Bba q&a study final white

EXECUTIVE SUMMARY

A new study using the Butler University Q&A Intelligence Index measures how various mobile
Q&A platforms deliver quality, accurate answers in a timely manner to a broad variety of
questions. Based on the results of our analysis, ChaCha led all Q&A platforms on mobile
devices.

Results of the study are based upon review of a large set of responses from each of the major
Q&A platforms, coupled with a comparison of disparate Q&A platforms that serve answers in
different ways. Our methodology included the creation of a new metric, termed the Butler
University Q&A Intelligence Index, which measures the likelihood that a user can expect to
receive a correct answer in a timely manner to any random question asked using natural
language. We asked questions via mobile services and randomized the questions to cover both
popular and long-tail knowledge requests.

*Mean accuracy of responses originally graded on a 5 point scale.

Here are a few details about some of the platforms tested:

ChaCha delivered the highest quality responses consistently across the largest group of
categories and question types, but did have occasional issues with Objective|Temporal and
Sports questions. Ask.com performed better than ChaCha in the single category of questions
tagged as Objective|Temporal.

Quora was proficient at answering difficult questions that require expert and extensive
explanations, but it was generally unable to deliver answers within 3 minutes for most
information searches on mobile devices. Quora answered only 24% of the questions at all, and
often the match found did not include a viable answer.

Siri did not perform nearly as well on this random sampling of popular and long-tail questions as
it did on a recent Piper Jaffray study, where results indicated that Siri correctly answered 77% of
questions (Elmer-DeWitt, 2012). Our study found Siri only accurately answered 16% of the
questions posed. The variance may be due to the types of questions asked and the testing
conditions. Piper Jaffr

mainstream questions.

-sponsored result on the search results page
(which often times was not fully visible as an organic search result on the presented page on a
mobile device) only presented an accurate answer about 50% of the time, according to the
Butler University Q&A Intelligence Index. On a mobile phone, when accounting for the clutter of
ads and the likelihood of extra clicks to achieve the answer, allowing for the answer to be within
the first non-sponsored search result might be considered generous. Again, this study differs
from the results found in the Piper Jaffray study, but differences are likely due to variations in
methodology. For example Piper Jaffray found that Google scores highest in terms of navigation
and information (Elmer-DeWitt, 2012).

hat Q&A platforms cannot rely on algorithmic search
results alone to deliver quality answers. Search Engine Results Pages (SERP) lack deep

effectively for long-tail questions like those posed in this study (De Virgilio, Guerra, &
Velegrakis, 2012).
To achieve a score above 50% or 60% on the Butler University Q&A Intelligence Index, it would
appear that Q&A platforms must supplement algorithmic document indexing with either:

Utilization of structured data
Semantic understanding via artificial intelligence (AI) or real humans

In terms of handling structured data more effectively, Google is promoting direct answers using
its new Knowledge Gra

Page 1 of 18
Property of Butler University, All Rights Reserved

an
algorithmic solutions.

Improved machine learning may eventually push past the algorithmic limitations of document
n Jeopardy) proved that intensive semantic
processing can help, albeit without cost-

Watson showed potential. The DeepQA project claimed 90% accuracy in answering 60% of
questions and 70% accuracy in answering 100% of questions (Ferruci, 2010). These results
translate to 54 and 70 Butler University Q&A Intelligence Index scores respectively.

We conclude that, without large advances in semantic processing and better utilization of
knowledge graphs, Q&A platforms can benefit from the timely injection of human semantic

Intelligence Index is an indication that just-in-time human-assisted Q&A can outperform
algorithm-only solutions.

Page 2 of 18

INTRODUCTION

-driven culture are turning to the internet and smart
phones to find the answers to their questions. Given this fact, the availability of accurate
answers to these questions by Q&A platforms1 is of growing importance. ChaCha Search, Inc
(ChaCha) commissioned the Butler Business Accelerator (BBA) to conduct an independent
assessment with quali
competitive position with respect to Q&A platforms in the industry, as it relates to quality. We
conducted a focused independent study examining comparative data across multiple Q&A
platforms in the marketplace. This white paper outlines the methodology, results and
conclusions of the study.

PROBLEM AND METHODOLOGY

We gathered our data in August of 2012 to assess the mobile experience when using Q&A
platforms. Our study evaluated mobile applica
app was not available, and in that case our analysts used a mobile website.

platforms. We selected these platforms because of their popularity and ability to provide
answers to a variety of posed questions. Appendix A describes all of the Q&A platforms
included in this study.

To assess the quality of the responses of these mobile Q&A platforms, we conducted mobile
research, which included the following activities:

a) Posing a sample set of questions to each of the Q&A platforms using a mobile
application when one was available and a mobile website when no application
existed

b) Recording all responses

c) Rating the responses from each Q&A platform for coverage and accuracy

d) Analyzing data and tabulating it in a summarized format

Attributes of Quality:

For the purposes of this study, in order to evaluate the quality of answers in terms of coverage
and accuracy for each Q&A platform, we defined the terms as follows:

Coverage: A binary decision based on whether a Q&A platform returned an answer
within 3 minutes. For search engines, we only reviewed the first non-sponsored2 result

1,2
See definitions in Appendix C

Page 3 of 18

provided because, on a mobile phone, the visual bandwidth is highly constrained,
especially when there are advertisements and navigational elements competing for

be easy to access without extra navigation.

Accuracy: The objectively determined correctness of answers, in relation to an answer
key, which an independent third party developed. The answers were then scored blindly
by Butler Business Accelerator analysts.

Measurement:

We developed a mechanism

variable.

Coverage: We determined coverage based on whether or not the Q&A platform gave an
answer for each question asked.
1 No, answer not provided. If the question was not answered within 3

2 Yes, answer provided within 3 minutes. The analyst logged the results as

Accuracy: We measured accuracy using an answer key, which a contracted independent
third party created. Then, two analysts used a blind answer key, meaning that the
scorers did not know which Q&A platform gave a particular answer. The analysts used a
reasonableness test to determine accuracy by rating the answer using the interval scale
below:

1 Incorrect answer
2 More incorrect than correct
3 Neither correct or incorrect
4 More correct than incorrect
5 Correct answer

When the answer to a question was subjective, we measured the answers using the
following scale:

1 Very illogical, inappropriate or one-sided
2 More illogical, inappropriate or one-sided (rather than logical, appropriate,
and impartial)
3 Neither illogical, inappropriate or one-sided or logical, appropriate, or
impartial

Page 4 of 18

4 More logical, appropriate, and impartial (rather than illogical, inappropriate
or one-sided)
5 Very logical, appropriate and impartial

For each answer, we had a total of four accuracy scores (researcher 1/scorer 1, researcher
1/scorer 2, researcher 2/scorer 1/, researcher 2/scorer 2). We calculated the average of each of
the four scores, and then used this data set for the analysis.

Question Selection:

Given the scope of this project, we tested 180 total questions during the Study. This number
was large enough to be significant and compelling, yet small enough to keep the effort
reasonable in scale.

Our study examined a random sample of both popular and long-tail3 questions from mobile
users. The long-tail nature of questions is described by David Ferrucci in his IBM DeepQA

frequent occurring <3% of the time. The distribution has a (Ferruci, 2010).

The question corpus in this Study began with 20,000 randomly chosen questions asked by third-
-day period. From
3,000 randomly selected questions, filters removed those with sexual references or profanity
and those that were confusing or incomplete. From the remaining questions, BBA again
randomly sampled to arrive at 180 non-duplicative questions. For more details about the
methodology BBA utilized to provide questions for this Study, see Appendix B.

Question Type and Category Assignments:

Once our project team selected 180 questions from the original corpus for testing on the various
services, ChaCha then aided in classifying those 180 questions in two ways: 1) a category-
based ontology that ChaCha uses as a base taxonomy and 2) a question-type matrix that
looked at the type of answer required of the question and some attributes of the question.

Evaluated Questions:

Before our team completed its analyses, an analyst labeled each question in the study by its
type and category in order to provide more detailed analysis. Because the Objective|Static type
consisted of 63 more questions than the next highest type or category, we randomly selected 30
questions from this type to be evaluated in order to make the sample size more comparable.
Using a smaller sample also helped to assure that we did not reach statistical significance
simply because of a large sample size. We chose the questions to be evaluated for this type
using the Excel random number generator in-between function. This assured that we selected

3
See definition in Appendix C

Page 5 of 18

the questions for evaluation completely at random. Because the function repeats numbers, we
used the first 30 different randomly generated numbers for evaluation in the study.

Answer Selection:

Many of the Q&A platforms gave one direct and definitive answer when posed with a question.
When one direct and definitive answer was given, we evaluated the accuracy of this answer.

However, if one direct and definitive answer was not given, we scanned any information that
was viewable in the first non-sponsored search result. If the answer was within a link, we only
used the description of the link on the search page. Our researchers did not follow links in
search of an answer.

If SIRI, IRIS or SpeakToIt automatically searched the web in response to being asked a
question, we evaluated the first answer of the web search. However, if the assistant offered to
perform a web search but did not search automatically, we did not evaluate the result of the web
search, but we noted this response.

Testing Conditions:

Our project team used the Apple iPhone 4 as the primary device for this study. We used
this device because of its capability to search using both mobile web browsers and
applications and its availability to our analysts conducting the research. The only time
the iPhone 4 was not used was when needed software was unavailable on this device.
When testing SIRI, we used the iPhone 4s; when testing IRIS, Google and SpeakToIt,
we used an Android phone.
We posed all questions from the same two iPhone 4 devices (with the exception of SIRI,
IRIS and SpeakToIt testing) from the same location at Butler University in Indianapolis,
Indiana, to ensure that internet signal strength was constant at all times.
The same two researchers conducted all portions of the study to ensure that notation of
answers was consistent.
One question was posed to all Q&A platforms before our researcher moved on to asking
the next question, so that the same question was answered by each platform within the
same time frame.
Our researchers posed all questions using the same wording. We typed all questions
with the exceptions of the questions asked to SIRI, IRIS and SpeakToIt. Because SIRI,
IRIS and SpeakToIt respond only to voice commands, we spoke the questions clearly
using the same words that we typed to the other Q&A platforms.

FINDINGS

In order to evaluate the results of this study, our project team created the Butler University Q&A
Intelligence Index using the scores for both Coverage and Accuracy. The Butler University Q&A

Page 6 of 18

Intelligence Index is a metric that represents the probability that a completely random search,
using natural language, will return an accurate answer within 3 minutes and in a useful format.

Summary results in terms of the two metrics evaluated and the Butler University Q&A
Intelligence Index are shown in Table 1.

Table 1: Summary Results

*Mean accuracy of responses originally graded on a 5 point scale.

Coverage:

To measure coverage, our project team calculated the percentages listed in Table 1 based on
the number of answers provided divided by the total number of questions.

From visual inspection of the data, the team concluded that there are not significant differences
in coverage between ChaCha, Ask, Bing, Yahoo, Google and IRIS. The results also show that

Page 7 of 18

these Q&A platforms are superior to the other competitors listed with regards to coverage based
on the percentage of questions answered.

In this case, we believe visual inspection is adequate, but difference of proportions testing to
deliver inferential statistics could also be warranted.

Accuracy:

ChaCha ranks highest in terms of Accuracy as compared to its competition (see statistics in
Table 2). On a 5 point scale, ChaCha was .40 points better than Answers & .466 points better
than Ask, which is very strong given the sample size.

Table 2: Mean Accuracy Summary Analysis for All Questions

All Questions

One factor ANOVA

Mean n Std. Dev
3.194 168 1.2349 Ask
2.754 168 1.2642 Bing
3.660 167 1.0974 ChaCha Analysis shows that in terms of accuracy
2.506 168 1.1015 Google ChaCha is statistically superior to its major
2.771 40 1.2587 Quora competitors. It is managerially superior to all
2.723 168 1.2709 Yahoo competitors except Ask and Answers.
3.260 119 1.3934 Answers
2.644 165 1.5336 IRIS
1.795 146 1.2369 Wolfram Alpha
2.189 132 1.2782 SpeakToIt
1.874 73 1.3399 SIRI
2.716 1514 1.3794 Total

ANOVA table
Source SS df MS F p-value
Treatment 443.3637 10 44.33637 27.36 2.05E-48
Error 2,435.6092 1503 1.62050
Total 2,878.9729 1513

Post hoc analysis

p-values for pairwise t-tests
Wolfram Alpha SIRI SpeakToIt Google IRIS Yahoo Bing Quora Ask Answers ChaCha
1.795 1.874 2.189 2.506 2.644 2.723 2.754 2.771 3.194 3.260 3.660
Wolfram Alpha 1.795
SIRI 1.874 .6615
SpeakToIt 2.189 .0099 .0900
Google 2.506 8.71E-07 .0004 .0327
IRIS 2.644 5.27E-09 1.82E-05 .0023 .3228
Yahoo 2.723 1.53E-10 2.16E-06 .0003 .1180 .5700
Bing 2.754 3.79E-11 9.19E-07 .0001 .0744 .4305 .8248
Quora 2.771 1.84E-05 .0004 .0115 .2371 .5717 .8317 .9400
Ask 3.194 1.10E-21 2.35E-13 1.67E-11 8.12E-07 .0001 .0007 .0016 .0591
Answers 3.260 4.00E-20 4.02E-13 4.04E-11 8.58E-07 .0001 .0004 .0009 .0357 .6660
ChaCha 3.660 2.30E-36 7.89E-23 1.64E-22 2.33E-16 5.64E-13 2.31E-11 9.91E-11 .0001 .0008 .0088

Our team conducted additional in-depth analyses on the accuracy metric to better understand
how ChaCha performs against the competition across different question classifications and

performance was either statistically or managerially significant compared to its competition.

For Table 3, we only recognized managerial significance when a Q&A platform had first
obtained statistical significance. Using this approach, the only time that there was managerial
significance was when the results first cleared the statistical significance screen. This
conservative approach does not take into account the idea that managerially meaningful
differences might exist, but are sometimes masked by either small sample size or high variance.

Page 8 of 18

Table 3: Summary of C
Platforms
All Type
Advice| Objective| Objective| Subjective|
Advice|Static
Personalized Static Temporal Static
Source Stat. Man. Stat. Man. Stat. Man. Stat. Man. Stat. Man. Stat. Man.
Answers x x x
Ask x x x x x
Quora* x x x x
Bing x x x x x x
Yahoo x x x x x x x x x x
IRIS x x x x x x
Google x x x x x x x x x x
SpeakToIt x x x x x x x x x x
SIRI x x x x x x x x x x
Wolfram Alpha x x x x x x x x x x

Category
Entertainment Language & Society &
Health Sci Tech Sports
& Arts Lookup Culture
Source Stat. Man. Stat. Man. Stat. Man. Stat. Man. Stat. Man. Stat. Man.
Answers
Ask
Quora* x x
Bing x x x x x x
Yahoo x x x x x x
IRIS x x x x x x x x x x
Google x x x x x x x x
SpeakToIt x x x x x x x x x x
SIRI x x x x x x x x x x
Wolfram Alpha x x x x x x x x x x

Sample size varied based on each category and type as w ell as the number of questions answ ered by each provider. *Quora data derived from a very small sample
size per lack of coverage by the provider.

CONCLUSIONS

Based on the results of our analysis of Q&A platform performance related to quality results,
ChaCha led all competitors in delivering accurate answers to varying types of questions asked
on mobile platforms.

This conclusion is derived from an in-depth study that included the review and evaluation of a
large sample set of actual responses, collected by the Butler Business Accelerator through the
use of a defined method for comparing disparate Q&A platforms that offer answers in different
ways. Our methodology included the creation and use of a unique metric we have termed the
Butler University Q&A Intelligence Index, which multiplies raw accuracy scores by coverage
rates to get at the percentage of times end users received a correct answer to random

Page 9 of 18

questions asked using natural language. The questions we tested had been previously asked
via mobile services and randomized to cover both popular and long-tail knowledge requests.

Below are a few details about some of the Q&A platforms tested:

ChaCha delivered the highest quality responses consistently across the majority of categories
and question types, but did not have the top position with respect to Objective|Temporal and
Sports related questions.

Quora was strong in answering difficult questions that required experts and extensive
explanations, but it was unable to deliver answers within 3 minutes for information search on a
mobile device. Only 24% of the questions were answered, and many times the match found did
not include a viable answer.

e Intelligence Index, but scored more than
eight percentage points lower than ChaCha. Ask.com did score higher than ChaCha on
questions from the Objective|Temporal category.

Our study found that Siri did not perform well on this random sampling of popular and long-tail
questions. In clear contrast to a recent Piper Jaffray study, where results indicated that Siri
correctly answered 77% of questions (Elmer-DeWitt, 2012), we found Siri only able to
accurately answer 16% of questions. The variance in results of the studies is mainly due to the

randomized, empirical question sample set for this BBA study.

Google provided a response 100% of the time, but the first non-sponsored result on the search
results page (NO COMMA) displayed an accurate answer only about 50% of the time, according
to the Butler University Q&A Intelligence Index. Again, this study differs from the results found in
the Piper Jaffray study, but differences are likely due to variations in methodology. For example
Piper Jaffray found that Google scores highest in terms of navigation and information (Elmer-
DeWitt, 2012), which may not have been tested at comparable levels in the two studies.

The results of our research support a hypothesis proposed by ChaCha that Q&A platforms
cannot rely on algorithmic search results alone to deliver quality answers to mobile users.

human questions and, therefore, cannot account effectively for long-tail questions like those
posed in this study (De Virgilio, Guerra, & Velegrakis, 2012).

To obtain a score above the 50-60% mark on the Butler University Q&A Intelligence Index, it
would appear that Q&A platforms must supplement algorithmic document indexing either with
utilization of structured data or with semantic understanding through artificial intelligence or real
humans.
In terms of handling structured data more effectively, Google is providing direct answers using

(Google, 2012).

Page 10 of 18

algorithmic solutions.

Improved machine learning will eventually push past the algorithmic limitations of document

exper
accuracy in answering 60% of questions and 70% accuracy in answering 100% questions
(Ferruci, 2010). These results translate to 54 and 70 scores for the Butler University Q&A
Intelligence Index respectively.

We conclude that, without large advances in semantic processing and better utilization of
knowledge graphs, Q&A platforms can benefit from the timely injection of human semantic
understanding into the Q&A experience. C
Intelligence Index is an indication that just-in-time human-assisted Q&A can outperform
algorithm-only solutions.

Page 11 of 18

LIMITATIONS

This study was a limited study in that our analysts conducted research on primarily the iPhone4
(device) and not the newest version of the iPhone. However, our two analysts asked a total of
3,960 questions, which produced numerous data points for use in evaluating the performance of
all key competitors in the Q&A platform space.

It is important to acknowledge that other mobile Q&A platforms are now available on devices,
and newer versions of specific applications are now available. For example, a new version of
SIRI is now available on the iPhone5. While these new solutions were beyond the scope of this
study, the evaluation of new versions of all platforms in a subsequent study may be warranted.

The following activities were out of scope for this study but may be considered for additional
potential research studies:

1. Hands-on development of, deployment of, and direct response compilation of the SMS
text-based and app-based survey technology.

2. Surveys to the mobile or web-based community at large to understand and assess end
user perceptions and experiences with Q&A platforms.

3. Repeat this study with newer or latest versions of technology and applications for key
Q&A platforms in the industry.

Page 12 of 18

REFERENCES

De Virgilio, R., Guerra, F., & Velegrakis, Y. (2012). Semantic Search Over the Web. Springer.

Elmer-DeWitt, P. (2012, December 20). In Piper Jaffray re-test, Siri raises her grade from a D to
a C. Retrieved January 2, 2013, from CNN Money:
http://tech.fortune.cnn.com/2012/12/20/google-now-vs-apple-siri/

Ferruci, D. (2010). Building Watson: DeepQA@ IBM Research. Retrieved December 20, 2012,
from Whitehouse: http://www.whitehouse.gov/sites/default/files/ibm_watson.pdf.

Google. (2012, May 16). Introducing the Knowledge Graph: things, not stringsl. Retrieved
January 2, 2013, from Google Blog: http://googleblog.blogspot.com/2012/05/introducing-
knowledge-graph-things-not.html#!/2012/05/introducing-knowledge-graph-things-
not.html

Page 13 of 18

About the Butler Business Accelerator

The Butler Business Accelerator, established in 2006, is an extension of the Butler University
College of Business. The BBA employs full time leaders and consultants steeped in consulting
experience across a vast range of industries. The BBA also utilizes both graduate and
undergraduate student interns to assist in research, investigation and deliverable development.
Finally, the BBA capitalizes on the breadth and depth of knowledge and experience within the
faculty of Butler to provide subject matter expertise where needed and warranted.

The BBA and its staff of professionals have worked with other service-oriented companies to
help bring clarity to their service offerings, bring new offerings to market, develop strategies and
tactics to better market their services to customers and educate internal stakeholders on the
value proposition of their service offerings. This result has been accomplished in companies in
various industries, including both traditional and web-based services organizations.

The project team for this study consisted of Ginger Lippert, Butler Business Accelerator project
manager, and Dr. Richard Fetter, Butler University associate marketing professor, as well as
Butler Business Accelerator marketing student analysts.

The student analysts performed the research testing under the oversight of Ginger Lippert. Two
-rater
e accuracy
portion of the study, so that the scoring for all questions could be compared.

Ginger Lippert and the student analysts conducted the analysis portion of the study with the
guidance and review of Dr. Fetter.

Page 14 of 18

Appendix A: Q&A Platforms

Q&A Platform Description Type Tested using
4
Answers.com A question and answer website that gives Crowdsourcing Mobile web
direct answers to posed questions.
Researchers used the mobile web version of
Answers.com because no mobile application is
available.
Ask.com A question and answer website and app that Crowdsourcing Application
provide direct answers to posed questions.
Bing.com A search engine that provides direct answers Algorithm-only5 Application
to some of the most commonly asked
questions and also provides a multitude of links
to other websites.
ChaCha A question and answer website and app that Human-assisted6 Application
provide direct answers to posed questions.
Google.com A search engine that provides direct answers Algorithm-only Application
to some of the most commonly asked
questions and also provides a multitude of links
to other websites. Researchers used the voice
recognition software of the Google Search
application in order to test the latest and most
advanced Google technology.
IRIS A personal assistant app for Android that Virtual Assistant 7 Application
responds to voice commands and answers
questions.
SIRI Virtual Assistant Built-in
commands and answers posed questions on
select Apple devices.
SpeakToIt Virtual buddy app for smart phones that Virtual Assistant Application
answers questions in natural language,
performs tasks, and notifies the user of
important upcoming events.
Quora.com Website and app that deliver answers and Crowdsourcing Application
content from people who share similar
interests.
Yahoo A search engine that gives direct answers to Algorithm-only Application
Answers some of the most commonly asked questions
and also provides a multitude of links to other
websites.
Wolfram|Alpha Website and app that deliver answers to Structured Data8 Application
questions based on a collection of built-in data,
algorithms and methods.

4, 5, 6, 7,8
See definitions in Appendix C

Page 15 of 18

Appendix B: Methodology for Question Selection

Our BBA project team used the following methodology to select questions for this study:

The corpus started with two sets of data, randomly sampled from incoming questions over a 30-
day period using Hadoop queries9
Dexetra, Inc. The second set was from incoming questions via the iPhone mobile app called

Initially, at the direction of our BBA project team, ChaCha sampled 10,000 questions from each
source for a total of 20,000. From this set, ChaCha, with our direction, then discarded questions
containing sexual references and profanity using a Regular Expressions-based filter. Then
ChaCha, at our direction, re-randomized the data to select only 3,000 questions from the 20,000
corpus using an Excel randomization function. These 3,000 questions were then manually
reviewed by two analysts who culled from the corpus only questions that 1) were not complete,
understandable questions, 2) contained significant grammatical errors, 3) were duplicative or
extremely similar to questions already in the resultant corpus.

From the resultant corpus, our BBA project team then randomized the questions again and
selected the first 1000 items.

We then selected 180 questions from the corpus of 1000 items using the Excel random number
generator in-between function. This technique assured random selection. We selected the first
180 non-duplicated questions using this method.

9
See definition in Appendix C

Page 16 of 18

Appendix C: Definitions

1. Q&A platform: A provider of answers to natural language queries.

2. Non-sponsored: A result not being displayed as a form of advertising.

3. Long-tail: Not focusing only on popular questions but rather the full universe of possible
questions. For example, on one graph showing questions asked, roughly 3% were
popular questions showing high volume while roughly 97% were low-volume questions
that together resembled a long tail.

4. Crowdsourcing: A type of Q&A platform that primarily solicits answers from a large group
of users. In most cases, the question and answer are asynchronous and the transactions
are completed on the web.

5. Algorithm-only: A type of Q&A platform that primarily uses a set of computer instructions
to search for specific data.

6. Human-assisted: A type of Q&A platform that primarily relies on human intervention to
supplement algorithmic results.

7. Virtual Assistant: A type of Q&A platform primarily designed for voice use on a
smartphone device. The user interface is generally a virtual personal helper, that
attempts to understand user intent across a variety of functions including Q&A.

8. Structured Data: A type of Q&A platform that primarily delivers computational results
from large sets of structured sets, generally based upon an entity attribute value model.

9. Hadoop query - A command sent to a data system that allows the storage and
manipulation of very large sets of data to, in this case, randomly sample a group of
questions from a larger corpus.

Page 17 of 18

Bba q&a study final white

More Related Content

What's hot (20)

Similar to Bba q&a study final white (20)

More from Greg Sterling (20)

Bba q&a study final white