COLLECTING KNOWLEDGE AND LEARNING LANGUAGES

WITH TOWERS OF KNOWLEDGE GAME

Dilyana Valkova Budakova, Mariyana Ivanova Ilieva

Technical University of Sofia, Branch Plovdiv, Sankt Peterburg Blvd. N: 61, Plovdiv, Bulgaria

Lyudmil Georgiev Dakovski

CLBME BAS, Acad. G. Bonchev Str., bl.105, 1113 Sofia, Bulgaria

Keywords: Playing game, Affective databases, Accumulating and evaluation knowledge.

Abstract: The article treats the problem of deriving and accumulating knowledge and data about people’s everyday

real-world knowledge as a background for the process of learning languages through games. The

development of a game and the introduction of a virtual agent–assistant into the game are supposed to

increase the interest among the users, to stimulate their motivation for language practice, and, at the same

time, to increase the quantity and the quality of accumulated knowledge. The results from a survey,

conducted among users, have been analyzed and generalized in this paper.

1 INTRODUCTION

There are three big common real-world knowledge

bases: Cyc (Douglas B. Lenat 1995), Open Mind

Common Sense (OMCS) (Henry Lieberman, Dustin

A Smith, Alea Teeters, 2007), (Push Singh. 2001)

and Thought Treasure (Mueller, E.T. 1998). Cyc is

the biggest of the three, having more than 3 million

entries about the world, followed by OMCS,

containing nearly half a million of collected

sentences. Thought Treasure comprises about

100 000 concepts and relations. There are smaller

real-world knowledge bases, for example, in

Bulgarian language. The use of data from the

knowledge bases OMCS and Gethi (L.Dakovski, D.

Budakova, 2003) is convenient, as the collected

sentences there (OMCS in English and Geti – in

Bulgarian) are presented in the form of knowledge

and are, to some extent, easy for processing and

analyzing at morphological and syntactic level. The

common real-world knowledge is presented in

OMCS in the form of sentences in English, reduced

to about 20 models of various sentences, expressing

different relations between the concepts. In Gethi

simple sentence models and models at concept level

are supported.

There are web-based games that aim, on the one

hand, to derive knowledge, and, on the other hand,

to provide entertainment to their users. For example

Peek-aboom (Luis von Ahn, Ruoran Liu, and

Manuel Blum.) – a game designed for segmentation

of objects in pictures; Verbosity (Luis von Ahn,

Mihir Kedia, and Manuel Blum) – a game for

collecting commonsense real-world knowledge; and

Gathi (L.Dakovski, D. Budakova, O. Obretenov, V.

Georgiev) – Gathering Information – game for

Gathering Descriptions, Analogues and Associations

via Internet aimed at forming Rational Behavior.

The article treats the problem of deriving and

collecting knowledge and data about people’s

everyday real-world knowledge (objects, events,

analogies and associations) as a background for the

process of learning foreign languages through

games.

The aim, on the one hand, is to create a system,

attracting the users both because of being useful for

their development and because of being fun. On the

other hand, what is aimed as well is the capability of

the system to collect knowledge inconspicuously

(without the explicit awareness of the users). Thus

collecting data is not the users’ purpose but a result

(side effect) from their language learning activities

realized through the system.

In order to achieve this goal the models,

supported by the system are built simultaneously in

a number of working natural languages. It is

474

Valkova Budakova D., Ivanova Ilieva M. and Georgiev Dakovski L. (2010).

COLLECTING KNOWLEDGE AND LEARNING LANGUAGES WITH TOWERS OF KNOWLEDGE GAME.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Artiﬁcial Intelligence, pages 474-478

DOI: 10.5220/0002700304740478

 SciTePress

assumed that such a system would be of interest to

users, learning one of the working languages and

being familiar or speaking either one or two of the

other languages. While playing, the users of the

system learn and develop their linguistic skills, and

the system collects knowledge and data in the

background.

The introduction of a game together with the

participation of an IVA in the role of an emotional

assistant is expected to increase the users’ interest

towards the programming system, to facilitate the

process of language learning, to keep their attention

live for a longer time and to lead to accumulating a

bigger quantity and a better quality of knowledge. In

order to check these expectations, three versions of

the programming system are offered.

The first version suggests forms, containing a

question aimed at collecting a particular type of

knowledge and they need an answer.

The second option includes a game, which

stimulate either giving answers to questions or

executing the tasks from the first version.

The third option is the same game combined with

emotional assistance by the Microsoft intelligent

virtual agent Peedy.

The results from a survey, conducted among

users of the system, have been analyzed and

generalized further in this paper.

2 STRUCTURE OF THE SYSTEM

The Figure 1 contains 16 basic forms. More of them

allows collecting specific knowledge, e.g., collecting

associations with two notions, two activities or a

combination of a notion and an activity; collecting

descriptions of notions etc. Some of the forms allow

building a hierarchy of knowledge. For example, the

forms, related to forming questions and answering

questions of the type “What do you need to ….”

allow building a hierarchy of goals by means of

transforming each answer into a new goal and then

asking the same question for the new goal. The

forms, related to comparing notions and activities

are also addressed to building a hierarchy,

depending on the comparative criteria, chosen by the

user, such as “better”, “bigger”, “more useful”

“more difficult” etc. The forms, related to describing

notions allow building hierarchies of the type

“something is” and “something consists of”, i.e,

hierarchies of categories and sub-categories.

Other forms allow collecting assessments of both

notions and activities. Some of the assessments are

emotional, need-related or rational. Others represent

Figure 1: Structure of the programming system.

the valence of a notion (positive, negative or

neutral), its level of functioning (e.g. a rude or

insulting notion, a jargon, a frequently/rarely used

spoken notion, a poetic one, or a term), and the

particular view-point for giving the assessment (the

observer’s, the doer’s or the affected person’s view-

point).

The form, offering a choice for colour matching

allows the user to form a particular, “his/her own”

colour match to each notion by simply dragging the

mouse to a colour field, thus receiving automatically

the corresponding RGB code. The users are also

offered a list of colour samples and the names and

Dictionary

Association

with nouns

Forming

word

expressions

Comparing

notions

Describing

notions

Color

matching of

Emotionality, Need

Rationality,

assessment of words

Association

with verbs

Association

with nouns

and verbs

Answering

by listing

Level and valence

assessment of words

Assessment of

sentences in

Comparing

activities

Forming

questions

Game for

collecting

goals

without MS

Agent Peedy

Game for collecting

goals and emotional

assistance by MS

Agent Peedy.

COLLECTING KNOWLEDGE AND LEARNING LANGUAGES WITH TOWERS OF KNOWLEDGE GAME

475

RGB codes of the most commonly used standardized

colours on the web to choose the most appropriate in

their opinion for the notion to be evaluated.

The form for vocabulary learning is the only one,

not intended to allow collecting knowledge. The

user here is shown a word given in two of the

working languages by random and is supposed to

enter the word in the third language. After a correct

answer he/she is shown a funny picture as a reward,

and he scores points in addition; after a wrong

answer, respectively, he/she can see the correct word

displayed together with an appropriate caricature

and no points are scored. The words are

accompanied by suitable pictures and sound files,

illustrating the right pronunciation in each of the

working languages.

The last two forms contain two versions of a

game aimed at collecting hierarchies of goals related

to the real world. These options will be considered

in details in the further sections of this paper. The

users’ opinion will be studied according to different

criteria.

Almost all of the forms offer the Microsoft

Agent Peedy, capable of pronouncing every text or

word in English. The Microsoft® Agent is a set of

programmable software services that supports the

presentation of interactive animated characters

within the Microsoft Windows® interface.

For the realization of the program system the

environment Visual Studio .NET 2005 is used, as

well as the programming languages C# and

JavaScript, and the Microsoft Agent character

Peedy. The data bases are accomplished in MS

Access. We have been developing our system by

adding new options and possibilities and you can

find it on: www.expertimental.net.

3 THE GAME

In order to collect a big quantity of correct

knowledge and to become popular, a programming

system should be useful for the users, i.e. it should

be funny, interactive, it should contain a competitive

element, not be very complicated, not detract the

user from the main goal, be pleasant, encourage the

user to enter newer and newer knowledge in the

three languages, and stimulate the correct knowledge

introduction.

For achieving these goals a game is realized,

which could be used for collecting hierarchies of

knowledge, e.g. - in the form, collecting hierarchies

of goals by answering the question “What do you

need to….”, or in the form, collecting knowledge

according to different criteria for comparing by

filling in the notions to be compared. For example:

A is bigger than B; B is bigger than C; C is bigger

than D, etc.

The idea is to have at least two players or two

teams with one or more players, i.e., two participants

in the game. Questions or criteria for comparison are

randomly given to them. The participants from the

two teams take turns and the system monitors the

process of taking turns. When the users from the

first team give their answer, the second team has to

approve it and vice versa. There are two special

buttons - for each of the teams - for the approval of

the answers. Only after being approved, an answer is

saved in the database and the user receives reward

points. The points are visualized in the form of three

balloons appearing on the screen and moving at

random speed in various directions. The users can

only claim their points after they pop the balloons by

clicking the mouse on each of them. The balloons

disappear and the updated number of points for the

corresponding team is visualized. This is a short

dynamic element within the game, requiring

quickness and aiming at psychological relaxation of

the user.

The second team is in turn after that. The team

turns are predefined by the program through

enabling or disabling the approval buttons for each

team. If the game continues this way only, then the

teams would gather the same amount of points.

However, while a team is in turn, it can decide to

“build a tower”. When “a tower” is announced, each

of the approved answers becomes the next question

for the team, giving it. In “a tower” mode the teams

do not take turns and the reward points are five, not

three as they are in the case with no tower

announced (ticked). The aim is to build the highest

possible towers of goals. When a team is not able to

continue building the tower, the tick is removed

from the tower check-box and the other team is in

turn.

The reward points are used to open a picture,

which is cut into 64 pieces and originally hidden.

Each of the teams opens its own picture. Five points

are needed to open just one of the 64 pieces of the

hidden image. Thus for opening the whole picture

320 points are required.

Both teams can use their points to open parts

from their hidden picture at each moment of the

game. Depending on the team in turn, points from its

score will be taken off and pieces from its picture

will be opened. Thus each team can click on the

pieces, already opened by the other team,

respectively. As a result, the game can continue for a

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

476

very long time and the teams can compete with each

other all the time until one of the pictures is

revealed. The winner is the team, which completely

opens its picture first. The shortest game is when a

team collects the 320 points needed to open their

picture entirely.

A further addition to the game will be to make a

piece already opened by one of the players worth

more points for the other player to open.

The advantage of the game is that it is easy and

encourages introducing knowledge due to receiving

reward points by answering questions and approving

answers. A way of validating knowledge is thus

suggested by the requirement for receiving approval

from the enemy team. As the two teams compete

with each other, they would not allow introducing

nonsense. The game evokes curiosity from the users

to see what the hidden image is. There is a

competitive element – which team will open their

picture first. There is a bit of dynamics and

relaxation while receiving points when a player has

to click on the balloons, flying randomly

(chaotically) on the screen.

4 THE EXPERIMENT

The aim is to check if the users like the

programming system and to find out which option is

considered to be the most appropriate. 25 users were

asked to test the system. They were offered three

versions of the developed program for collecting

goals by answering the question “What do you need

to…”. The users were supposed to evaluate the

versions by the following criteria: entertainment,

interactivity, intelligence, and adoption, as shown in

Figure 2.

The three versions of the programming system

will be further denoted shortly as it follows:

1. questions-answers; - a version in which the

users directly answer questions, written in a database

beforehand.

2. game without MS Peedy; - this is the game,

described above, without the MS parrot Peedy.

3. game with MS Peedy; - the same game but

this time Peedy participates actively by prompting

the next move, giving advice on the way to become

a winner, calculating the number of points up to a

moment, showing curiosity about the hidden picture,

announcing the reward points, being happy with

each of the caught balloons, compassioning the

players in case of a difficult question etc.

Figure 2: Experimental results.

The three versions of the programming system

were to be evaluated according to each of the criteria

by the numbers from 1 to 10.

Figure 2 illustrates the results from the

conducted experiment. The results show that the

most interesting and preferred option is the program

with the intelligent virtual agent (IVA) Peedy.

It was surprising that the biggest part of the

users, who tested the program, gave equal number of

points to the three options of the program according

to the criterion of intelligence, i.e., they consider that

the introduction of the IVA does not contribute to

the intelligence of the system. It was unexpected,

that there are users, who do not like the IVA at all;

they say that the virtual agents are irritating and

prefer turning them off. Another unexpected result

is that the first option of the program – the one with

direct questions and answers received a higher

appraisal than expected and many users shared that

if they wanted to study a language, they had better

concentrate on their language practice only and

avoid any other elements.

5 CONCLUSIONS

The article illustrates some initial results of the

carried-out survey among users, who worked with a

knowledge accumulating system as a background

process to learning foreign languages. The

COLLECTING KNOWLEDGE AND LEARNING LANGUAGES WITH TOWERS OF KNOWLEDGE GAME

477

introduction of a game and the presence of an

intelligent virtual agent (IVA) within the game

acting as an emotional assistant increase to a certain

extent the users’ interest in the system and keep

them involved for a longer period of time.

Evaluation has been given of three versions of

the program system

The first version suggests forms, containing a

question aimed at collecting a particular type of

knowledge and they need an answer.

The second option includes a game, which

stimulate giving answers to questions from the first

version.

The third option is the same game combined with

emotional assistance by the Microsoft intelligent

virtual agent Peedy.

The experiment shows that there are different

types of users, with different preferences,

understanding and priorities. They are classified here

conditionally and absolutely subjectively as

ambitious, emotional, and demanding-conventional.

The ambitious users are the ones who just want

to study without distractions and without playing

games. Sometimes, just for a break, they are inclined

to pass into the category of the emotional users and

play for a while, but definitely with the Peedy

version of the game.

The emotional users are those, who prefer to

play and learn with the help and in the company of

the MS Agent Peedy. These are the more

communicative and happy users.

The third category of users is the group of the

demanding traditionalists. On the one hand they are

accustomed to the standard text interfaces and prefer

them. On the other hand, the ease of communication

with the new user interfaces somewhat frightens

them. Therefore they tend to move to the category of

the so-called here ambitious users.

It was therefore decided to support the three

versions of the system – direct question-answers;

game without an IVA and game with an IVA.

It is expected that with accumulating more

knowledge about the world we know and with

building a more complicated behavioral system for

an IVA, the users will appreciate their intelligence.

ACKNOWLEDGEMENTS

This work has been supported by the Technical

University Sofia, Project 092ni067-17 “Program

system for multi-language teaching of people and

intelligent virtual agents”, 2009

REFERENCES

Hugo Liu, Ted Seler, and Henry Lieberman, 2003

Visualizing the Affective Structure of a Text

Document MIT Media Laboratory, Copyright is held

by the author/owner(s). CHI 2003, April 5–10, 2003,

Ft. Lauderdale, Florida, USA. ACM 1-58113-630-

7/03/0004.

Henry Lieberman, Dustin A Smith, Alea Teeters, 2007

Common Consensus: a web-based game for collecting

commonsense goals MIT Media Lab 20 Ames St. E15-

X Cambridge, MA 02139, USA +1-617-253-031, IUI

2007.

Luis von Ahn and Laura Dabbish. Labelling images with a

computer game. 2004 In CHI ’04: Proceedings of the

SIGCHI conference on Human factors in computing

systems, pages 319–326, New York, NY, USA,

2004.ACM Press.

Luis von Ahn, Mihir Kedia, and Manuel Blum. 2006

Verbosity: a game for collecting common-sense facts.

In Proceedings of ACM CHI 2006 Conference on

Human Factors in Computing Systems, volume 1 of

Games,pages 75–78, 2006.

Luis von Ahn, Ruoran Liu, and Manuel Blum., 2006

Peekaboom: a game for locating objects in images. In

Proceedings of ACM CHI 2006 Conference on Human

Factors in Computing Systems, volume 1 of Games,

pages 55–64, 2006.

Timothy Chklovski and Yolanda Gil. 2005 Improving the

design of intelligent acquisition interfaces for

collecting world knowledge from web contributors. In

K-CAP’05: Proceedings of the 3rd international

conference on Knowledge capture, pages 35–42, New

York, NY, The USA, 2005. ACM Press.

Douglas B. Lenat., 1995. CYC: A large-scale investment

in knowledge infrastructure. Communications of the

ACM, 38(11):33–38, 1995.

Push Singh. The public acquisition of commonsense

knowledge, 2001.

Mueller, E.T., 1998. ThoughtTreasure: A natural lan-

guage/commonsense platform (Online).

http://www.signiform.com/tt/htm/overview.htm. 1998.

L.Dakovski, D. Budakova, O. Obretenov, V. Georgiev. M.

Kolarov, 2003. Information system for Gathering

Descriptions, Analogues and Associations via Internet

with a view of forming Rational Behavior,

International seminar and a scientific session

“Automation, Electronics and Informatics” FEA

Proceedings, volume II, 20 June, 2003, TU – Branch

Plovdiv, pp. 50-57, Bulgaria.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

478