Keywords
language learning expert system chat ot emerging technologies
Introduction
Emerging technologies open possi ilities for language teaching and learning that have not een possi le in the past. Language learning has traditionally een and still remains handicapped y a variety of scarcities greatly in uencing the e ciency of the language learning process.
To exploit the ne possi ilities these emerging technologies ena le in the language learning process e need to re-examine our notions and roles of teachers classrooms and materials and leave ehind those that provide fe er ene ts than are provided y alternatives. To ta e a step in this direction I attempt to put aside pre-existing ideas of language learning as e have experienced it and consider an ideal language learning process free of many of the current constraints. aving esta lished a target paradigm I ill loo at ho current and emerging technologies can e applied to create a language learning system more suited to this ideal.
In order to eep the discussion at least relatively pragmatic I con ne the discussion to currently availa le and emerging technologies. For this discussion emerging technologies ill e defined as technologies that have potential to e applied to language education proof of concept has een esta lished ut have not een applied to language learning to the degree that they have een implemented and evaluated either through practice or academic study. This rather tight de nition should eep the discussion ithin the realm of hat ould e possi le to achieve ithin the next fe years if aggressively pursued. hich of course means that physical enhancements neural implants and uploading lessons to the rain remain outside this de nition.
Preconceivedassumptions
e have reached the current state-of-the-art of the eld of language education in an environment of scarcity of teachers location materials time and cost. Each of these has had its e ect on ho the eld has ecome hat it is and each also imposes limitations on ho and ho e ciently and e ectively language is taught and learned.
s technologies evolve it is important that e let go of paradigms that continue to eep us ithin
Gordon Wilson
Foundations of an Individualized Language-
Teaching Expert System
old unnecessary constraints and stop us from exploiting hat may e etter alternatives.
Much of hat I descri e here could e udged as o vious hen it is pointed out ut at ris of appearing simplistic I ant to attempt to expand the paradigms of hat is possi le and preferred ecause although potentially o vious hen ma ing these decisions it is not uncommon to remain
ithin our paradigms ased on hat has or ed in the past under old technological conditions.
Exampleoftheinfluenceofparadigm
ecisions a out class size o er a good example of ho paradigm can in uence perspective. hen e thin of prefered class size generally e ould udge smaller teacher to student ratios as etter than larger. o ever still thin ing from a classroom scarce-teaching resource perspective it is also common to assume that having say under ten students ould e prefera le to having a 1 1 ratio.
In the paradigm of the classroom or traditional tutoring environment a 1 1 ratio ould limit the interactions that ould ta e place. It ould remove the possi ility of the learner occasionally moving into the role of o server of interactions force them to e the focus 100 of the time and limit the variety of interactional partners. It ould e easy then to conclude that an ideal teacher student ratio ould include more than one student. o ever stepping outside of traditional paradigms and assumptions of teacher scarcity ould it not e etter to have additional instructors that the student could o serve interacting hen the focus is o the student This is not practical under traditional paradigms ut emerging technologies-- here the teachers are computer ased--open up this possi ility.
The follo ing section of is meant as a paradigm shifting exercise to precede the presentation of a non-traditional approach to language learning teaching and is in no ay an attempt at a thorough deconstruction of the eld. I can say that I have a heavy ias to ard the elief that technologies material or conceptual lead methodological approaches hich tend to lag ehind.
Towardanideal
So hat form ould the ideal language teaching and learning process ta e My first step in attempting to conceive and ideal process is as a mental exercise to remove each of the scarcities mentioned a ove. I am not attempting to argue at this point that each item here is achieva le only suggest that the full or partial removal of each of these scarcities ould have a positive e ect on the teaching and learning process potentially shifting the target ideal out of the classroom.
Numberofteachers
emoving from the e uation the constraints of limited teaching resources e can imagine allocating a private instructor or even multiple private instructors to each learner as in the a ove scenario that alone could e argued to create a great improvement in the progress of the student.
There are de nite ene ts to having other students in a class rather than al ays having a one-on- one session ith the teacher ut ouldn t it e even etter if rather than other students ho ma e
errors and have limited no ledge of the target language additional instructors ere included in the learning experience
Location
emoval of the constraint of location the learning experience could e ith the learner herever they are any availa le time in any appropriate location the lesson could ta e place.
Materials
I imagine non-scarce materials to e free and instantly availa le lessons and activities covering the continuum of di culty levels. comprehensive cro d-sourced i ipedia-style lesson repository pro ect is not eyond the scope of hat could e accomplished.
Time
Time concerns oth the teacher and the learner. Short of physiological or chemical alteration of learners the time ta en to learn appears to e xed. There do not appear to e any improvements to this on the horizon. I ll have to consider the time a student spends engaged in learning as an unchanging varia le for this discussion. o ever I ill consider the potential e ects of non-scarce teacher time cf. the num er of teachers . I elieve that e can consider that unlimited resources of teacher time ould have a positive e ect on learning.
Knowledgeofthestudent
nother scarcity that is easily ta en for granted is the teacher s no ledge of individual student s no ledge and a ility. e have course prere uisites uizzes and tests and our interaction ith students that allo us to form a general idea of each student s level ut there is much e don t no . emoval of this scarcity ould mean that the teacher ould have a thorough no ledge of every s ill and item of no ledge the learner possesses and also therefore everything that they do not no .
So an ideal teaching and learning process ould have all of these ualities. The teacher ould e availa le any here and any time the student ished for a lesson it ould no exactly hat the student needed and dra lessons and exercises from an exhaustive pool of materials to provide individualized instruction of exactly hat the learner needs at the time. ssuming that learner needs include conversational interaction it could serve as a target language spea ing companion or even multiple companions.
I ish to argue that each of these scarcities has the potential of eing reduced or removed y emergent technologies and could result in a much more efficient and effective language teaching and learning process. nd ith this as a target I ill descri e in the sections elo the form that a
learning system of this ind could ta e.
AnIndividualizedLanguage-TeachingExpertSystem
n expert system is an artificial intelligence that emulates the decision ma ing and reasoning of a human expert ac son 199 . I ll descri e in the follo ing sections the form that a language- expert system could ta e. The model is made up of process and data nodes. s seen in Figure 1 the processes are represented y the hite oxes and the data is represented y the grey oxes.
Except for the lesson an all data nodes are created y a process node and then used y the next process node in a cycle. There are three main process nodes in the cycle consisting of
. There are t o inputs into the main cycle
the and the created y the .
The former eing a repository to e accessed for and the latter to inform the .
Figure1.Language-TeachingExpertSystemWorkflow
CoreProcesses
E ective curriculums include feed ac that allo s the process to continually adapt to the current needs of learners . . ro n 199 . The model in this study is uilt upon three core processes that result in a feed ac loop of continual adaptation in uenced primarily y a needs assessment updated ith each cycle. eing cyclical there is no clear starting point from hich to descri e the entire process as each process is in uenced y that hich preceded it. I egin ith the
hich ill ma e reference to the preceding processes hich ill in turn e ela orated on further as e progress through the cycle.
NeedsAssessment.
The receives the data from the data node and
data node. The creates the that ill e then used
for . The the process esta lishes hat
no ledge and s ill the learner has ac uired and hat they have not. The
node adds to this from the learners hat li ely has priority to the individual learner.
LessonSelection.
ta es in the data hich could e characterized as everything the esta lished that the learner cannot do ith extra eighting given to s ills and no ledge that may e of immediate importance or interest to the learner. The
process then eights this data ta ing into account di culty levels and retrieves the appropriate lessons from the . This process ould also incorporate a spaced retrieval system to select previously learned content that are ready for revie . aving done this the
process then produces the output data referred to in Figure 1 as the Lesson.
InteractionwithLearner.
The process receives the that as chosen y the
and presents it to the learner. This could ta e a variety of forms and of course some could resem le traditional interactions ho ever lessons and practice could e incorporated into casual conversation as asides as are done ith human companions ho spea a learner s target language. I imagine that there is an ideal ratio of study to situational language use ut I have not found reference to it yet in the literature. Let s assume for discussion that it is 1 20.
side from delivering the lesson the process ould use this stage to pad the interaction ith conversation hich is strategically interspersed ith language items in need of revie dra ing on a spaced retrieval algorithm.
side from delivering and revie the results of interactions ith the learner ould also e continually assessing learning outcomes referred to in Figure 1 as hich
ould then create the data output referred to as .
EnvironmentalScan
External to the main cycle is the . The attempts to create
a thorough profile of the learners linguistic environment potentially including daily conversation the learners audi le and visual receptive physical environment receptive media environment and receptive and productive digital environments. This is essentially all language and possi ly the voca ulary of the o ects in the learner s physical surroundings percepti le to the learner in their
daily life.
It is not di cult to imagine ho this ould e done ith the digital environment ith oogle and various government agencies having pioneered methods through collection of mar eting data and domestic intelligence collection such as the ISM electronic data mining program exposed y Ed ard Sno den . o ever Microsoft too this further ith it s Lifelogging pro ect in hich eara le audio and video recording devices ere attached to participants in an attempt to collect data from the participant s physical daily environment ara et al. 200 . In addition ell 200 in his similar MyLife its pro ect captured a lifetime s orth of articles oo s cards s letters memos papers photos pictures presentations home movies videotaped lectures and voice recordings and stored them digitally extending it to include phone calls instant messaging transcripts television and radio.
hile there is potential for misuse of data of this readth a potential enefit also exists and collection of data of this ind paired ith data collected from the
process ould result in an extremely lo scarcity of no ledge of the learner and provide rich data from hich to create a very ro ust of each individual learner.
ComprehensiveLessonBank
There are a great num er of lesson repositories Moodle.org pro a ly eing the most idely no n. To e most e ectively exploited a lesson repository ould rst need to e comprehensive.
ith pro a ly millions of English lessons eing created daily the redundancy in the system is enormous ith the very est lessons eing lost to the ma ority of teachers. lthough lesson repository pro ects have made attempts to ma e uality premade lessons idely availa le they remain scattered and far from comprehensive. I imagine a i ipedia-style pro ect that is continuously expanded improved and curated. I place this outside of the language learning expert- system model ecause the process involves other actors hereas all other process in the system are carried out internally y the expert system therefore the enters the model as data.
So hat a out text oo s as lessons I do not consider text oo s here ecause text oo s are a symptom of the classroom paradigm and deliver lessons as a pac age intended to meet the needs of a num er of individuals in the classroom ma ing them specialized to none of the learners. This meets the needs of the traditional classroom paradigm ut is not compati le ith the system I descri e.
TheBot
The nal element of the model to e descri ed is the ot hich ill serve as the interface et een the system and the learner. This technology is emerging in the form of a chat ot. chat ot is a computer interface that allo s humans and computers to interact using natural human language.
This is one of the more exciting emerging technologies. omputers that linguistically interact ith humans have een speculated on and have appeared in ction for decades. t ell 1999 proposed the serious potential of a computer- ased conversational language-teaching system near the turn of
the millenium. nd more recently s et al. 201 descri es the emergence of chat ot technology into the mainstream
Face oo Messenger had zero ots in Fe ruary 201 and over 1 000 y uly 201 . it too pple more than seven months to reach that mar ith apps. i interactive has more than 20 000 chat ots. In the rst seven months that i interactive allo ed romoted hats its 200 million registered users exchanged 0 million messages ith ots.
Even more recently many usinesses are adopting chat ots as a ay to interact ith customers eo Lee 201 . Through voice recognition and speech synthesis voice ased chats are ma ing it possi le to have a conversation ith your computer as you al do n the street. eep learning engaging neural net or s is eing applied to these technologies to improve them further ie et al. 201 . ithin prescri ed domains chat ots can pass o as human. sho oel as descri ed y Moln r Sz ts 201 created a chat ot system to ans er the 10 000 in uiries that ere received from students regarding his class. The chat ot replied to students for an entire year ithout eing identi ed as not eing a human.
These technologies could e developed to specialize in interactive conversational language teaching interactive lessons administering tests and assessments small tal and conversation.
s an element of the process the learner
sinteraction ith the chat ot contri utes to the data.
Userexperience
The main features of the user experience ould e the hard are and the chat ot. The chat ot could ta e the role of a companion interacting ith the learner throughout the day engaging them in conversation exploiting teacha le moments to spontaneously insert a lesson into the dialog commenting on errors it o served in the learner s production and providing alternatives or normal error correction. There could also e time set aside for more explicit lessons especially hen visuals ould etter serve the lesson. For sit-do n lessons the chat ot could serve as a reference source and provide feed ac . This all could regardless of location serve as an immersive language learning experience.
The hard are depending on the level of comprehensiveness chosen y the learner could include the collection of all receptive and productive digital sources as ell as audio and video of the physical environment. This ould re uire cameras and audio recording devices. ecently these can e physically at least rather non-intrusive. It ould also e possi le to include augmented reality a digital layer added onto the perceived environment. This could add la els or comments to o ects in the environment or potentially translate text in the environment into the target language. For a
asic and still rather glitchy sample of this see the camera function of oogle s Translate app.
References
s . . Facemire M. ogan . 201 . The State f hat ots ilot hat ots s art f our pp Mo ile
Strategy. .
t ell E. S. 1999 . pp. 1- 2 . The ritish ouncil.
ell . emmell . 200 . digital life. - .
ro n . . 199 . . einle
einle u lishers 20 ar laza oston M 0211 .
eo M. Lee . . 201 . hat ot as a e usiness ommunication Tool The ase of aver Tal Tal . 1 1- .
ac son . 199 . . ddison- esley Longman u lishing o. Inc.
Moln r . Szűts . 201 The ole of hat ots in Formal Education. IEEE 1 th International Symposium on Intelligent Systems and Informatics.
ara . Tu eld M. M. Shad olt . 200 . Lifelogging rivacy and empo erment ith memories for life.
1 1 -1 2.
ie L. Lee T. Ma M. . 201 . uest Editorial dvances in eep Learning for Speech rocessing.
9 9-9 1.
201 11 2
