A Contextual Help Browser Extension to Assist Digital Illiterate Internet Users

A Conte xtual Help Bro wser Extension to Assist Digital Illiterate Internet Users Christos K outsiaris School of Science and Computing South East T echnological Univ ersity Ireland github .com/unseen1980/acr o-helper Abstract —This paper describes the design, implementation, and ev aluation of a browser extension that provides contextual help to users who hover over technological acr onyms and abbr evi- ations on web pages. The extension combines a curated technical dictionary with OpenAI’s large language model (LLM) to deliver on-demand definitions through lightweight tooltip ov erlays. A dual-layer artificial intelligence (AI) pipeline, comprising Google Cloud’ s Natural Language Processing (NLP) taxonomy API and OpenAI’ s ChatGPT , classifies each visited page as technology- related bef ore activ ating the tooltip logic, thereby reducing false- positive detections. A mixed-methods study with 25 partici- pants evaluated the tool’s effect on reading comprehension and information-r etriev al time among users with low to intermediate digital literacy . Results show that 92% of participants reported impro ved understanding of technical terms, 96% confirmed time savings over manual web searches, and all participants found the tooltips non-disruptive. Dictionary-based definitions were appended in an average of 2,135 ms, compared to 16,429 ms for AI-generated definitions and a mean manual search time of 17,200 ms per acronym. The work demonstrates a practical, real- time approach to bridging the digital literacy gap and points toward extending contextual help to other domains such as medicine, law , and finance. Index T erms —Contextual Help, Digital Illiteracy, Br owser Ex- tension, Natural Language Pr ocessing, Large Language Models, Accessibility I . I N T R O D U CT I O N The rapid growth of digital technology over the past fiv e decades has been driven in large part by adv ances in semi- conductor density , as captured by Moore’ s Law [21]. As devices ha ve become smaller and more capable, technology has permeated almost e very aspect of daily life, bringing with it an ev er-e xpanding vocab ulary of abbreviations, acron yms, and technical jargon [11], [13]. Online content, including ne ws portals, e-commerce sites, and government services, increasingly contains this specialist language. For users who do not interact with technology on a deep le vel, encountering unfamiliar terms such as CPU , SSD , or API can block comprehension, reduce engagement, and e ven cause social or financial harm [8]. This group is commonly described as digitally illiterate : individuals who lack the skills, kno wledge, or confidence to use digital tools effecti vely [15]. Existing remedies, such as formal education programs, ICDL curricula, and UNESCO frameworks [12], [26], require institutional coordination and long lead times. They do not help a user who encounters an unknown acronym right now , mid-sentence, on a technology news page. This paper presents Acro Helper , a Chrome browser ex- tension that addresses this gap by injecting inline definitions at the point of need. The extension automatically detects the page category , identifies technical terms, and attaches hover - activ ated tooltip definitions without requiring the user to open a new tab or interrupt their reading flo w . An empirical study with 25 participants v alidates its ef fecti veness and measures its performance. I I . R E L A T E D W O R K A. Digital Illiteracy Digital illiterac y is broadly defined as the inability to use technology for reading, writing, communicating, and accessing information in a digital context [15]. Martin and Grudziecki [16] describe a three-lev el model (Digital Competence, Digital Usage, and Digital Transformation), where the lowest lev el cov ers foundational skills, kno wledge, and attitudes. This work targets primarily Level I users. Ke y factors associated with low digital literacy include age, education lev el, and socioeconomic background [18], [22]. Even in dev eloped countries the problem is substantial: a report by Accenture found that one in fi ve Irish adults under 34 self-rates their digital skills as average or below av erage [1], while the Netherlands reports 15% of the population lacking advanced digital skills [10]. Pew Research Center projections suggest the trend toward a more technology-driven world will deepen these inequalities without intervention [2]. B. Acr onyms as Barriers to Compr ehension Shulman et al. [23] demonstrated that exposure to jargon reduces readers’ interest in similar articles compared with jargon-free equi valents. Appelman [3], [4] showed that readers dev elop negati ve sentiments towards acronyms they do not understand, and that such exposure can discourage further reading of technology-related content. For users who already hav e lo w digital skills, this creates a feedback loop that prev ents kno wledge acquisition. C. Contextual Help T ools Sev eral prior works have built contextual help systems for users with limited digital skills: • Tipper [9] is a bro wser extension providing contextual help on icons, links, and buttons; usability tests with seniors were strongly positiv e. • LemonAid [7] is a crowdsourced, selection-based help system for web applications, achieving a result for 90% of help requests. • Y eh et al. [28] demonstrated screenshot-based contextual help for desktop GUIs, validated across 60 real tasks. • Y ada v et al. [27] showed that SMS-based acronym retriev al can assist semi-literate users outside browser en vironments. • NASA Acronyms [17] is an open-source browser exten- sion with over 25,000 space-related acronym definitions, av ailable in Chrome and Firefox. None of these prior tools combine AI-driv en page classifi- cation with LLM-powered, context-aw are acronym definition deliv ery in a fully automated pipeline. D. NLP for W eb Content Classification Atkinson and V an der Goot [5] showed that text-mining techniques applied to news articles can detect content cate- gories in near real time. T okenization, stemming, and lemma- tization [6] improve classification accuracy during prepro- cessing. Lämmel [14] demonstrated that XPath-style selectors improv e DOM parsing performance. This body of work un- derpins the content-extraction and classification pipeline used in Acro Helper . I I I . R E S E A R C H Q U E S T I O N S Three research questions guide this work: RQ1 Can a contextual help browser extension improve the comprehension of online articles containing technical terms among users with low digital literacy? RQ2 Does the contextual help tool positi vely af fect information-retriev al time for these users? RQ3 How can AI assist a browser extension in accurately classifying web pages by content domain? The central hypothesis is that showing on-demand defini- tions, activ ated by mouse hover , eliminates the need to leav e the active tab, increases comprehension, and reduces the total time spent resolving unknown terms. I V . S Y S T E M A R C H I T E C T U R E A. Extension Building Blocks Acro Helper is built with the Plasmo framework [19], which enforces best-practice folder structures, provides T ype- Script support, and manages bundling. The extension follo ws the standard Chrome extension model, composed of: • Content script — runs in the context of every visited page, reads and modifies the DOM. • Background service worker — listens for bro wser ev ents and routes messages. • Manifest v3 file — declares permissions, entry points, and metadata. • T ab page — a bundled React page providing a full-text dictionary search interface. All content-manipulation logic runs asynchronously and non-blocking, so the browser’ s normal rendering pipeline is not affected. B. F our-Phase Pr ocessing Pipeline When a page finishes loading, the extension ex ecutes four sequential phases, illustrated in Fig. 1. Phase A Sanitise & Parse Phase B AI Classification Phase C T erm Detection Phase D DOM T ooltip Injection Google NLP T axonomy API OpenAI ChatGPT T echnical Dictionary API OpenAI ChatGPT Page load ev ent Hover tooltips in DOM Fig. 1. Four-phase pipeline of the Acro Helper browser extension. Dashed arrows indicate calls to external AI services. Phase A — Content Sanitisation and Parsing. The extension clones the document and strips noise elements (headers, footers, cookie banners, navigation menus, sidebars, and