THE LEARNING CURVE

​A MAGAZINE DEVOTED TO GAINING SKILLS AND KNOWLEDGE

THE LEARNING AGENCY LAB’S LEARNING CURVE COVERS THE FIELD OF EDUCATION AND THE SCIENCE OF LEARNING. READ ABOUT METACOGNITIVE THINKING OR DISCOVER HOW TO LEARN BETTER THROUGH OUR ARTICLES, MANY OF WHICH HAVE INSIGHTS FROM EXPERTS WHO STUDY LEARNING. 

Natural Language Processing In Education

​After all, there is already evidence that NLP is capable of parsing and summarizing arguments, helping writers improve their prose, and motivating writers to revise essays.

But what can we do to improve NLP in the classroom and bring the technology to students who can benefit the most from it?  What are the most powerful potential NLP applications in education?

It might be better to start by asking: What is not needed in NLP?

NLP has successfully and typically been used in educational settings to identify problems in student grammar and mechanics and to provide holistic scores for five-paragraph essays. While these areas may be important, NLP has the potential to address more significant problems that learners and teachers face in the classroom.

Natural Language Processing For Reading and Writing

First of all, developments in NLP can help students learn to write better essays by providing formative feedback (i.e., actionable feedback on specific essay parts) that can be used during the revising process to improve more than just grammar and mechanics.

For instance, NLP can help identify the presence of absence of important discourse elements like claims, arguments, and evidence. In addition, NLP can provide feedback to learners about the organization of an essay. These NLP solutions can be combined into automatic writing evaluation (AWE) systems that can provide low level feedback (e.g., tips about vocabulary) or higher level feedback (e.g., advice about the cohesion of discourse).

Ultimately, this formative feedback mixed with holistic feedback will allow students to revise essays multiple times before final submission to teachers. Teachers will then receive essays that are better developed, especially at lower levels of text features allowing them to focus on elements of writing that are more difficult to assess through computational algorithms, including argumentation, style, and organization.

Beyond improving students’ language skills directly, NLP features can also be used to help educators better understand what is happening cognitively with their students.

Another limitation of NLP writing studies is that they have historically focused on independent, persuasive essays that require no background knowledge (i.e., five-paragraph essays written on general topics). However, these types of essays do not accurately reflect writing expectations in the classroom.

Thus, researchers have begun to expand the writing tasks to which NLP can be applied to include source-based essays where writers are expected to integrate external documents, narratives, summaries, and expository texts.

Beyond writing, NLP can also go a long way toward helping struggling readers in the classroom. NLP algorithms can provide automatic feedback to students about the strength of self-explanations and summaries of reading samples, both of which are key elements of reading comprehension. However, teachers often don’t have the time to provide students with detailed and individualized feedback on these tasks.

Newer readability formulas based on NLP can also help educators better match texts to students to ensure reading assignments are suitably challenging and productive. NLP readability formulas can calculate more accurate readability scores that outperform traditional formulas such as Flesch-Kincaid Grade Level.

These new metrics provide information about the complexity of language in terms of vocabulary, cohesion, and syntactic complexity. Not only are they better predictors of reading speed and comprehension, but they better match cognitive models of text processing. They also work for a variety of different readers and genres. NLP techniques are even being used in text simplification algorithms that can automatically modify texts to make them better fit with the reading skills of students.

Finally, NLP can be especially useful for English Language Learners (ELLs). By providing feedback on both form and function of language components, NLP can improve the rate and quality of language acquisition.

NLP can also allow ELLs more opportunities for practice in their second language outside the classroom. For instance, online language tutors can provide feedback to learners about grammatical, syntactic, and lexical errors as well proficiency assessment. NLP can also help assess proficiency level for ELLs and track their progress over time.​​

NLP can thus both improve the quality of instruction within individual assignments and help educators improve the learning environment more broadly.

Natural Language Processing For Learning Behavior and Motivation

Beyond improving students’ language skills directly, NLP features can also be used to help educators better understand what is happening cognitively with their students. By analyzing language use in the classroom NLP can help identify and predict students’ mental states during learning. Studies in this area are still nascent, and should be improved upon.

But recent research provides evidence that text features in students’ written and spoken production both in-person and online can be predictive of success in math and science domains. Automatically assessing mental states can provide teachers with a better understanding of how well their students are prepared to learn. This information may help teachers better manage the classroom, identify struggling students early on, and improve student learning.

NLP can also identify individual differences in learners. Recent research has focused on NLP predictors of students’ individual differences, including vocabulary knowledge, working memory, and identity.

NLP can also help monitor affective factors via monitoring tools that are important in learning (i.e., engagement and boredom), which could be used to signal the need for transitions and spaced practice. The basic idea is that the language students produce can be strong indicators of cognitive and knowledge-based skills, all of which are dynamic elements of learning that can affect success in the classroom. Stealth assessment of these skills may help teachers more effectively tailor instruction to individual needs.

Finally, it’s also worth noting that NLP can be used to study less traditional educational metrics like successful collaboration in the classroom. Indeed, researchers have started to apply social-network analysis approaches to language data to find patterns of collaboration among students in online discussion forums and within MOOCS.

These NLP approaches can identify semantic trends in discussion, discussion leaders, and the genesis of ideas, all of which can help teachers better understand learning in the classroom. Recognizing class leaders and key collaborators may be an effective means to develop classroom partnerships and peer mentoring to help ensure success for all learners.

NLP can thus both improve the quality of instruction within individual assignments and help educators improve the learning environment more broadly. It’s still early days, but the research progress so far suggests that NLP can have a profoundly positive impact on learning.

-Scott Crossley is a professor of Applied Linguistics and Learning Sciences at Georgia State University.

Leave a Comment

Your email address will not be published.