Once a text has been annotated by our NLP-based tools and a corresponding termbase established, any part of the text can be transformed into a series of exercises, designed by didactic experts to help learners commit new knowledge to long-term memory.
Technological advances in recent years enable an unprecedented level of personalization. Your customers now also expect personalization from digital learning products, i.e.:
- that they can create their own custom curriculum
- that information is available to them at the tap of a finger
- that immediate feedback is given once they submit any kind of request
L-Pub enables content owners to transform their texts into materials for personalized learning – in an NLP-driven, standardized process that is cost-effective and delivers high value to your customers.
All that is needed is a reading text (e.g. an article, book or website) and a termbase (definitions of terms, descriptive images or other reference content). L-Pub then performs a mash up of your text and the termbase to create of a series of automatically generated exercises.
The learner then simply has to choose a word from your termbase which they would like to commit to memory and a sentence from your text to use for the exercises. The exact nature of these exercises will depend on the learning topic, but they typically consist of two types:
- Recognition: The first set of exercises begin with multiple choice questions based on the selected term, including so-called distractors and other words they are learning.
- Recollection: The more advanced exercises require the learner to fully retrieve the word from memory (without a choice of possible answers), by showing them only the definition or the sentence with the word omitted.
Since the right answer is stored in the termbase, we can always give the learner immediate feedback on how they are doing and also gather statistics on their progress, which they and/or an instructor can access at any time.
The automatically generated exercises used by L-Pub were developed in connection with our first tool for language acquisition, the vocabulary trainer app vobot.me. Since everyone in the development team was a language learner and two were didactic experts, we knew exactly what kind of exercises we needed to create.
For vobot, we were working with the Hueber Learner’s Dictionary German-English / English-German. Since it includes a database of nearly 100,000 sample sentences, the challenge was creating them on-the-fly, i.e. not programming them each individually by hand. Based on natural language processing and our own specialized annotation technique, we were able to develop a robust and reliable algorithm after almost 3 years of work.
Although we began with language learning, we soon recognized that our algorithm could easily be applied to any text-based learning scenario.
We are constantly improving this algorithm as a part of our on-going software development. In collaboration with TU Darmstadt, we have received a grant to bring considerably deeper sophistication to these exercises by 2019. The research includes the use of machine learning to precisely and dynamically adapt the exercises to a learner’s progress.
At the moment, the sequence of the exercises appears random to the learner. In reality, a number of criteria are already factored in to determine exactly when an exercise appears:
- The total number of exercises in the learner’s list
- The last time they successfully or unsuccessfully completed the exercise
- If they did all levels of an exercise in a short period of time (mass repetition usually does not result in knowledge being committed to long-term memory, so a resurfacing of the exercise after a certain period of time would then be scheduled)
- On request, we can change the algorithm to support a customer’s preferred repetition method (Leitner, graduated interval, cued recall etc.)
Further methods will be integrated over time, but already today we are proud to have such a robust algorithm that stands out from other learning methods in three key ways:
- All exercises are context-based
- Learners actively choose what they want to learn, so both boredom and overloading are avoided
- Our software immediately gives feedback on a learner’s answers, so they objectively know at all times how they are doing
Does personalized learning really work?
There has been significant research into discovery learning and its efficiency in a wide range of educational scenarios. It is only challenged when it is not accompanied by any form of guided learning whatsoever. Personalized learning, in our eyes, should not replace traditional instruction, but be used in tandem with it, to maximize learner motivation and encourage continued learning outside of the classroom.
L-Pub distinguishes between two main types of personalization, both of which we aim to pursue in all our products:
- Choice of learning content
- Choice of learning context
The first, choice of learning content, means the learner is control of what they are learning, at least with regards to:
- when to enter the learning mode
- where to dive deeper
- what to focus on next
- how much time to spend on each exercise
What about curriculum requirements by the school district or other educational authorities? Choosing the learning content does not necessarily mean students can skip required materials, it simply means they can define the sequence and manner in which they work through the tasks.
Choice of learning context is a criterion we believe is vital to learner motivation and success. For instance, if a teacher wants to cover a generic topic like prepositional verbs, they could let each learner choose their own context, e.g. one learner may choose a music magazine, another may choose a graphic novel, yet another may choose a book about gardening. An initial task may be for the learners to identify the prepositional verbs in their chosen context. If these texts were also pre-annotated by L-Pub, the learner could also practice those verbs in automatically generated exercises, all in their chosen context.
Study on the effectiveness of adaptive learning
Our research partners at the Technical University of Darmstadt are currently conducting a study on the effectiveness of adaptive learning for learners of German as a foreign language. If you, or anyone you know, is currently learning German, we encourage you to participate in the study! It requires 2-3 hours of your time every week over the course of one or two months. Learn more about the study and how to participate here.
Advantages of automatically generated instructional materials
- Any text can be annotated and transformed into instructional materials
- Saves significant time in the generation of instructional materials
- Provides considerable value-add to customers incl. functions like automated feedback to learners and statistics on learning progress for instructors
- Can be integrated into existing LMS (learning management system)
- Easy assignment of tasks and monitoring of learner progress
- Automatic correction of learner classwork or homework
- Follow their own curiosity
- Personalized learning
- Constant feedback on learning progress
L-Pub can deliver all of the above features in products which work both online and offline on various devices.
You may wish to have a customized version of our products or have our software implemented in your own ecosystem. In this case, we are happy to provide you with a white label or SaaS solution.
Get a quote
Are you interested in integrating automatically generated instructional materials in your digital offer? Whether via an API, encapsulating our code in yours or creating a re-branded version of our existing tools, get in touch so we can start the conversation!