regarding H1N1 prevention? What can you do to              related problems (against generic information), one is not comfort-
prevent yourself from getting infected?                                  able discussing these issues with the people they know.
   Scenario 2: You are a GT student living in the dorms. You be-            What we learnt from the interviews is that people connections
lieve that you have become infected with H1N1. What are the treat-       were a crucial aspect of Healthbuzz system. Collaboratively seek-
ments offered by GT? What does GT advise you to do (including            ing information with similar users lead to more refined and satisfied
class schedules) if you become infected?                                 search experience than solitary Web search. We had thought that
   Besides interviews, we asked users to rate on 3 different param-      providing Web based results in a conversation would be a crucial
eters. The averaged results are given in Figure 2 for the 3 questions    aspect of the system. This came out as a surprise to us when users’
where 3 indicates Strongly Satisfied.                                     did not show an active interest in the web based recommendations
                                                                         alongside in the conversation. Unless pointed to explicitly by one
    • The level of satisfaction with time spending on finding the         of the participants, the users generally didn’t pay attention to what
      results you want. (Timeliness)                                     was showing up as the recommended articles. We also realized that
    • The level of satisfaction with the results you find. (Results)      we needed to restrict these recommendations to just a few really
                                                                         good recommendations with highlighted answers to their questions
    • How easy it is to use the system? (Ease of Use)                    when it was required (using Natural Language Technologies and
                                                                         AI techniques).
                                                                            Participants of the study enjoyed the use of chat based conversa-
                                                                         tions to gather information. It was noted that some users valued the
                                                                         ability to create a group chat by inviting others to the conversation.
                                                                         Due to the nature of the information being exchanged within the
                                                                         Healthbuzz system, privacy was a major concern expressed by all
                                                                         participants. Participants stressed the need for anonymous commu-
                                                                         nication when discussing health related issues. Some users of the
                                                                         system were concerned about the amount of time it would take to
                                                                         find the health information they were looking for. These concerns
                                                                         centered around two topics: the interaction method of chatting and
                                                                         the response time needed to answer questions. Concern was noted
                                                                         by a participant that the method of interacting through chat would
                     Figure 2: Evaluation Results                        take more time than doing a web search. In her words, “Chatting is
                                                                         for people who have time to spend."
   First, we conducted a ‘within subject’ experiment for our user
                                                                            We learnt that the socially powered search feature and the ability
testing. Our participants were asked to perform scenario-based
                                                                         to collaboratively search together and solve issues with real people
tasks using Healthbuzz and web search. The usability of the Health-
                                                                         instead of search engines is a very powerful medium and highly
buzz was evaluated through quantitative methods by using a 7-point
                                                                         personalized. Websites like are doing this effectively us-
Likert scale survey. This survey was used to ask the participants’
                                                                         ing IM based messaging bots. People spent more time on Health-
level of satisfaction with the time they spent, the results they got,
                                                                         buzz system as compared to solitary web searches, exploring more
and the easiness of performing the tasks. The primary intent of this
                                                                         questions and arguments, engaging naturally into the conversation
usability testing was to determine if our participants could perform
                                                                         accepting and sending conversation invites by/to others. We also
the assigned tasks and their subjective level of satisfaction with us-
                                                                         learnt that a Healthbuzz like system has many other applications
ing the systems.
                                                                         revolving around sharing experiences, engaging the local commu-
   Based on interviews with students, we found that web search-
                                                                         nity into a dialogue about different issues, events and opinions.
ing is the most frequently used method when looking for informa-
tion related to health issues. However, students addressed that it
is difficult to find a place for discussing local health related issues