Thursday, November 19, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News

AI Agent Automates Complex Web Forms for Higher Education

November 18, 2026
in Science News
Courtney Benton
By Courtney Benton Scienmag Editorial Profile - Science and Technology Policy
Reading Time: 5 mins read
0
AI Agent Automates Complex Web Forms for Higher Education

AI Agent Automates Complex Web Forms for Higher Education

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

Researchers have developed a new artificial intelligence system designed to automate the completion of complex digital forms in higher education and administrative settings. The system, named KAFA, or Knowledge-Augmented Form-Filling Agent, integrates large language models with structured knowledge bases to handle the varied and often inconsistent layouts of online academic platforms. Published in the International Journal of Educational Technology in Higher Education, the study addresses a persistent bottleneck in university administration, where students and staff frequently must re-enter identical personal and professional data across multiple unrelated systems.

The primary challenge identified by the authors is the lack of unified standards in web form design. Institutional platforms often use different field names, validation rules, and dynamic interaction logic for the same type of information. The study identifies that rule-based tools struggle with synonymous field relationships, such as when one platform labels a field 'Email' while another uses 'Contact Address,' or when validation rules vary significantly. Traditional rule-based automation tools struggle with these variations because they rely on fixed templates and static page structures, which fail when websites are redesigned or when fields are generated dynamically based on user input.

To overcome these limitations, KAFA employs a multimodal personal knowledge base that combines three distinct data storage methods. Structured information, such as names and degrees, is stored in a MySQL relational database. Unstructured long-text content, including project descriptions and self-evaluations, is encoded into semantic vectors using Sentence-BERT and stored in a Milvus vector database. Additionally, a Neo4j graph database models the relationships between entities, such as linking a specific degree to a particular university or a job title to an organization. This tri-source approach allows the system to retrieve information based on exact matches, semantic similarity, or relational logic, depending on the nature of the form field.

The system utilizes the Chrome DevTools Protocol to parse the Document Object Model of web pages, extracting both structural and semantic attributes from form fields. By analyzing the context of each field, including labels, placeholders, and surrounding text, the DeepSeek large language model generates a semantic representation that can be matched against the knowledge base. This process enables the agent to understand the intent behind a form field, even when the terminology is non-standard or ambiguous. The system then selects the most appropriate action, such as text entry, dropdown selection, or file upload, to complete the field accurately.

A key innovation in KAFA is its tri-source collaborative retrieval and fusion strategy. Rather than querying all data sources simultaneously, the system uses a hierarchical triggering mechanism. For standardized fields, it first attempts an exact match in the relational database. If the field requires descriptive content, it activates vector retrieval to find semantically similar text. For fields involving entity relationships, it uses graph reasoning to traverse connections in the knowledge graph. If the results from these sources are ambiguous or conflicting, the large language model acts as an arbiter, fusing the candidates based on confidence scores and predefined trust hierarchies to determine the final value.

To evaluate the system, the researchers tested KAFA on 44 public recruitment websites, which were selected as a proxy for higher-education service forms due to their similar profile-intensive structures. The dataset comprised 1,301 unique form fields, ranging from simple text inputs to complex dynamic sections requiring multi-step interactions. The forms varied in complexity, with field counts ranging from 4 to 70 per form. This diverse test set allowed the team to assess the system’s performance across simple, medium, and complex scenarios, simulating the variety of administrative tasks encountered in university settings.

The experimental results demonstrated that KAFA significantly outperformed traditional rule-based baseline methods. While the baseline achieved high precision, its recall and accuracy dropped sharply on complex forms, falling to 45.52% accuracy. In contrast, KAFA maintained high performance across all complexity levels, achieving an overall F1-score of 95.58% and an accuracy of 91.66%. On simple forms, the system achieved perfect accuracy, while on medium-complexity forms, it maintained an accuracy of 92.48%. Even on the most complex forms, which involved intricate interaction logic and numerous fields, KAFA achieved an accuracy of 86.93%, representing a substantial improvement over the baseline.

Ablation studies further highlighted the importance of the specific components within the KAFA architecture. When the core large language model was swapped from DeepSeek-V2.5 to Qwen-7B, performance decreased, with DeepSeek showing superior recall and accuracy across all complexity levels. This suggests that the choice of language model significantly impacts the system’s ability to handle complex semantic understanding and logical reasoning. Additionally, the study compared the performance of KAFA with historical action guidance against a no-guidance strategy. The guided approach, which reuses successful action sequences from previous similar forms, improved recall by 58.2% on medium-complexity forms and 95.9% on complex forms, underscoring the value of trajectory reuse in enhancing efficiency and robustness.

The system also incorporates an adaptive error correction mechanism that responds to validation feedback from web pages. When an input triggers a validation error, such as an invalid phone number format or character limit exceeded, the system parses the error message to infer the implicit constraints. It then uses the large language model to generate a compliant input value, replacing the invalid entry and continuing the form-filling process. This reactive approach allows KAFA to handle dynamic validation rules that are not explicitly defined in the initial form structure, ensuring smooth completion of forms with strict input requirements.

Despite its strong performance, the study acknowledges several limitations. The evaluation was conducted on public recruitment forms rather than live institutional education-service platforms, meaning that further validation is required on real admission, registration, and student service systems. While recruitment forms share many profile-oriented fields with educational forms, cross-domain transfer may be affected by institution-specific terminology, authentication workflows, and policy constraints. The authors note that practical deployment in higher education would require robust privacy-preserving measures, including consent management, access control, and encrypted storage of sensitive student and staff data.

Future research will focus on deploying the system on actual educational platforms and enhancing its ability to handle highly complex dynamic forms with multi-page dependencies and third-party embedded widgets. KAFA is designed to automate and personalize administrative tasks, which the study suggests may enhance efficiency and user experience by reducing the need for repetitive data entry. By combining large language models with structured knowledge bases and intelligent retrieval strategies, the system offers a robust solution for navigating the heterogeneous digital landscapes of modern academic institutions.

The implications of this research extend beyond form filling to broader areas of educational technology and human-computer interaction. Online form completion for tasks such as university admissions, course registration, and institutional reporting requires efficient and accurate automation. KAFA combines CDP-based page parsing, tri-source knowledge retrieval, and historical trajectory reuse to improve semantic field-value alignment and operational robustness in form-filling workflows.

By leveraging the capabilities of large language models and multimodal knowledge bases, KAFA demonstrates the potential for AI to handle complex, real-world tasks that require both semantic reasoning and precise execution. The system’s ability to adapt to dynamic form structures and validation rules highlights the importance of combining advanced AI techniques with practical engineering solutions. Integrating large language models with structured knowledge and guided interaction strategies allows KAFA to automate administrative tasks, helping to enhance efficiency and user experience.

Subject of Research: Knowledge-augmented form-filling agent for higher education services

Article Title: Knowledge-augmented form-filling agent for higher education services

Article References: Sang, H., Liu, M., Tang, J., Yang, Y., & Chen, Y. (2026). Knowledge-augmented form-filling agent for higher education services. International Journal of Educational Technology in Higher Education, 23(1), Article 36. https://doi.org/10.1186/s41239-026-00612-x

Image Credits: AI Generated

DOI: 10.1186/s41239-026-00612-x

Keywords: Knowledge-augmented, form-filling, agent, higher, education, services

Cite Scienmag News

Courtney Benton. (November 18, 2026). AI Agent Automates Complex Web Forms for Higher Education. Scienmag. https://scienmag.com/ai-agent-automates-complex-web-forms-for-higher-education/

Courtney Benton. "AI Agent Automates Complex Web Forms for Higher Education." Scienmag, 18 November 2026, https://scienmag.com/ai-agent-automates-complex-web-forms-for-higher-education/. Accessed 19 November 2026.

Courtney Benton. "AI Agent Automates Complex Web Forms for Higher Education." Scienmag. November 18, 2026. https://scienmag.com/ai-agent-automates-complex-web-forms-for-higher-education/

Tags: agentEducationform-fillinghigherKnowledge-augmentedservices
Share26Tweet16
Previous Post

Artificial Intelligence Transforms Solid-State Impedance Matching for Radio-Frequency Power Supplies

Next Post

Interpretable Machine Learning Models Predict Deterioration in Steel Box-Girder Road Bridges

Related Posts

Narrow Margins Match Wide Resection in Liver Cancer After Conversion Therapy
Medicine

Narrow Margins Match Wide Resection in Liver Cancer After Conversion Therapy

November 19, 2026
Moral Diversity in Clinical Care Complicates Ethical Responsibility for Medical Artificial Intelligence
Medicine

Moral Diversity in Clinical Care Complicates Ethical Responsibility for Medical Artificial Intelligence

November 19, 2026
Mango and Soursop Leaf Extracts Lower Blood Glucose in Diabetic Rats
Medicine

Mango and Soursop Leaf Extracts Lower Blood Glucose in Diabetic Rats

November 19, 2026
Densely Temporal Causal Network Proves Competitive in Short-Range Weather Forecasting
Science News

Densely Temporal Causal Network Proves Competitive in Short-Range Weather Forecasting

November 19, 2026
Nepal School Study Shows Low-Cost Upgrades Greatly Reduce Earthquake Hazards
Earth Science

Nepal School Study Shows Low-Cost Upgrades Greatly Reduce Earthquake Hazards

November 19, 2026
Sobol and Morris Sensitivity Measures Best Identify Influential Pixels in Neural Networks
Technology and Engineering

Sobol and Morris Sensitivity Measures Best Identify Influential Pixels in Neural Networks

November 19, 2026
Next Post
Interpretable Machine Learning Models Predict Deterioration in Steel Box-Girder Road Bridges

Interpretable Machine Learning Models Predict Deterioration in Steel Box-Girder Road Bridges

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Narrow Margins Match Wide Resection in Liver Cancer After Conversion Therapy
  • Moral Diversity in Clinical Care Complicates Ethical Responsibility for Medical Artificial Intelligence
  • Mango and Soursop Leaf Extracts Lower Blood Glucose in Diabetic Rats
  • Densely Temporal Causal Network Proves Competitive in Short-Range Weather Forecasting

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading