Philip Mackenzie is a master student at Nanjing University. Philip’s research involves applying NLP techniques to academic review
This is the second lecture of Coderbunker’s EdTech series. Our speakers, two linguistic experts and three NLP experts, shared their knowledge on human language learning and machine learning.
Tiffany Wingert, the Content Editor at EF Education First made her speech from a learner’s perspective. “Most of the apps give users reward once they accomplish certain tasks but this external motivation is short-term. You are not doing the task because you want to learn a human language. Only internal motivation is long-term,” Tiffany explained. She suggested that applications focus more on intrinsic motivation rather than gamification. She also pointed out that some apps such as Duolingo have not prepared learners with conversations. She expected that learning apps would dynamically adapt contents to each user in the future.
Maijaliisa Mickols, who has been working with digital learning for 5 years at EF Education First, shed light on some basic principles of second language acquisition and pedagogical approaches. She clarified several differences between first language acquisition and second language acquisition such as “conscious vs unconscious” and how adults rely on “patterns” when they are learning a second language. Another difficult of second language learning was that learners usually had limited access to their second language. In addition, she claimed that “Learning and acquisition are different terms. Acquisition is memorization while learning is to have something ingrained in your language system.”
Gregory Orton, the front-end web developer and product manager at EF Education First shared his insights on the technology side of human language learning. “Why do adults turn to apps to learn a second language? We hope technology could help us to improve feedbacks, to give production and generate stimulus to internal motivation.” Gregory also pointed out the that there was something missing, “We already have learning contents for learners, but what we don’t have is how we serve the contents to people.” Gregory said this could be solved by figuring out the roadmap between points of contents in the future.
Philip Mackenzie, who has a bachelor degree in psychology, is doing a research in applying NLP techniques to systematic review automation. He focused on introducing NLP (Natural Language Processing) and NLTK (Natural Language Toolkit). Phillip also explained how NLTK could assist both learners and teachers.
The last speaker, Giuseppe Tomasello, specializes in engineering, international management and NLP for human language models. He also attended the first session of EdTech. Giuseppe further explained some technical aspects of Philip’s speech. Word embedding, a set of techniques that map words or phrases from the vocabulary to vectors of real numbers, has wide applications on machine translation, text classification, sentiment analysis, spam detection, etc. He also showed us some tools that could be used in the human learning process, such as Word2vec, which “leads to much improved NLP as it understands the general semantic relationships”.
Resources from speakers
Don’t count, predict! A systematic comparison of context-counting vs. context-predicting semantic vectors (Useful paper to understand the differences and similarities among and between word2vex)
If you also want to share your insights about human language learning and machine learning, you are welcome to join our third session of EdTech!
Follow us on Wechat!
We just sent you an email. Please click the link in the email to confirm your subscription!
OKSubscriptions powered by Strikingly