JAPASSIST: AN ASSISTANT FOR STUDYING JAPANESE WHILE

SURFING THE WEB

Shiori Daigo

, Rosario Giunta

, Giuseppe Pappalardo

, Emiliano Tramontana

Facolta’ di Lingue e Letterature Straniere

Dipartimento di Matematica e Informatica

Universit

a di Catania

Keywords:

Web personalisation, user assistance, e-learning.

Abstract:

The web offers an increasing amount of pages written in several languages. Given the variety and the quality

of some pages, these can be helpful as an additional language learning support. Indeed, when studying a

foreign language, reading newspapers, magazines, etc. is one of the preferred ways for learners to improve

their abilities. However, without an adequate support they would be forced to make extensive use of the

dictionary.

This paper proposes Japassist, a browsing assistant that supplies users with a customisable amount of sug-

gestions, looked up into dictionaries, useful to read and study the Japanese language while surﬁng the web.

Japassist aims at supporting not only users who need a beginner-level understanding of Japanese documents,

but also those having some comprehension of the Japanese language. Actually, given the Japanese language

intrinsic complexity, some help to read or appreciate the meaning of uncommon kanji symbols may be useful

even for native speakers.

We have employed the aspect-orientation technology to customise a bare web browser with the ability to assist

users reading and studying Japanese. This design approach allows the web browser classes to be unaware of

assistance classes, thus promoting low coupling between different concerns.

1 INTRODUCTION

Given the large amount of web pages written in sev-

eral different languages, it is generally useful to pro-

vide web users with suggestions about the meaning

of foreign text in visited pages. As it happens when

studying a foreign language, reading the text avail-

able from web sites offering newspapers, magazines,

etc. greatly helps learners to improve their abilities.

Moreover, the electronic version of a page is ideally

suited to be customised, with information looked up

into dictionaries, so that learners can enjoy a promptly

available support.

This work proposes Japassist, a browsing assis-

tant that supplies users with a customisable amount

of suggestions useful to read, translate and study

Japanese pages while surﬁng the web. Japassist runs

on the user host, which makes it is easy for the user to

activate, deactivate, or choose the desired amount of,

suggestions.

Support for understanding a foreign language web

page is only limited currently to single word trans-

lation (Google, 2005b) and (often misleading) whole

text translation (Google, 2005a; Altavista, 2005).

Several features make Japassist assistance differ-

ent from other supports: (i) words can be separately

listed and explained; (ii) words and characters can be

highlighted to ease user reading and understanding;

(iii) text can be supplemented with reading assistance

for kanji

, hiragana

and katakana

; (vi) Japanese

words, as well as their composing characters, can be

enriched with more than one English translation and

suggestions on the context where they should be used;

(v) grammar particles can be highlighted, thus giving

hints on grammar rules regulating text; ﬁnally, (vi)

at the architectural level, the use of aspect-oriented

programming (AOP) (Kiczales et al., 1997; PARC,

2005) affords the ability to develop a personalised

web browser, without affecting the original browser

System of Japanese writing using Chinese characters.

A kanji can be a word on its own or a part of a word.

System of syllabic writing for Japanese, especially

used for function words and inﬂections.

System of syllabic writing for Japanese, especially

used for words of foreign origin.

245

Bleisch S. (2006).

ELML, THE E-LESSON MARKUP LANGUAGE - Developing Sustainable e-Learning Content Using an Open Source XML Framework.

In Proceedings of WEBIST 2006 - Second International Conference on Web Information Systems and Technologies - Society, e-Business and

e-Government / e-Learning, pages 245-251

DOI: 10.5220/0001254502450251

 SciTePress

source code with changes.

As far as the software architecture is concerned, a

simple Java browser has been enhanced by connect-

ing it with Japassist assistance functionalities. Con-

nection is obtained by means of AOP, thus maintain-

ing the two main concerns we have focused on, i.e.

browsing and assistance, clearly separated.

Thanks to the noted assistance functionalities,

Japassist is useful not only as an automatic tool for

translation, but also as an assistant for users wish-

ing to learn, or improve their knowledge of, Japanese,

while surﬁng. Actually, given the Japanese language

intrinsic complexity, users having a certain com-

prehension still need help for understanding many

other kanji (among the around 50 thousand existing

kanji (Kubota, 1989)). Moreover, some support to

read or appreciate the meaning of uncommon kanji

and words is useful to native speakers, or to advanced

level learners.

The sequel of the paper is as follows. Next section

presents the detail of assistance functionalities. Sec-

tion 3 describes the software architecture of Japassist

and its connection to a web browser. Section 4 exam-

ines the related work. Finally, section 5 draws some

conclusions.

2 ASSISTANCE

FUNCTIONALITIES

Japassist assistance functionalities can be grouped

in three categories: (1) counting of kanji and words

occurrences, (2) kanji and words highlighting, and

(3) suggestions about reading and English meaning.

These are described in the following three subsections

respectively.

2.1 Counting Kanji and Word

Occurrences

Japassist holds and updates the count of occurrences

of kanji and words as two ranked lists, i.e. kanji-

count and wordcount, to reﬂect the web pages vis-

ited by the user. The kanjicount and wordcount lists

count how many times each kanji and word, respec-

tively, has been found while surﬁng the web. In a cus-

tomised way, these two lists suggest users the most

useful kanji and words. This is similar to what the

Halpern dictionary (Halpern, 2003) provides, i.e. pro-

viding ranked lists of words and kanji on the basis of

their use in daily newspapers.

Figure 1 shows the dialogue window that, upon

user request, reports the wordcount and kanjicount

lists, in the left and centre panels, respectively.

On this dialogue window, the user can select a word

or kanji, in the left and center panel respectively, to

obtain in return the following information:

1. in the bottom panel of Figure 1, the reading

and meaning in English, gathered by looking up

the electronic EDICT

(Breen, 2005a) and KAN-

JIDIC

(Breen, 2005b) dictionaries for words and

kanji respectively;

2. in the right panel of Figure 1, the list of existing

words that use the selected kanji, produced by scan-

ning the EDICT dictionary for a word including the

kanji.

Figure 1: Counting kanji and word occurrences.

2.2 Kanji and Word Highlighting

The aim of characters and word highlighting is to

make it easier for learners to remember the most im-

portant kanji and to suggest the usage of grammar

particles and character sets. Four different highlight-

ing modes are at user disposal: (i) Jouyou and Jinmei

kanji, (ii) customised list of kanji or words, (iii) gram-

mar particles, and (iv) character sets colouring.

EDICT consists of a text ﬁle in which a row is an entry

of the dictionary. Each entry contains a word, its reading

and meaning. Most of the words have a single reading and

meaning. However, words with multiple readings and/or

meanings are dealt with by separate entries.

Similarly to EDICT, in the KANJIDIC ﬁle each

row/entry contains a kanji, its readings, meanings and some

additional data such as its frequency (Girardi, 2005). Each

kanji is found in one row only.

WEBIST 2006 - E-LEARNING

246

Figure 2: Kanji highlighting, furigana reading and translation tooltip.

In March 1981, the council of Japanese language

submitted the list of 1945 kanji, called Jouyou kanji,

that are most used in daily life, press, broadcasting,

law and ofﬁcial documents; and the list of 166 Jinmei

kanji that are used for names, which became 284 in

1992 (Shinmura, 1994).

In highlighting mode (i), when a web page is re-

quested by the browser, Japassist ﬁnds all Jouyou and

Jinmei kanji in it and sets their background colour to

green. The background colour is changed by adding

the span tag (World Wide Web consortium (W3C),

2005a) around the kanji into the HTML source of the

visited page.

The background colour varies from light green to

dark green according to the kanji classiﬁcation, i.e.

the kanji that are usually learnt in the lower classes

of elementary school have a darker background than

those in higher classes (information on which classes

a kanji is taught in is available in KANJIDIC). Fig-

ure 2 shows an example of the green background of

kanji and words on the Yahoo! Japan home page.

In highlighting mode (ii), Japassist marks as light

grey the background of a user deﬁned list of kanji or

words. This list is initially set by Japassist, with the

most frequent kanji found in the visited web pages

(see kanjicount in section 2.1), and can be modiﬁed

by the user through a provided front-end.

As for highlighting mode (iii), yellow is used to

mark the background of hiragana grammar particles.

These particles are used for distinguishing the role

(subject, object, etc.) of nearby words (see e.g. the

yellow background colour of hiragana in Figure 2).

Finally, in order to help users determine to which

writing system, i.e. kanji, hiragana or katakana, a

character belongs to, Japassist can be set to mark

the text using an associated colour for the foreground.

E.g. we have chosen black for kanji, blue for hira-

gana and dark grey for katakana. Figure 3 depicts a

rendered web page fragment whose foreground colour

reﬂects whether text is in the hiragana or kanji script,

while background colour highlights grammar parti-

cles and Jouyou and Jinmei kanji.

Beginner-level learners can take advantage of text

colouring, since a typical difﬁculty they face is to dis-

tinguish between character sets.

The current version does not use a grammar analyser,

therefore some hiragana, used as a phonetic character, could

be mistakenly recognised as a particle.

JAPASSIST: AN ASSISTANT FOR STUDYING JAPANESE WHILE SURFING THE WEB

247

Figure 3: Hiragana and kanji highlighting.

2.3 Suggestions for Reading and

English Meaning

According to user preferences, Japassist can decorate

the text of visited web pages with help information, to

supply, both for words and kanji, some aid in terms of

reading, transliteration, English translation or English

semantic explanation. These are orderly described in

the following items.

• Reading assistance is available for kanji and can be

written in hiragana and katakana. This can be use-

ful for learners that can read hiragana and katakana,

but have a scarce familiarity with the harder kanji

script. Kanji can be read in three different ways:

on, kun and nanori. Our support can be set to pro-

vide the three different readings.

Figure 2 shows several examples of reading sug-

gestions for kanji provided by furigana, i.e.

(smaller) text inserted on top of a word or kanji.

• Transliteration aims at providing the phonetic

equivalent of hiragana, katakana and kanji into ro-

maji, i.e. roman writing. When the transliteration

suggestion has been selected, the Japanese text of

the visited web page is substituted with the translit-

erated version, as shown e.g. in ﬁgure 4. In this

transformation, the shape and structure of the doc-

ument remain almost unaltered.

Transliteration in romaji is useful for beginner-

level learners that are not acquainted yet with the

hiragana and katakana writing systems.

• English translation assistance is available for words

written in hiragana, katakana and kanji. It provides

users with the corresponding English word or the

explanation of the meaning in English, according to

EDICT and KANJIDIC. The user can set whether

s/he wants to obtain the main or all the meanings

corresponding to a kanji or a word.

Since each kanji can have several meanings, the

appropriate one is obtained by relating a kanji

with contiguous ones, e.g. our support provides the

translation corresponding to the longest set of kanji

found into KANJIDIC.

Figure 4: Transliteration of a web page in romaji.

Once a combination of assistance suggestions are

activated the user can surf through several web pages

and automatically be supported by the selected assis-

tance. The various assistance suggestions are at the

user’s disposal and, upon selection, they can be pro-

WEBIST 2006 - E-LEARNING

248

vided in three assistance modes: tooltip, furigana and

append mode.

The tooltip assistance mode provides the user with

a small yellow window containing additional infor-

mation when the mouse is pointed over a kanji or

a word. This assistance mode allows Japassist to

supply suggestions unobtrusively, i.e. the web page

remains unaltered, and is meant for advanced level

learners that need not have a large amount of sugges-

tions. Figure 2 shows a web page including a transla-

tion tooltip.

The furigana assistance mode uses ruby HTML

tags (World Wide Web consortium (W3C), 2005b) to

insert on top of a word or a kanji the additional sup-

porting text (this additional text is said furigana). This

assistance mode, unlike the previous one, does not

conceal the additional text, so that the web page for-

mat becomes slightly different than the original page.

This assistance mode is more suitable for medium-

level learners. Figure 2 shows a web page including

several fragments of furigana text providing reading

suggestions.

The append assistance mode provides the addi-

tional text by mixing it with the original text. Sugges-

tions are added just after the word or kanji they refer

to, and both the word or kanji and the suggestions are

enclosed within square brackets. Although append

mode largely modiﬁes the original page layout, it can

be particularly useful for beginner-level learners that

need a large amount of suggestions.

3 ASPECT ORIENTED

SOLUTION

The two main concerns we deal with are browsing and

assistance. By means of the proposed aspect-based

software architecture these concerns are clearly inde-

pendent, i.e. no classes implementing the ﬁrst concern

should be aware of classes implementing the second

concern and vice versa. Browsing and assistance con-

cerns are connected to each other by means of the Ad-

dOn aspect. This aspect provides the desired ability to

develop a personalised web browser without inserting

any changes into the source code.

Japassist assistance functionalities have been im-

plemented as a namesake Java package, whose classes

receive a HTML page and enrich it with the additional

data corresponding to the desired suggestions. The

JBrowser package implements browsing functionali-

ties.

JBrowser behaviour is customised with the results

Japassist provides by means of the AddOn aspect.

According to the aspect-oriented approach, AddOn

speciﬁes the JBrowser operations, e.g. method invo-

cations, that need to be connected with the methods of

Japassist. Figure 5 shows an overview of the archi-

tecture and highlights the classes of the two packages

that are connected to each other by the AddOn aspect.

JBrowser

Japassist

HighlighterByteGetter

<< aspect >>

AddOn

JListenerWebBrowser

PageSetterJapassist

Segmenter

Figure 5: Aspect-oriented architecture for JBrowser and

Japassist.

3.1 Application Concern JBrowser

JBrowser is a simple Java web browser, consisting

of classes that handle, for a given web address, page

downloading, rendering, etc.

Rendering is based on an existing component,

such as Microsoft Internet Explorer or Mozilla for

Windows, and Mozilla for Linux. JBrowser con-

nects with the chosen non-Java rendering compo-

nent by means of the Java Desktop Integration Com-

ponents (JDIC) libraries (Java Community, 2005).

Within JBrowser, the class WebBrowser of package

org.jdesktop.jdic.browser provides a blank area into

which web pages can be displayed by invoking the

setURL(aUrl) method. In turn, this class invokes na-

tive methods (through JNI) of the available browser.

In addition, the said JDIC package provides in-

terface WebBrowserListener that allows an instance

of the class implementing this interface to be noti-

ﬁed when some events occur on the WebBrowser in-

stance it is registered to. In this context, class JLis-

tener is an implementation of WebBrowserListener

and its methods e.g. downloadStarted(), download-

Completed(), etc. are invoked when the corresponding

events occur.

3.2 Assistance Concern Japassist

The assistance functionalities of the package Japas-

sist are provided by classes Japassist, PageSetter,

Segmenter, Dictionary, Highlighter, OccurDialog and

Occurrences.

Class Japassist is implemented on the basis of the

Facade design pattern (Gamma et al., 1994) and is

responsible to invoke the methods providing the user

requested assistance support within the visited web

page.

PageSetter is intended to change the original

HTML page source in order to insert into existing

JAPASSIST: AN ASSISTANT FOR STUDYING JAPANESE WHILE SURFING THE WEB

249

links their absolute path. This allows links within the

local enriched copy to still work correctly.

Segmenter provides methods allowing a sequence

of characters, taken from the web page, to be split into

words. Since in Japanese text words are not separated

by blanks, in order to detect whether a sequence of

hiragana, katakana or kanji constitutes one word, we

need to search the longest sequence of characters that

matches a word listed in EDICT.

Dictionary holds the EDICT and KANJIDIC dictio-

naries, and the list of grammar particles. Its methods

make it possible to look up words and kanji, and re-

turn their readings and meanings.

Highlighter enriches HTML source by adding tags

allowing text to be coloured (cf. section 2.2). It also

manages tooltips.

OccurDialog is responsible to draw the window

showing word occurrence (cf. Figure 1) taking as in-

put the data organised as class Occurrences. Occur-

Dialog is instantiated when the user clicks on the cor-

responding button of the browser window.

3.3 Aspect AddOn

Aspect AddOn is responsible to connect the browser

with assistance activities.

AddOn deﬁnes pointcut init() that is responsible to

insert some buttons and checkboxes into the browser

front-end when an instance of WebBrowser is created.

These buttons allow the user to: (i) select the desired

suggestions that will be inserted into pages while surf-

ing the web; and (ii) interact with additional Japassist

dialogue windows (e.g. cf. section 2.1).

At run-time, Japassist should insert suggestions

into visited web pages when their download is ﬁn-

ished. For this, AddOn deﬁnes pointcut completed(),

see listing in Figure 6, that captures method docu-

mentCompleted().

pointcut completed() :

execution (public void

documentCompleted(WebBrowserEvent));

Figure 6: A pointcut capturing the termination of pages

download.

Figure 7 shows the sequence diagram for the ac-

tivities that are started once a page download has

completed. The deﬁned AddOn after Advice, i.e. the

code that is executed after the pointcut has captured

control from JBrowser, provides means to bring con-

trol to class Japassist that starts a set of activities.

Firstly, the bytes corresponding to the HTML page

are extracted by means of a supporting class ByteGet-

ter. Secondly, PageSetter is invoked to make ex-

isting links absolute. Then, class Japassist invokes

other methods to insert the assistance text into the web

page. In the example of Figure 7, highlighting is ob-

tained by invoking class Highlighter.

Finally, the modiﬁed page returns to class Japas-

sist that updates the original page, rendered by the

browser, with the one enriched with suggestions.

:JBrowser :Highlighter

:ByteGetter

:Japassist

document-

Completed()

startGetSet()

create()

getPage()

create()

makeAbsolute()

colorText()

setURL()

Pointcut:

downloading ended

Advice:

activate Japassist

Captured by

execution(public void documentCompleted(...))

:PageSetter

Figure 7: Sequence diagram for the assistance activities.

4 RELATED WORK

Google Language Tools (Google, 2005a) and Ba-

belﬁsh (Altavista, 2005) are some examples of ser-

vices available on the web that provide text transla-

tion from Japanese to English. Once such a transla-

tion service has been requiested, it can remain active

to operate on the following visited web pages. These

services are meant as a translation support rather than

an e-learning tool like Japassist.

Googlebar (Google, 2005b) is a tool that pro-

vides single word translations, by means of a tooltip,

from English to other languages, including Japanese.

In general, this is more suitable to Japanese native

speakers than to Japanese learners. Japanese learn-

ers having had an English word translated into kanji

would need some further hints, such as e.g. how to

read it, what are the related meanings, how to use it.

Thus, this tool has a quite limited e-learing support

for Japanese.

WEBIST 2006 - E-LEARNING

250

JquickTrans (Systran, 2005) is an e-learning tool

that allows users to create the kanji and word lists they

need to study. This tool uses EDICT and KANJIDIC

dictionaries. The aim of JquickTrans is to provide an

advanced electronic dictionary with useful annotation

facilities, which is quite different from enriching web

pages with e-learning support.

The Japanese WorkBench project (Winiwarter,

1999) is an e-learning tool for the Japanese language

that, using a web-interface, provides support for read-

ing and understanding kanji and words. A text is

transformed by WorkBench so that each kanji and

word is provided with a link that, upon clicking, ﬁlls

a separate frame with readings and meanings.

Compared with the latter tool, Japassist affords a

more advanced support, in that: (i) its assistance func-

tionalities are more numerous and dynamically avail-

able while surﬁng the web; (ii) selected suggestions

can be set to interfere as little as possible with the text

and the user’s reading activity; and (iii) additional dia-

logue windows are available to list useful and frequent

words and kanji.

At the architectural level, the use of aspect oriented

programming avoids any interference between assis-

tance and browsing functionalities, hence any need to

modify or adapt browser code.

5 CONCLUSIONS

The Japassist e-learning support for Japanese lan-

guage has a rich set of features that can satisfy

the needs of beginner and advanced level learners.

These features range from counting occurrences to

text colouring and highlighting, as well as explain-

ing the meaning of kanji and words into English. All

the e-learning aid is available while browsing the web

on-line by enriching the visited web page.

A simple Java browser has been enhanced and

connected with Japassist assistance functionalities

through an appropriate aspect. This provides means

for the two main concerns we have focused on, i.e.

browsing and assistance, to be kept clearly separated.

Performance of Japassist strongly depend on the

amount of required suggestions and on the length of

the web page. On a 2.6GHz Pentium 4, 1GB RAM,

host updating the HTML source of Yahoo! Japan

home page to have the several suggestions shown in

ﬁgure 2 takes around 3 seconds. However, the origi-

nal page is rendered once download is completed, and

suggestions are added in the background, thus the said

delay often does not affect the user.

The preliminary experimental tests with Japanese

language learners have shown that the provided aid

is very appreciated. By using Japassist, learners ﬁnd

it easier to read and comprehend Japanese text both

with teachers and on their own. Besides reducing time

and effort for manually looking up dictionaries, the

adjustable amount of suggestions allowed learners to

ﬁne-tune the extent of additional text, letting them fo-

cus on just what they needed.

REFERENCES

Altavista (2005). Babel Fish Translation. WWW.

http://babelﬁsh.altavista.com.

Breen, J. (2005a). EDICT Project Home Page. WWW.

http://www.csse.monash.edu.au/ jwb/edict.html.

Breen, J. (2005b). KANJIDIC Project Home Page. WWW.

http://www.csse.monash.edu.au/ jwb/kanjidic.html.

Gamma, E., Helm, R., Johnson, R., and Vlissides, R.

(1994). Design Patterns: Elements of Reusable

Object-Oriented Software. Addison-Wesley. Reading,

MA.

Girardi, A. (2005). Word Frequencies. WWW.

http://ftp.cc.monash.edu.au/pub/nihongo/wordfreq.

Google (2005a). Google Language Tools. WWW.

http://www.google.com/language tools.

Google (2005b). Google Toolbar. WWW.

http://toolbar.google.com.

Halpern, J. (2003). The Kodansha Kanji Learner’s Dictio-

nary. Kodansha International.

Java Community (2005). Java Desktop In-

tegration Components (JDIC). WWW.

http://javadesktop.org/articles/jdic/.

Kiczales, G., Lamping, J., Mendhekar, A., Maeda, C.,

Lopes, C. V., Loingtier, J. M., and Irwin., J. (1997).

Aspect-Oriented Programming. In Proceedings of

the 11th European Conference on Object-Oriented

Programming (ECOOP’97), volume 1241 of Lec-

ture Notes in Computer Science, Berlin, Germany.

Springer-Verlag.

Kubota, Y. (1989). Grammatica di Giapponese Moderno.

Ca Foscarina.

PARC (2005). AspectJ Project. WWW.

http://www.parc.com/research/projects/aspectj/.

Shinmura (1994). Koujien, 5th edition. Iwanami Shoten.

Systran (2005). JquickTrans. WWW.

http://www.coolest.com/jquicktrans/.

Winiwarter, W. (1999). A Language Learning Environ-

ment for Assisting Foreigners in Reading Japanese

Web Pages. In Proceedings of the 5th International

Congress on Terminology and Knowledge Engineer-

ing, Vienna.

World Wide Web consortium (W3C) (2005a). Hy-

perText Markup Language (HTML). WWW.

http://www.w3.org/MarkUp/.

World Wide Web consortium (W3C) (2005b). Ruby Anno-

tation. WWW. http://www.w3.org/TR/ruby/.

JAPASSIST: AN ASSISTANT FOR STUDYING JAPANESE WHILE SURFING THE WEB

251