2021/10/03 speaking test: NatTos




NatTos makes a paradigm shift in the speaking test from
  • - subjective to objective.
  • - experience-driven to data-driven.
  • - excrusive to sharable.
  • - ready made to creative.
  • - time-consuming to immediate.
  • - high cost to low cost.
  • - by humans to by AI.


もくじ
  1. Challenge
  2. Solution
  3. Learning Impact
  4. Return on Investment
  5. References

..[↓] 1
 1 Challengeもどる
もくじへ

  • To be able to speak English, you need someone to talk to.
  • 　
  • If you have someone to talk to, you can know immediately whether they understand you or not, and if they cannot understand you, they will tell you how to fix it.
　　
  • There are already many such tools, but some fatal problems have not been resolved.

  • Speed for students: Quickly feedback in line with brain cognition and working memory in about 3 seconds.
    • 　
  • Ease of use for teachers: Can create exercises based on each week's lesson content within a few minutes.
    • 　
  • Simple to understand: Anyone can easily figure out where mispronounced and how to improve them.

  • We are going to make a paradigm shift in scoring speaking tests from subjective and time-consuming by humans to objective and immediate by AI.
..[↑][↓] 2
 2 Solutionもどる
もくじへ

  • NatTos is a sophisticated tool for English teachers and learners. Its features are scoring and correcting the learner's speech instantly and completely automatically, which is only 3 seconds. Not only scoring, but correction is also very important. Because...

  • Learners can see
    1. what they were saying. So NatTos displays both what the speaker intended to and how it was heard.
    2. the red pen error marks.
  • This two-way approach is the first and finest in the world.
    • 　
  • See figure 1.
  • Try a sample drill fig. 3.
  • Teachers can
    1. easily create and share a web-based speaking drill like a paper test in minutes.
    2. view the learning history within 10 second, calculated and displayed by text mining. See fig. 4,
..[↑][↓] 3
 3 Learning Impactもどる
もくじへ

  • See figure 2.

  • 3 engines system of NatTos
    • is an integrated technology.
    • (1) Teachers find the speaking test easier.
    • (2) Students get 3-second readable responses.
    • (3) While testing, teachers receive statistically processed feedback.
    • 　
  • 3 seconds quick and readable responses
    • makes a student easily adjust his pronunciation with a quick.
    • 　
  • 3 cycles of NatTos
    • allows teachers seamless access to students' activities, feedback lecture and lesson planning.
    • 　
  • 2 stages of on and off NatTos, exercise and evidence-based feedback lecture,
    • enchants students' tuning power.
..[↑][↓] 4
 4 Return on Investmentもどる
もくじへ

  • NatTos was uploaded 2020/6 and used in 3 classes at Tohoku University successfully in the autumn.
    • 　
  • The teacher wrote task texts according to his lesson section and uploaded to NatTos.
    • 　
  • Students tried to speak task texts with great eagerness. When they mispronounced, they tried again and again, and about 40-70% of them cleared within five trials. See fig. 4.c.
    • 　
  • Over 30,000 utterances were stored and we found a Japanese learners' weak point "he/she error".
    • See fig. 1 and 4.a and 4.d.
  • AI may discover more weaknesses that haven't been pointed out so far.
..[↑] 5
 5 Referencesもどる
もくじへ

  1. Spring, Ryan; Tabuchi, Ryuji (2021). Assessing the Practicality of Using an Automatic Speech Recognition Tool to Teach English Pronunciation Online, Journal of English Teaching through Movies and Media Vol. 22, No. 2, (May 31, 2021), pp. 93-104. https://doi.org/10.16875/stem.2021.22.2.93
  2. 田淵龍二（TABUCHI Ryuji）・Ryan Spring (2021). 　「スピーキング指導可視化ツール開発と指導意思決定支援―AI自動赤ペン添削で弱点を図示する学習履歴マイニング / Speaking instruction visualization tool development and instruction decision support - Learning history mining that illustrates weaknesses with AI automatic red pen correction」　. 第26回人工知能学会 インタラクティブ情報アクセスと可視化マイニング研究会（The Japanese Society for Artificial Intelligence: Interactive Information Access and Visual Mining）. @online
  3. 田淵龍二（TABUCHI Ryuji）・Ryan Spring (2021). 　「スピーキング3秒採点と赤ペン添削システム開発とオンライン授業―人手による主観評価から機械（AI）による客観評価へ / Speaking 3 seconds scoring and red pen correction system development and online lessons - From subjective evaluation to objective evaluation by machine (AI)」　. 言語処理学会（The Association for Natural Language Processing）第27回年次大会(NLP2021) E2:テーマセッション：言語教育と言語処理の接点(1). @online