On Language Acquisition and Pedagogy

Lee Kezar, Class of 2019. This reflection-based blog is designed for the Language Acquisition and Pedagogy (LAP) course at Rhodes College. Questions, comments: email kezal-19 (at) rhodes (dot) edu

[kezarlap]

Shadowing a Language Teacher

For my assignment involving shadowing a language teacher, I chose to review an ASL teacher who I used to have, named Sheila Chappell. The class was ASL 1, which met once a week for two hours at night. I shadowed her several times, only intervening when she asked me a question.

The typical class looked like this:

Reviewing last week's vocabulary

New signs and practice

15 minute break

30-item receptive quiz

Some kind of game or activity involving ASL production

I noticed that the students were most engaged for the reviewing session and the game, but less engaged during the new signs portion (which usually took upwards of 20 minutes because of all the nuance behind each word).

Sheila’s sense of humor and focus on individual students made the class very productive, in my opinion. While I wish there was more exposure to conversation, I must acknowledge that this is level 1 and the students probably wouldn’t understand much. Additionally, as a Deaf woman who exclusively uses ASL, the students got plenty of exposure by just trying to keep up with what she was saying.

One other critique is that when practicing new signs, she would frequently ask students from across the room to pair up and repeat a fill-in-the-blank kind of sentence (e.g. “My name is __, it’s nice to meet you”). This classic method of repeating (reminiscent of audiolingual, except there is no audio) is notorious for being dry and disengaging, especially since she went one by one.

Lastly, another strength I noticed was frequent positive feedback. An emphatic “RIGHT!!” or the sign for APPLAUSE/YAY would ensue after a student demonstrated exemplary work, such as signing a personally challenging sentence, or going beyond what she expected. On the other hand (pun not intended), minor mistakes were usually ignored, and severe mistakes warranted a correction and request to try again. I think this strategy for correction works well, but could use a little more individuation. For example, she could pick sentences which the student would find personally challenging, perhaps making the tedious process of practicing more interesting.

[kezarlap]

Research in Affective Computing

In my previous post, I discussed Affective Computing and its potential to fill gaps caused by Autism Spectrum Disorder. I also briefly nodded to my planned work in this field, without going into any detail. I’d like to expand on this a little bit, since it has strong connections to Language Acquisition and Pedagogy.

My research project will produce an artificial intelligence which produces emotionally-charged sentences. However, unlike prior research, this will be able to produce multiple emotions simultaneously, closer emulating human affect. In a sense, it will acquire a more precise language of emotion.

“Acquire is rather contrived,” you might say. And I would agree, computers don’t really learn the same way we do. They learn by guessing the right answer, comparing it to the actual answer, then automatically tweak itself to do better next time. This is called the training phase of an unsupervised learning algorithm. What makes this possible is the essential component of having thousands if not millions of training samples to throw at it.

And as it so happens, we have this training sample for affect. It’s called “Depeche Mood”. You can see it in action here — and it’s quite fun to play with. I will use this database to go the opposite direction: from emotion to words.

So they don’t quire acquire the “language of emotion” the same way we do, but they achieve a decent level of accuracy which makes them potent. I have said before how affecting computing is useful for helping people with autism, namely by identifying emotions explicitly when implicit understanding is lacking. But I am also interested in seeing how an AI which can produce emotions might be useful in conversational settings. I think regardless of ASD, a computer which is more compassionate and understanding is both exciting and useful.

There are dozens of applications for an emotionally intelligence computer that come to mind, so I can’t begin to do this idea justice in this post, so I will intentionally leave it open for your mind to contemplate. To help, here is the trailer for a science fiction film named Her which explores a specific application of this technology a few years into the future:

[kezarlap]

Connection: ASD, ASL, and the Language of Emotion

Today in class, we heard a student presentation on autism spectrum disorder (ASD) and language. As someone stated rather concisely, autism is known for fostering great strength in some areas, and impediments in other areas. Language is something that falls into both categories, as when someone with autism is experiencing a sensory overload and loses their ability to communicate, or on the other hand, finds language learning very easy and enjoyable.

We discussed ASL and its utility when these oral breakdowns occur, assuming the person is also able to retain motor control in their hands and that the audience also knows ASL (I’m not sure which is a bigger stretch). ASL benefits from simpler grammar than English, and therefore might be less cognitively taxing. In more severe cases of communicative disorders caused by ASD, ASL might be the only available option.

Another connection we made is to emotion, and how some people with autism have difficulty understanding how other people feel from body language, leading to a different and sometimes problematic perception of what is social acceptable. For example, one person in the audience recalled a student on the spectrum who didn’t quite seem to understand looking at a  watch repeatedly meant “I’m in a hurry and need to go”.

Computer scientist Rosalind Picard, whose son has autism, decided to dedicate a large part of her career to assisting people with autism in their differences of ability. She called this field Affective Computing, for its focus on emotional processing. The video below briefly describes some highlights in her work.

How does this pertain to language acquisition and pedagogy? Well, I believe that language is far more than strings of morphemes. They include body language and the language of emotions, things which are typically not explicitly taught yet widely understood. I’d like to continue Picard’s work by integrating Natural Language Processing (described in the previous post) into Affective Computing. Tomorrow I will receive a $4,500 grant to do exactly this research, and I look forward to connecting it to what I have learned in this presentation and course.

[kezarlap]

Connection: LAP and Natural Language Processing

The ability for computers to learn language is remarkable. Using an algorithm which closely mirrors the human brain (called an artificial neural network), an AI can be given lots of examples of natural language, and slowly improve its accuracy.

A basic example is learning the past tense of regular and irregular English verbs. This algorithm (called FOIDL) quickly learns the regular verbs (e.g. walk becoming walked, color becomes colored), then overgeneralizes it to irregular verbs (e.g. drive becoming drived), producing a small dip in accuracy around 100 samples.

Perhaps what is most interesting about this Natural Language Processing algorithm is how closely it mirrors the U-shape curve demonstrated by children acquiring languages, who also overgeneralize when learning things such as past tense forms of verbs.

I am interested in using what I have learned in LAP as I study Natural Language Processing. We have provided many valid critiques of technology in language learning (e.g. lack of humanity in a human-centric field), but I think that they are all surmountable with advanced artificial intelligence approaches. For example, it is not unimaginable that an AI will learn individual preferences and maximize comfort, how to speak articulately and listen intently to each mistake, etc. In fact, I think that once AI can accomplish these things, they will be more useful to students than any teacher could.

[kezarlap]

Research in ASL Pedagogy

The above image is the poster I made for my research in Deaf Studies. It is interdisciplinary, drawing on aspects of Deaf Studies, Linguistics, Educational Studies, and Sociology. Its focus is the mixing of ASL and English, a phenomenon provoked by ASL instructors and students with damaging consequences.

Full sized image

[kezarlap]

Connection: LAP and Deaf Studies

A little over a year ago, I was notified of my acceptance into the Deaf Studies Fellowship, which overjoyed me. I have been immersed in Deaf culture and ASL since the summer of 2016, in part due to an initial spark (dating an interpreter) but also due to some inherent fascination within me. The fellowship was, to me, an opportunity to transform my deep-rooted fascination into a tangible work, a project.

Through the Deaf Studies Fellowship, I was able to practice what I learned in LAP. For one, we had an ASL class every Wednesday, but on Mondays and Saturdays the tables turned, and I taught ASL to kids (pictured) and other college students. The latter experience – teaching ASL – was particularly enjoyable, because the topics covered in LAP translated directly into my teaching. For example, with my two Monday students (who are also peers of mine), I emphasized conversation as much as possible. We read children’s books, watched song interpretations, and had conversations with each other. Pedagogical research in ASL recommends avoiding SimCom (the simultaneous communication of ASL and English) because it is cognitively taxing for the teacher and therefore leads to mistakes, I thought it was a great tool at the basic level to associate ASL signs with English words.

A different part of the fellowship, I was also able to review scholarly research in ASL acquisition and pedagogy. Among the approximately 10 articles I annotated, many of them discussed ASL acquisition among native English speakers. One discussed how ASL learners tend to fill knowledge gaps with English, creating a mixture of ASL and English. Another discussed the “oralism” approach, where deaf students are taught exclusively in spoken and written English, and how it robs deaf children of strong linguistic skills.

[kezarlap]

Language Center: Re-envisioned

Vision Statement: A socially oriented, dynamic, learning space for student, educators, and community partners interesting learning language and culture. Items 

Fundraise for language center remodeling

Remodel for kitchenette in entry room

Purchase comfortable furniture, stray away from desks

Remove computers: personal laptops are quickly replacing them

Commission community artists representing a variety of cultures for wall murals (contrast to blank walls -- cold and unwelcoming)

Reach out to community partners about offering cultural-based classes

Concerns

Fire hazard from kitchenette

Increased chance for stains

Cost/fundraising

Budget:

Enhanced support for clubs and extracurriculars -- $1000/year

Couches -- $2,000

Kitchen(ette) -- $10,000

Immigrant education -- $0

Art -- $1000

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Thursday, March 1st, 2018.

[kezarlap]

This row of computers in the Paul Barret, Jr. Library is, in my opinion, the worst space to learn a language, because it minimizes classmate interaction while also being in a highly public and distracting atmosphere. While I believe computers are quickly improving and flexible, human interaction remains the best way to learn.

[kezarlap]

The East Village Common Room is my ideal space to learn a language, because it has the necessary technology to share media (a projector, screen, TV, and audio system) but focuses on comfort and human interaction, reducing the affective filter and encouraging conversation.

[kezarlap]

Most Useful Teaching Approach

I find the communicative approach to be the most useful. Unlike Krashen’s “i+1”, this approach builds many things simultaneously, and aligns closely with Ortega’s model of practice. The book mentions that we ought to practice what we want to learn. So... practice communicating! Casual conversation, question-answering/asking, describing: these are all things that we use on a daily basis, and need to be practiced. If we break them down into pieces (e.g. lessons on these features distinctively, or a step further, by individual pieces of these features), we risk losing their context and what it truly means to communicate.

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Tuesday, February 20th, 2018.

[kezarlap]

Point of Interest No. 2

In our discussion with Professor Wanske, we talked about Middlebury Language Schools at length. I found this program to be very interesting, as it institutionalizes (and commercializes) the immersion method. I traditionally thought of immersion as travelling to a country and embracing the language and culture. This, however, only requires a plane ticket to Vermont.

One particularly interesting feature of this program was the Language Pledge, a promise given sharp teeth by its punishment upon breaking: expulsion from the program. It does give me anxiety just thinking about it (and is a notable difference from simply travelling), and I wonder if this helps more than it harms, but does ensure maximum immersion.

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Thursday, February 8th, 2018.

[kezarlap]

Most Useful Theory

In my opinion, Ortega’s model for practice is both very useful and valid, perhaps the most so. It aligns with psychological research, concise, and translates easily into pedagogical practice. It says three things: practice should be integrative, meaningful, and focused on task-essential forms.

I believe this theory drives the crowned immersion method, but can scale to the level of a classroom which meets only a few hours per week. It also promises fun and responsiveness, two things I mentioned in my beliefs of LAP.

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Tuesday, February 6th, 2018.

[kezarlap]

Point of Interest No. 1

Among the many things that LAP has taught me, perhaps the most surprising and intriguing is the concept of an affective filter (p 106). As a computer scientist, I frequently neglect affect in favor of straightforward mechanics and ensuing methodology. This concept adds emotion, and asks that the classroom be not only well-structured (i.e. uses effective models) but also human and comfortable.

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Tuesday, January 30th, 2018.

[kezarlap]

Reflection: Beliefs about LAP

I believe that languages are much like other people: complex, flexible, and ever-changing. As such, I am strongly attracted to teaching models that acknowledge these things. I draw inspiration from how friends and lovers form: not only frequent interaction, but mutuality, responsiveness, and fun.

This blog is part of the Language Acquisition and Pedagogy Course at Rhodes College. This post is my submission for the assignment due Tuesday, January 30th, 2018.