This document discusses new approaches to search beyond just keyword searches. It proposes searching by emotion to find information that makes users feel positive emotions. It also discusses creating personal agents that conduct goal-oriented searches on topics of interest to the user and provide regular reports. The document outlines the technical frameworks needed for these new search approaches, including information retrieval, text mining, natural language processing and machine learning.
3. Search by Keyword only?
• Page search, image search, news search,
blog search, feed search etc.
• Other choices except “keyword”?
• The goal of searches is that find data you
want.
• They can’t find information you feel
positive, e.g. happy, kind, novel, optimistic.
4. Case study
• Financial recession
– Unhappy, suffering, pessimistic and
negative thinking.
• Surfing web
– Not only searching!
– A lot of netizenes surf web aimlessly
every time.
5. Search by Emotion
• Find information make positive emotions
for you, like happy, kind, novel, optimistic.
• Subjective vs. Objective
• Cross-lingual and Cross-region
• Happy, Novel, Kind, Excite
• Psychology of Emotion
– 發現恐懼和快樂都能傳染 . ( 英國醫學期刊 )
– 情緒傳染 , 情緒轉化 .
6. Strategy
• Sample the a lot of news or articles make
user happy, optimistic, even think it
positively.
• Vote or rate them via the all sorts of user,
and then raise accuracy and identification.
• Extract a large number of positive “key
word” and “key phrase” factors through
analysis, then group factors into the
several emotion types finally.
8. Search history only?
• The accuracy of results restricted
by keywords you inquire.
• The results at once before
periodically
• Google Alerts
9. Case study
• Plan your lunar year vacation.
• Seek a job you care.
• Track the stocks or funds you invest.
• Watch the some news headline you
concerned in.
• Refer to others’ suggestions or
opinions.
10. Personal Agent
• The goal-oriented search
– Who, What, When, Where and Which.
• Search and watch topic you interested in.
• Show you reports regularly.
• Domain-specific issue/event.
• Immediate vs. Regular
• Raw data vs. Processed information
11. Strategy
• Each agent is scheduled to do searching,
extracting, voting or rating.
• Each mission is limited in:
– Domain-specific issue/event
– Who, What, When, Where and Which.
• Build several core comparator, e.g. price, salary...
• Extract the summary of articles through analysis,
then group summaries into the several sections
finally.
12. Framework
• Technical scope
– information retrieval & extract
– text mining
– natural language processing
– machine learning
– intelligent agent
13. Distributed computing
• Execute the distributed computing
supported by all users, like searching,
extracting, voting or rating.
• One eSobi, one computing agent.
• Web servers only act as the
coordinator and collector.
14. Conclusion
• eSobi in the world (Ant) vs. servers
of Google, Yahoo, Amazon (Elephant)
• Patent family
• Q&A