Page 22 - Winter Issue 2018
P. 22

Deep Language Learning
(5) cloud-based virtual and augmented reality applications other mobile devices (e.g., virtual reality [VR] goggles and
that extend or replace the user’s physical environment artificial intelligence [AI] -enabled eyewear).
through simulation of a variety of situations and envi-
ronments. computer-Assisted Pronunciation
These, along with advances yet to come, will transform the Evaluation and Training
learning eXPerienCe, n0'E 0nlY for language instruction but Pronunciation training is where CALL has long deployed
also for pedagogy in general. cutting-edge technology (Eskenazi, 2009). Several early proj-
What is the current State of language learning technology’ ects used speech technology to evaluate a student’s fluency,
and where is it heading? Before answering, let’s first review Pronunciation Proficiency’ and Comprehension‘ An example
. ’ ’ der ro’ect in which Iapanese students were
the history of CALL (Bax, 2003). 1s SR1 S Autogra P l
evaluated on their ability to speak English intelligibly. An al-
A Brief History Of c°mputer_Assisted gorithm was developed to emulate intelligibility judgments
Language Learning oftnative speakers biit lacked remedial feedbacknfome of
Computers were introduced into language instruction around thls technology was mcorhorated mm Ijhonepass ’ an an"
1960 to supplement programmed classroom instruction. Al- t(_)mat1c_ Srstenl for evaluatlng 3 Students fluency and Prof“
though the technology was primitive by today’s standards, Clencr In Enghsh (Bernstein and Cheng’ 2007)‘
early CALL projects demonstrated a potential for enhancing Both academic (e_g_, Camegie.Me110n, Hong Kong, M11;
the Pedagogical elrPerienee- One eXarnPle is the Pregrarnrned Nijmegen, KTH Stockholm) and commercial (e.g., Carn-
Logic for Automatic Teaching Operations (PLATO) Project egie speech, Duofingo, Rosetta stone”, SR1, Transparent
(UniVer5itY 0r lllin°i5 at Urhana‘CharnPaign)> Which inCl11d' Language‘) teams have developed technology that evalu-
ed °nline te5tihg> tutoring: and Chat r°°rn5- ates pronunciation using methods adopted from ASR. At
Over the years, the quality of CALL improved, driven by hrst glances ASR aPPear5 a Perfect rnateh r°r CALL ln Plaee
advances in interactive media and technology (Warschauer ‘it a language teaeher> WhY h°t leave the tetliilrn °r tiitorihg
and Healy’ l998)_ lh the 19605 and 19705) CALL foeused on to an algorithm embedded in the cloud? It’s available 24/7,
drill and practice lessons in which a computer presented a hever tires °r 5ieken5> and ‘l°e5nt gt’ °h Vaeati°n- However:
Stimulus and the Student responded with (hopefully) the ASR-based CALL has its drawbacks. For one, ASR doesn’t
correct response. This was the “structural” (or “restricted”) elaS5itY individual 5Peeeh 5°ilh‘l5 with great Precision (e-g-s
phase of CALL. Beginning in the late 1970s and extending Greehherg and Chang 20O0)- Like humans: ailternatie 5Y5‘
through the early 1990s, CALL entered its “communicative” terns ‘l°ht tleeetle 5Peeeh Stiilhtl hY 5°iln‘l but rather relY ‘in
Phase) which emphasized more natural ways of Speaklhg clever engineering to infer what the speaker said (or should
and listening. have said). They do so by culling information from a variety
f t‘ . .,l t’ , ‘l, l’ h-
with me  owe worm we we; and émiffmedia :.;::‘:::::.:::::‘:;:S::.%...;C:.;::..e::;.:;‘.“;:.:::::..
technology in the 1990s, CALL entered its integrative phase, . , , . .
_ _ _ _ for conventional ASR (e.g., Amazons Alexa, Apples Siri, and
in which the pedagogy was incorporated into a broad range of . . .
_ _ _ _ _ _ _ Google Voice), such supplementation can be a serious draw-
communication scenarios representative of daily life. During . . . . .

_ _ _ _ _ _ _ back for CALL applications. This is due to the uncertainty
this time, CALL applications offered graphics, animation, au- d. th .d . f .fi h d k
dio and text all in lessons that combined s eakin listenin Surmun mg 6 I entity 0 Specl C Speec Soun S (a ia-

élt d’ _ _ Ch 11 d S 2%” g’ g’ “phonetic segments” or “phones”). To better understand the
tea mg’ an wmmg ( ape e an aum’ )' problem, let’s consider a hypothetical example. The word
The key to effective language learning is for the student to use “pan” consists of three phonetic segments represented by
the foreign language as much as possible. Constant practice the symbols [p], [ae], and [n] (brackets denote individual
and feedback is essential. A shortage of language instruc- segments). An ASR system might correctly identify the ini-
tors and classroom time makes a compelling case for CALL tial and final consonants ([p] and [n]) but misidentify the
because it offers instruction anytime, anywhere. Although vowel [ae], in which case the word initially “recognized” is
CALL was originally designed for desktop and laptop com- “pin” rather than (the spoken word) “pan.” However, the
puters, its future likely lies with smartphones, tablets, and vowel’s misclassification would probably be overridden by a
an | Acuuseics Thday | Winter 2018

   20   21   22   23   24