Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

AI-based chatterbots and spoken English teaching: a critical analysis

No description
by

Chen Te-Hai

on 13 September 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of AI-based chatterbots and spoken English teaching: a critical analysis

animated
facial expressions
, such as blinking of the eyes, smile, anger, idle, etc.
Natural Communication
AI-based chatterbots and spoken English teaching: a critical analysis
Computer Assisted Language Learning
Vol. 22, No. 3, July 2009, 269–281

A presentation by
Ted
Guoquan Sha
Chatterbots based on dialogue management and developed with
AI
,
TTS technology (text to speech)
, and
computer animation
can engage the learner in
computer-simulated human–human
conversations.
Its
natural

language social interaction
opens new horizons for
spoken
language teaching/learning.
A classroom survey
was conducted in which the students showed
keen interest
and computer-simulated human–human interaction was frequently observed.
Abstract
The dilemma of spoken language instruction
Multimedia environments and CALL can
effectively promote
students’ abilities such as
listening
comprehension and
reading
comprehension but they are rather
limited
in their effectiveness when dealing with
output abilities
such as
speaking
and
writing.
‘‘A key feature of
natural communication
, which is seen as a characteristic of communicative activity is that it involves
the production of a message
in real time’’ (K. Johnson & H. Johnson, 1999).
Real situation
In the classroom where
CLT
is adopted,
the native speaker

dominates
the classroom, broadcasting news or stories and the majority of students remain
silent
.
When the native speaker spends all his time among the students, talking with them one after another, each student shares a very
limited time speaking
.
Other factors that may inhibit natural communication:
class sizes
being too
large
students feeling too
shy

students being too concerned about the requirements for passing a
written exam
to take spoken language seriously
Pair work is another common practice, through which the teacher tries to motivate the student to talk.
1.the teacher turns himself into a
monitor
or
supervisor
and sometimes simply watches the students practicing
2.
cross-cultural communication

no longer exists
between two non-native speakers
3. a lack of
authentic sources
of the language
What chatterbots can do
The fastest way to learn a foreign language, is by living with a native speaker.
Can a computer act as a human being and live with the learner of a foreign language?
The chatterbot is one answer.
Highlights of chatterbots
chatterbots are computer programs designed to simulate human conversations in
voice
or
text form
, or
both
They have been applied in specific domains such as
customer services
or
FAQ
(frequently asked questions)
services
in business.
The experiment discussed in this paper mainly involves
Verbot 4.1
(see Figure 1),
Ultra Hal Assistant 6.0
(Ultra Hal), and
Jabberwacky
, among which Verbot is the most valuable for spoken language teaching.
Verbot and Ultra Hal are characterized by
three
up-to-date technologies
1. The knowledge base (the brain)
compiled in
artificial intelligence markup language (AIML)
or other similar programming languages, it is the main and indispensable part of every chatterbot
Verbot
adopts the algorithm of
pattern matching
and
Ultra Hal keyword spotting
For example, all the wh-questions constitute a
pattern
and all the questions beginning with ‘‘What is/are/do . . .’’ constitute a subpattern respectively.
The
patterns
and
subpatterns
(or rules and child rules, as in Verbot) form a tree-like structure called
a decision tree
(see Figure 2).
the conversational partner is actually
not the machine
– which can hardly generate responses on its own –
but the human
who writes into the knowledge base based on naturally predictable inputs.
When the
keywords
are recognized, the stored responses are
retrieved
and put out.
2. Text to speech (TTS) and speech recognition engines (SRE)
an
optional
component of the chatterbot but
extremely necessary
if
voice chat
instead of text chat, is to be implemented in language teaching
Natural speech is produced by selecting segments from a
real
,
pre-recorded human speech database
. This means, however, a separate
large database
for each male or female voice has to be installed.
3. Agent characters
The most important
animation is the
lip-sync
, which makes them look like life-like humans when speaking.
Speech recognition engines (SRE) will engage the learner in interactions with the chatterbot most naturally by
listening
and
speaking
.
However, problems may arise when the learner makes
too many hesitation pauses
, factors including
pronunciation
,
acoustic environment
,
speaking rate
, and
training of the recognition engine
.
Text chat
, also known as
keyboard-based chat
, can be a better way for the learner is rather
relaxed
,
free of speed constraints
and he can
check
what he is going to say before he presses the enter key.
Ultra Hal Assistant
http://www.zabaware.com/assistant/
http://www.jabberwacky.com/
Jabberwacky
Evaluation of chatterbots
It needs to be pointed out that, unfortunately, most chatterbots are
not aimed
at
language teaching
.
A number of the Ultra Hal’s responses seem
irrelevant
to the input questions
except that they contain the keywords
, which have been identified in the input questions.
Chatterbots can be
perfect
at
providing general knowledge
, which concerns one of the major functions of language:
the informative function
(Hawes & Thomas, 1994).
Chatterbots are at their
best
when the dialogue runs in a
single question-and-answer sequence
. In the same way, chatterbots are not supposed to ‘‘understand’’ a joke or story told by the learner and therefore often fail to give relevant responses.
A classroom survey
Voluntary sampling was performed:
15
subjects self-selected from approximately 150 college students (three classes of first and second graders) including
two males
and
13 females
, half of the total participants having experience of talking with a native
Table 1. Characters chosen by the subjects (%).
Julia Melissa Voyager Genie Ka
53 20 13 7 7
47%
saying that the character was ‘‘
prettier than others
’’
27%
stated that they
didn’t mind the appearances
of the characters
7%
agreed that the most impressive aspect of Verbot was the character’s
facial expressions
and
33%
noted its
beautiful look
Generally, all the students showed
strong interest
in the human-chatterbot chat,
13%
feeling ‘‘
amused and started laughing
’’ and
80%
feeling ‘
‘interested
’’. No one (0%) felt bored or nervous.
To reveal how much they were interested, a further question was asked: ‘‘
If you had a computer of your own, would you like to have Verbot at home to chat with?
’’
73%
said ‘‘
yes
’’ and 20% ‘‘no’’.
To investigate the effectiveness of communication, the following question was asked: ‘‘
How many of your questions does Verbot give good or correct answers to?
’’
No one (0%)
said that all of the questions asked had returned
the right answers
,
33%
admitting ‘‘
most of my questions
’’,
49%
saying ‘‘
only a few of my questions
’’ and
27% being sometimes not sure what the chatterbot meant
.
The most divergent responses were given to the question ‘‘
Do you find any technical problems that make it difficult to communicate with Verbot?
’’
The survey showed that
typing error (20%)
,
spelling error (13%)
and
unnaturalness of Verbot’s voice (27%)
had caused difficulty.
Furthermore,
20%
had problems with the
vocabulary
that Verbot used.
A solution might depend on the
in-depth study
of the
students’ favorite topics, setup of a knowledge base
that covers the whole range of the FAQs.
It can be concluded from the survey that
the knowledge base is critical
to the well-being of the chatterbot.
The survey also shows
the importance
of the role played by the
realistic animations
of an agent character.
Correct pronunciation
, which can be developed by using special software,
is a must
before the shift is made from text chat to voice chat.
Chatterbot-based language learning is not a cure-all methodology.
Like many other computer programs, chatterbots are still at
an early stage
in evolution and await further development.
Conclusion
Full transcript