Filetype PDF | Posted on 21 Sep 2022 | 3 years ago
Authentication
digital resources
68x Filetype PDF File size 0.56 MB Source: www.cse.iitb.ac.in
Hindi Wordnet for Language Teaching: Experiences and Lessons Learnt
1 2 1
HanumantRedkar , RajitaShukla , SandhyaSingh ,
1 1 1
Jaya Saraswati , Laxmi Kashyap , Diptesh Kanojia ,
1 1 1
Preethi Jyothi , Malhar Kulkarni and Pushpak Bhattacharyya
1
Indian Institute of Technology Bombay, Mumbai, India.
2
Bennett University, Greater Noida, India.
{hanumantredkar, rajita.shukla38, sandhya.singh}@gmail.com,
{jayasaraswati, laxmi.kashyap, dipteshkanojia}@gmail.com,
{preethijb, malharku and pushpakbh}@gmail.com
Abstract ance of one entity entails the appearance of the
other in the mind. Thus, it was found that subjects
Thispaperreportstheworkrelatedtomak-
respond quicker than normal to the word ‘nurse’ if
1
ing Hindi Wordnet available as a digital
it follows a highly associated word such as ‘doctor’
resource for language learning and teach-
(Church and Hanks, 1990). This property of the
ing, and the experiences and lessons that
mentallexiconisstructurally built in the Wordnets
were learnt during the process. The lan-
and manifests itself in the lexical and semantic re-
guage data of the Hindi Wordnet has been
lations which are encoded in it. Thus, a Wordnet
suitably modified and enhanced to make
is a ready resource of vocabulary of a language,
it into a language learning aid. This aid
whichcapturesassociativelearninginitsstructure.
is based on modern pedagogical axioms
Conventionalsourcesofvocabularylearning,such
and is aligned to the learning objectives
as the dictionaries and thesauri, do not have these
of the syllabi of the school education in
relations due to the very nature of their composi-
India. To make it into a comprehen-
tion. This is the motivation to present Hindi Word-
sive language tool, grammatical informa-
net as a tool for vocabulary learning and teaching.
tion has also been encoded, as far as these
Thesecondmotivationisthefactthateducation
can be marked on the lexical items. The
is undergoing rapid digitalization. Innovative in-
delivery of information is multi-layered,
struction techniques, which can cater to the tenets
multi-sensory and is available across mul-
of anywhere, anytime, any size learning, flipped
tiple digital platforms. The front end has
classroomapproachandblendedlearningenviron-
been designed to offer an eye-catching
ments, are the need of the hour. Such technology
user-friendly interface which is suitable
based learning solutions can help in better learner
for learners starting from age six onward.
engagement in classrooms and can also be used
Preliminary testing of the tool has been
ininformalteaching–learningenvironments,asthe
done and it has been modified as per the
delivery of knowledge is in a multi-sensory mode
feedbacks that were received. Above all,
andisavailableacrossvariousdigitalplatforms. In
the entire exercise has offered gainful in-
the form of a digital resource, the language learn-
sights into learning based on associative
ing aid that is presented here also seeks to redress
networks and how knowledge based on
the long standing problem of the burden of the
such networks can be made available to
school bag of young learners in India. This peda-
modernlearners.
gogical application of Hindi Wordnet will address
all the above issues and provide a resource which
1 Introduction
will cater to the various gap areas of learning.
A Wordnet is a large digital lexical database of
The rest of the paper is organized as follows:
a language in which information is organised
Section 2 provides the related work. Section 3 dis-
around cognitive synonym sets or synsets (Fell-
cusses the need for association capturing in lan-
baum, 1998). The underlying basis of such orga-
guage learning, which provides the basis for using
nization are the word association studies in psy-
Wordnet as a language learning resource. Section
cholinguistics, which proved that our mental lexi-
4 presents the digital aid developed for Hindi lan-
con is structured on associations, i.e. an appear-
guage teaching and learning. Section 5 discusses
1
http://www.cfilt.iitb.ac.in/wordnet/webhwn/index.php the process, experiences and lessons learnt while
developing this aid. Section 6 briefs the field test 3 LanguageLearningthrough
and user feedback followed by conclusion and fu- Association Capturing
ture work in section 7.
3.1 Needforassociation capturing
Association capturing lies at the core of language
2 Related Work
learning. Thetermassociationisusedheretorefer
In digital educational technology, the literature
to the connection or relation between ideas, con-
shows that psychological aspects of language
cepts, or words, which exist in the human mind
learning has been explored under various condi-
and manifest in such a way that an appearance of
tions due to continuous advancements on the tech-
one entity entails the appearance of the other in
nological front. The research indicates that multi-
the mind (Sinopalnikova, 2004). Learning or in-
modal learning has always resulted in better re-
structionstrategiesmustbeabletoencapsulatethis
tention (Dale, 1969). When the information en-
association for the creation of better pedagogical
ters the system through various senses, it helps the
techniques, as associative networks help not only
brain to circumvent the limited processing capa-
in understanding new knowledge but also to retain
bilities of each individual senses and allows for
it firmly in the mind.
greater total information to be processed (Clark
The understanding of how the meaning of a
and Paivio, 1991). With technology in place,
wordisunderstoodandretainedinthehumanmind
the ease of multi-modal learning environments
is of crucial importance as vocabulary plays an im-
have been studied in different settings (Mayer and
portant role in all the competencies of language
Moreno, 2003; Moreno and Mayer, 2007; Shams
learning, such as speaking, reading, and writing.
and Seitz, 2008; Sankey et al., 2010). Mobile As-
Methods of vocabulary learning too have moved
sisted Language Learning (MALL) is also being
beyond the traditional ways which were based on
explored as mobile technologies are becoming an
the behavioristic theory (Demirezen, 1988) of lan-
integralpartoflifestyle. Thefindings(Yang,2013)
guage learning to the modern methods of vocabu-
showthat MALLhasnotreacheditspotential and
lary learning (Nation and Newton, 1997). The lat-
it is moving towards being the new stage of Com-
ter are basedonthecommunicativetheory(Brown,
puter Assisted Language Learning (CALL). Peda-
2000), where understanding a word involves com-
gogical experts have stressed on the need for im-
mitting to memory its form, capturing its meaning
proving various approaches to enhance the will-
and finally knowing how and where to use it.
ingness of the learners for self-directed technol-
The semantic network theory (Collins and Lof-
ogy to maximize the technology potential for lan-
tus, 1975; Collins and Quillian, 1972; Rips et al.,
guage learning (Lai et al., 2016). The use of gam-
1973;Smithetal.,1974)statesthataword’smean-
ification is seen as an effective pedagogical strat-
ing is defined as “whatever comes to mind when
egy which can engage and motivate the learner to
someonesaystheword”or”youshallknowaword
learn in a relaxed environment, which is funda-
by the company it keeps” (Firth, 1957). It implies
mentaltoanylearning(WerbachandHunter,2012;
that a word’s meaning is represented in the mental
Figueroa Flores, 2015).
lexicon by a set of associations of that word with
In language learning, using semantic network otherwords. Thus,themeaningofawordisunder-
relations for learning new word helps in better un- stood as collections of associated concepts . Also,
derstanding of its meaning (Lin, 1997). The word- it has been proved that the syntactic category of
net, a semantic based rich lexical resource, has a word and the associated words that come to the
been used for various language learning applica- mind is the same (Fillenbaum and Jones, 1965).
tions such as - the semantic and lexical relations Since it is rare in any discourse for adjacent words
between synsets enables the learners to know the to be from the same syntactic category, therefore
connotationsofawordalongwithitsvariouspossi- this cannot be explained as association by being
ble contexts (Hiray, 2015). A gamification system contiguous. This association is because a word’s
based on wordnet was used to assess the depth of meaning is represented in the mental lexicon by a
word knowledge of a learner(Brumbaugh, 2015). set of nodes and the links between them. Here the
Asystemwasexperimentedforsimilarlookingand nodes represent concepts whose meaning the net-
nearsynonymswordlearningbasedonwordnetfor work is trying to capture, and the links represent
English language learners (Sun et al., 2011). relationships between concepts.
These nodes and links translate into word rela- Another evaluation was done based on four dif-
tionships and meaning relationships or lexical and ferent criteria, which were, (i) Number of Rec-
semantic relations. These lexical and semantic re- ommendations – those recommended by most of
lations have been found to be cognitive universals the ten ESL resources, (ii) Authority – authors of
(Lin,1997),i.e.,theserelationsarefoundinalllan- the sites should be related to the field of language,
guages. Exploring the various relations in the se- (iii) Simplicity – in terms of presentation, be user-
mantic field such as hypernymy, hyponymy, etc., friendly andmaterialclassifiedasperdifferentlev-
consolidates a new word’s position in the student’s els of learning and (iv) Currency – the site is reg-
mind. ularly updated. Here a total of 6 useful sites were
marked, out of which Princeton WordNet received
a ranking of 4. These results can be projected on
3.2 Associative network based Wordnet for
HindiWordnetaswell,sincestructurallyWordnets
language learning
are similar. Thus, empirically, a strong case for us-
As stated above, the learning and teaching of vo-
ing wordnet for vocabulary teaching is made.
cabulary with associative networks is helpful as
new words are presented in semantically related
group. Thus,itisbestwhenadjectivesaretaughtin 4 ADigitalAidforHindiLanguage
clusters around antonymic pairs, nouns are taught Teaching and Learning
as hyponymsofothernounsorwithsynonymsand
verbs are presented as groups of troponyms or en-
In order to employ association capturing in learn-
tailments. For learners, meaning is captured and
ingandtocatertotheneedsofrapiddigitizationof
retained when, for example, it is said that a horse is
education,thevastlexicaldatabaseofHindiWord-
a kind of an animal (hypernymy relation), uniform
net has been transformed for pedagogical purpose
is opposite of diverse (antonymy relation), snore is
in the form of an e-learning tool - Hindi Shab-
2
a part of sleep (entailment relation) and cultivate
damitra . In this tool, Hindi wordnet data is modi-
has as its object land (argument relation). Such
fied / simplified and further augmentedwithaudio-
associations are not captured in conventional dic-
visual features and grammatical properties, and is
tionaries as much of the structural information is
presented in a learner-friendly format for language
omitted from them and can only be provided by
teaching and learning. Depending upon the under-
semantic networks like Wordnets. A classroom
standinglevelofauserandthepurposeofuse,this
teacher will certainly not be able to provide as-
digitalaidfollowstheselectiveinformationpresen-
sociative information to students, by merely using
tation approach. Henceforth, in this paper, this e-
a conventional dictionary. A thesaurus does give
learning tool will be referred as Digital Aid.
synonyms, but lack in lexico-semantic relations.
Moreover,boththedictionariesandthesauridonot
4.1 Whyselective information presentation?
givethegrammaticalfeaturesthathavebeenadded
to the teaching aid (discussed in section 5.4).
The Digital Aid caters to various types of users,
Various tests have proven that amongst the
viz., school students, teachers, parents, language
plethora of web-based resources for language
learners, content managers, proofreaders, natural
teaching and learning, which have come up due
language processing researchers, mobile/web ser-
to the exponential growth of the Internet and the
vice providers, tourists, etc. As per the need
WorldWideWeb,PrincetonWordNethasemerged
of these users the information content has been
as one of the most reliable, authentic and useful
moulded, keeping in sight the cognitive load that
sites (Hiray, 2015). Infirst of twosuchevaluations,
the user may be able to cope with. This can avoid
three criteria, which were, (i) Non-commerciality
unnecessary learning efforts, wastage of time and
–suchsites were freely available, (ii) Adequacy –
the burden of information overloading. In Digi-
those that covered all the 570 word families fea-
tal Aid, this is achieved through the multi-layered
turing in the AWL (Academic Word List) 1 and
information rendering approach, suitable for both
(iii) Authenticity – the academic vocabulary ex-
formal and informal learning environments.
ercises belong to the site itself. The result of this
evaluation has marked 14 useful sites in which the
th 2
Princeton WordNet stood 4 . urlhttp://www.cfilt.iitb.ac.in/hindishabdamitra/
4.2 Multi-modal learning - a psychological understand and learn the concept. Audio pro-
aspect nunciationofasearchedwordandapictureor
an illustration of a given concept is provided.
Humansaregeneticallyprogrammedtocommuni-
Apart from this, the corresponding English
cateandunderstandthingsthroughamulti-sensory
WordNetsynsetis provided. For those words
learningsystem. Thebrainprocessestheinputsre-
which do not have a corresponding synset it
ceived from different senses to comprehend con-
is given a bilingual mapping (Singh et al., ).
cepts. As per the principles of multimedia learn-
ing (Gilakjani et al., 2011), multi-modal learning
• Level 2 (Intermediate): In level 2, the inter-
leads to comprehensive learning of a concept with
mediateusersaretargeted. Here,usersareex-
higher retention rate. They have shown to stimu-
pected to have basic knowledge of Hindi. In
late thinking and learning as compared to only text
this level, all the information from level 1 is
basedcomputerinterfaces(ClarkandPaivio,1991;
rendered and additionally, more grammatical
Mayer and Moreno, 2003; Moreno and Mayer,
features such as gender for nouns, antonym
2007;ShamsandSeitz,2008;Sankeyetal.,2010).
for verb and adverb, type of verb, countability
For language learners, having a multi-modal e-
for adjective, type of adjective, spelling vari-
learning tool provides motivation for learning the
ation, etc. are appended.
second language along with the independence to
learn at one’s own pace (Lai et al., 2016). It also
• Level3(Proficient): Inlevel3,additionalin-
provides the confidence of learning and handling a
formation is presented which is necessary for
digital device with ease from an early age.
the intended user. Here, instead of simplified
In Digital Aid, information is provided in the
gloss and simplified example sentence(s), the
form of text as well as audio-visual inputs. The
Hindi wordnet’s original gloss is rendered.
textual information pertains to the gloss (original
This is because, it is expected that these users
Hindi wordnet or simplified), word usage, syn-
have a good grip over Hindi and can under-
onyms, grammatical features, lexico-semantic re-
standthecomplexityofalanguage. Fromthis
lations, ontological information. It also has audio
level onward it is very important that the con-
pronunciations of words and pictures/illustration
cept should be clearly explained, so that the
of concept, etc. This helps in learning and under-
learners can understand the fine grained dif-
standing the concept with great ease.
ference between two synsets. Figure 1 shows
ascreenshotofthetoolrenderingawordभजन
4.3 Multi-layered presentation - an
(bhajana, hymn) at Proficient level.
incremental learning approach
The presented Digital Aid is designed keeping in
mind various aspects of language teaching and
learning. It is a five layered model where selec-
tive informationisrenderedineverylayer,depend-
ing uponthetype/needoftheuserandhis/hercog-
nitive competence. This multi-layered model is
structured in a level-wise and class-wise manner.
Level-wise information presentation
The level-wise module is designed for informal
setup where a user can select any of the five levels
Figure 1: Screenshot of the Digital Aid rendering
depending upon his level of expertise. The infor-
a word भजन (bhajana, hymn) at Proficient level
mation is rendered level-wise as follows:
• Level 1 (Beginner): Level 1 is meant for • Level4(Advanced): Atlevel4,otherseman-
users who are new language learners. Here tic relations such as hypernymy, hyponymy,
information such as simplified concept defi- etc. are introduced along with all the infor-
nition, word usage, synonyms, grammatical mation presented at level 3. Here, all the
features is selectively presented. The gloss available synsets in Hindi wordnet and all the
is simplified so that the beginners can easily grammatical features are rendered.
no reviews yet
Please Login to review.
