With language being at the heart of many teaching interactions, the ability to analyse classroom exchanges and school documents provides large banks of data that can tailor the patterns for progress across educational institutions.
How does technology deliver this uniquely, and what can we expect to achieve?
Natural language processing (NLP) is the branch of artificial intelligence concerned with computing the meaning of text and voice data using linguistic rules, statistical models and machine learning. Conversely, NLG generates outputs that seem to be spoken or written by a human from data. Together, these technologies allow computers to process and generate human language with nuance for context, intent and sentiment.
Chances are, you have interacted with NLP. After all, computer programs driven by NLP are behind the rapid response of Alexa or Siri prompted for the weather, voice-operated GPS in cars, language convertors, and chatbots.
So what does NLP bring to the table when enhancing education?
While a human certainly could derive conclusions from observing classrooms or documenting trends across school records, their data is necessarily less complete than automation that could consistently capture all lessons, and process transcripts or generate documents at a much higher efficiency without fatigue.
Additionally, when data is analysed through more objective methods, unique insights can be gleaned that may have otherwise been overlooked, giving us new perspectives for effective approaches in education and highlighting the most impactful practices at present.
NLP in education has focused on developing educational software and identifying strategies for improvement. Towards the latter, scholars use NLP to mine educationally relevant language data to provide an empirical basis for system design through large scale, text-as-data methods. Software applications were broader. Primarily with teachers in mind, NLP has been used to try to automate tasks that traditionally have required manual effort, such as creating curriculum or assessment materials, whereas software for formative language assessment and for tutoring sought to support students’ progress too.
Below are example use cases of NLP in education, showcasing its far-reaching potential.
Education Systems is an in-house project for combining all the aspects of running a school under one school management information system (MIS). A chatbot integrated in the application makes it easier for teachers and school leaders to quickly find the information they desire.
Since teaching style and involvement can be heavily dictated by the activities of a particular lesson, the conventional approach of a human evaluator visiting a lesson not only limits the information available about typical practice, but also places a lot of pressure on individual teachers undergoing observation.
The intent behind this idea was to provide consistent, ongoing feedback to educators instead. Can it capture features of teaching aligned with human raters’ observations though? It seems promising, as the study found that the measures of teacher practices developed using NLP showed alignment with traditional observation protocols.
This case demonstrates how NLP methods can provide insights into students’ learning processes. With online courses increasing, it seems pertinent to learn ways engagement is enhanced remotely.
Analysing language data from student exchanges revealed how valuable exposure to more interactive peers was for students less likely to engage in online discussion— going as far as to increase their probabilities of passing the course and enrolling in the next term. It also helped identify groups inclined to interaction, and conditions facilitating the participation of individuals.
Hiring is an opportunity for schools to find educators with values that align with their goals. Yet beliefs are difficult to measure, prompting the exploration of NLP for providing information for hiring. When structural topic modelling (STM) was used with administrative data from thousands of applications to a school district in California to decipher the underlying beliefs of applicants by analysing essays they had submitted, distinct themes and understandings of inequities had emerged.
Interventions can be complex and multidirectional, which makes it difficult to evaluate how their outcome was achieved and thus decide the aspects of programs to retain or terminate. In a bid to shed light on those mechanisms, researchers analysed thousands of school improvement planning and implementation reports. They identified 15 coherent reform strategies, including ones significantly associated with reductions in chronic absenteeism and improvements in student achievement.
While improvements in tuning models and results extraction mean we are starting to see more large-scale NLP deployments, to democratise AI and NLP, applications need to be accessible to domain experts— in this case teachers and other school staff— rather than being confined to workers with technical knowledge of data science and coding. Low to no-code platforms will put the ability to create and use educational systems applying NLP in the hands of the people most experienced navigating that sector or specific school.
Secondly, more research is needed to determine how teachers and principals perceive automated measures, but it seems intuitive that people are likelier to embrace systems they understand and even informed the creation of, so no code platforms may reduce negative perceptions of automation in schools. This is important, because even if senior leaders are keen to rollout new technology, they have to consider resistance amongst staff and ensure no one feels left behind.
Besides that, multiple modes of technology is necessary sometimes to analyse documents using NLP as relevant information may be displayed in images, calling for computer vision. Intelligent automation platforms that make it easier to include integrations can be helpful here.
At the same time, NLP is shaping intelligent automation too. With language being humans natural mode of communication, computer interfaces capable of understanding natural language are more powerful and user-friendly, suggesting that such interfaces are the next step as human-computer interaction evolves.
NLP in the education sector can help identify the best teaching and learning practices by driving analyses of linguistic data at a greater, inimitable scale, and allowing for new perspectives. It can also be applied in human-facing interfaces, such as school management apps, to return information quickly via chatbots, or to help save teachers’ time by powering assessment software and generating common documents.
Future work to promote wider adoption of this technology in schools include increasing their accessibility to teachers via no-code platforms and enacting the setup of relevant recording and transcribing systems for automated data collection, and equipping NLP programs with the ability to generalise to better handle different age groups and subjects.
Would you like to exchange ideas with us on the subject of digital transformation and process automation without obligation? Let’s talk!