Pavel Braslavsky - Questions Online: What, Where, and Why Should we Care?

Questions Online: What, Where, and

Why Should we Care?

Pavel Braslavski

09/04/2015

2

QUESTION ANSWERING

3

Question Answering

• A long tradition of research

• Text document collection + natural language questions concise answer

• TREC QA 1999-2007 – text search, some parsing and textual entailment

• Mostly factoids – How much folic acid should an expectant mother

get daily?

– What is the population of the Bahamas?

4

QUESTION IN WEB SEARCH

5

Question Queries (QQ)

• About 2-3% of all queries

• The share is growing (!)

• Somewhat longer

• More unique queries

• Worse results compared to keyword queries (at least until recently)

• how is the leader (i.e. not factoids)

6

7

Question types

как сделать что такое как правильно можно ли сколько стоит о чем что делать как узнать как установить где купить from 1B Yandex QQ

8

Why more questions?

• Natural UX

• Low search proficiency

• Users switch to questions when fail to formulate a good keyword query

• Question + answer online (more later)

• Voice search (!)

9

10

11

12

13

Modern Search

• Information need as questions

– More context, relations

• From terms to concepts, from text search to inference

– Google Knowledge Graph, Facebook Graph Search (i.e. more knowledge-rich approach)

• Concise answer

14

QUESTION IN SOCIAL MEDIA

15

16

Community QA

17

CQA-related tasks

• Quality estimation

• Questions & answers recycling

• Expert finding

• Etc.

18

Real World vs. CQA data

0

0,02

0,04

0,06

0,08

200

400

600

800

1000

1200

ARI rate vs. 'flu' topic

0

0,1

0,2

0,3

0,4

0,5

-2

-1

0

1

2

3

4

5

rain vs. 'runny nose' topic

20

Asking Friends

Questions on Twitter & FB

• ~3% of tweets convey an information need

• Correlate with search queries

• Recommendations, opinions

21

Facebook

References

• Children ask Yandex, https://company.yandex.ru/researches/figures/2013/ya_kids_questions.xml

• Google 2014 Trends: top questions, https://www.google.com/trends/2014/story/top-questions.html

• Ch. Manning, P. Nayak. Web Question Answering (lecture slides), http://web.stanford.edu/class/cs276/handouts/lecture19-Web-QA.pptx

• D. Ferrucci et al. Building Watson: An Overview of the DeepQA Project. AI Magazine, 31(3), 2010.

• A. Shtok et al. Learning from the Past: Answering New Questions with Past Answers. WWW’2012.

• B. Pang and K. Kumar. Search in the Lost Sense of “Query”: Question Formulation in Web Search Queries and its Temporal Changes. HLT '2011.

• Z. Zhao and Q. Mei. Questions about questions: An empirical analysis of information needs on Twitter. WWW 2013.

• Q. Liu et al. When web search fails, searchers become askers: Understanding the transition. SIGIR’2012.

• M. Ringel Morris, J. Teevan, and K. Panovich. What do people ask their social networks, and why?: A survey study of status message Q&A behavior. CHI’2010.

22

• Put RuSSIR pic here

• Annual event

• 100+ participants

• 4th RuSSIR: Voronezh 13-18 September

• http://romip.ru/russir2010/

23

RuSSIR 2015 Saint-Petersburg, 24--28 August 2015 Application deadline: 20 April, 2015 http://russir.org

24

Pavel Braslavski

[email protected]