Enhancing Mobility Access through User-Contributed Accessibility Data

Enhancing Mobility Access through User-Contributed Accessibility Data

Crystal Merlin D Souza, Sonal Riva Gonsalves , Gururaja S, Nishmitha J

Department of Computer Applications

St Joseph Engineering College, Mangalore, Karnataka, India

Abstract – Navigating public spaces can be a challenge for individuals having mobility impairments due to the lack of accessible infrastructure and real-time information. Through a web-based and mobile system that allows users to access and submit accessibility information for public spaces, this study presents a crowdsourced solution to this problem. User- submitted information about facilities like elevators, ramps, wide doors, and accessible restrooms is gathered by the platform. The system automatically creates accessibility tags from user descriptions using lightweight natural language processing, with the option for inspection manually via an admin dashboard, to increase data consistency and minimize verification effort. To find accessibility trends and coverage gaps, especially in urban settings, the gathered data is examined. The system includes a recommendation feature that makes suggestions for places based on accessibility tags from a user's previously bookmarked places in order to enhance user experience further. This study demonstrates how automatic tagging, structured moderation, and community engagement can be utilized to generate scalable, reliable tools that enhance mobility access and empower people with disabilities. Evaluation results indicated consistent tagging accuracy and a positive response to QR-based interactions, validating the systems usability in real-world settings.

Keywords Accessibility, crowdsourced data, mobility challenges, natural language processing, public space navigation, recommendation systems, smart tagging, user- generated content

1. INTRODUCTION

   The lack of real-time accessibility information highlights the necessity for a platform that not only gathers and shares such data but also verifies and enhances it over time. Crowdsourcing collecting input from a wide community of users offers an effective way to achieve this. When used for accessibility mapping, it allows people to work together to report features like ramps, elevators, wide entrances, and accessible restrooms, helping to build a dynamic dataset that truly reflects real-world conditions.

   This paper introduces a system that solves these mobility problems by leveraging user-provided information. Accessibility data is contributed by users as they visit public places, sharing their insights directly on the app. Administrators checks these submissions through an interface that supports automatic tagging powered by efficient NLP techniques. The system enhances this with a recommendation tool, suggesting places to visit by matching tags from locations users have already bookmarked. This paper also

   analyzes the collected data, identifies any gaps in geographic and feature coverage, and proposes an ongoing feedback loop to further boost reliability and ease of use. In addition, it examines methods for improving the user experience through automated configuration options. The systems structured moderation, adaptive personalization, and real-time features set it apart from existing solutions like WheelMap and AccessNow, providing users with clear advantages.

2. LITERATURE REVIEW

   There have been various research studies to check out how mobility access for physically disabled individuals can be improved through technological solutions, primarily in the domains of crowdsourcing and mobile apps.

   González and Parra [1] in their research discussed how crowdsourced information can be useful in making mobility- impaired people more accessible to smart cities. They emphasized the point that traditional approaches based on pure official government surveys or formal data are not sufficient, time-consuming, and disconnected from the ground realities of citizens. Crowdsourcing was proposed as a way to engage users actively to map accessible surroundings and identify infrastructural barriers.

   Degbelo et al. [3] proposed a participatory and open-mapping approach to facilitate inclusive city spaces. They highlighted the potential of citizen-generated information to augment city data sets to uncover gaps and inconsistencies, particularly in the mobility and navigation domain.

   Khanna and Yadav [2] considered challenges and possibilities of crowdsourcing for social benefit, with particular emphasis on its application in accessibility and public infrastructure projects. They mentioned such important issues as data quality, user motivation, and scalability of the system issues that are checked while designing this project. Several mobile-based applications such as WheelMap, AccessNow, and AXS Map have previously attempted to collect accessibility data via user contributions.

   This study builds on these insights by merging participatory data collection with automatic tagging through natural language processing, incorporating an admin review process, and designing a scalable system that excels in accessibility mapping. By blending crowdsourcing with intelligent

   automation, we can find cost-effective and adaptable solutions to mobility access challenges in everyday situations.

3. SYSTEM DESIGN AND IMPLEMENTATION

   The suggested method is an internet-based and mobile platform designed to assist individuals with mobility disability in coping with accessibility barriers. Users can contribute, search, and navigate details on accessible features of public venues. Crowdsourcing, administrative moderation, automatic tagging by natural language processing, and a tailored recommendation module are all embraced by the system. Scalability, data reliability, and cross platform functionality are highly prioritized within the design. The system also facilitates QR based sharing and retrieval of accessibility information in order to ease access in real-world situations.

   1. Architecture

      The mobile application's user interface, backend APIs and data storage all exist within the service oriented architecture. The frontend is done with flutter. The backend is built with Node.js, Express.js and RESTful APIs for data submission, and user authentication. The database is MongoDB and includes cities, submissions for places, and user profiles. All authentication is through a token-based login system.

   2. User Roles

      The user roles defined in the system are regular users and administrators. Regular users can sign up, sign in, and submit accessibility information about public places they have visited, search approved places, bookmark approved places and receive personalized recommendations on which approved places to check out based on their bookmarks. Administrators have access to the dashboard that allows administrators to review and moderate the places that users submit, change auto-generated tags, and filter places submitted by rows.

   3. Functional Components

      Our platform allows users to add new locations. Adding a new location will require users to fill out a form that asks for location name, type, description, and an image upload. Users can submit the information, and the location will be in pending status, which means that it is not displayed to the public until the information is approved by the admin.

      The user descriptions are processed in the back end by a NLP processor in order to generate accessibility tags. The system identifies certain predetermined words such as step free access, lifts, ramp, and converts them into accessibility tgs. Making use of an auto tagging system allows users to submit location descriptions in a consistent format and requires very little manual labor to incorporate the submissions.

      Fig. 1. Add Place interface with auto-detected accessibility tags based on user-submitted description.

      Fig. 1 depicts the user-facing form used to submit a new location in the WheelBuddy mobile app. The user must enter a location name, location description, coordinates, and optionally an image. As the user enters their location description, the system will use natural language processing to identify and highlight for the user appropriate tags related to accessibility (e.g. "ramp", "elevator").

      Fig. 2. Flowchart showing the regular user interaction flow

      Fig. 2 depicts the steps from registration to location discovery, along with QR code look ups and the interaction flow with regular use.

      Moderation is performed using a specific admin panel. The administrators can filter pending submissions by status, submission date or city of origin. Moderators can change the auto-generated tags for each entry before accepting or rejecting it. Approved locations will automatically be included in the public map view.

      Fig. 3. Flowchart showing the admin workflow

      Fig. 3 illustrates the admin moderation flow, emphasizing the quality control procedures and review process.

      Once approved, a location appears on the interactive map, allowing users to search by location, keyword, or accessible attributes like "wheelchair restroom" or "wide entrance." Every map marker provides both deep exploration of surrounding areas and accessibility features.

      A rule-based recommendation engine looks at a user's saved locations to find other sanctioned locations with matching tags (i.e. "ramp", "elevator") to improve personalisation. It's lightweight to maintain, and simple to comprehend, therefore a recommendation model offers real time personalised suggestions for every user.

      Moreover, the type of technology allows for users to create QR codes for each sanctioned site. When the user scans the site QR code, using the in-app scanner or QR reader, it takes users to a details page with verified access features.

   4. Data Flow

   The system has a logical data flow that starts with user input and concludes with individualized sharing and discovery. When a user wants to submit a new location, the user will input the location coordinates, description, and if required photos. The backend has a natural language processing engine that reviews the place description and all the details have been provided. The result is processed to suggest accessibility tags like ramp, elevator, step-free access which are tagged to the place details.

   Now the submitted result can be reviewed by Admin after it has gone into a pending state. The dashboard admins can analyze the tags and content, perform manual alteration, and

   approve or reject the location. When it is given, the location records will be added to the approved dataset and it will be publicly visible on an interactive map and will be searchable in the app.

   As users navigate and bookmark places, the system develops a tag profile of what access elements get saved the most often. The tag based recommender system would make required suggestions for other authorised locations, based on tags tagged in the bookmarked place in a user's profile.

4. DISCUSSION

1. Impact of Crowdsourced Accessibility Data

   WheelBuddy showcases how social engagement can effectively bridge the gap in accessible information for public spaces. By enabling people with mobility challenges to share real-time updates about features like ramps, elevators, and wide entrances, the platform helps users make more assured and informed decisions regarding their trips. This community-driven approach not only improves quality and relevance of the information available, but also allows the resource to naturally expand and adapt as people contribute across different areas.

2. Balancing Automation with Human Oversight

   By automatically suggesting tags using NLP, users dont have to manually enter every detail, making it easy and faster to submit new places. Human moderation through the admin dashboard makes sure that all submitted information stays accurate and trustworthy. This combination of smart automation and careful oversight helps keep the system reliable, up-to-date, and easy for everyone to use.

3. Personalization through Tag-Based Recommendations

   By looking at tags that users have bookmarked, the tag-based recommendation system recommends related places. This now enhances usability by helping users quickly discover places that meet their own accessibility needs.

   Fig. 4 illustrates the detail screen for a bookmarked retrieval location which in this case is Kempegowda International Airport. Each detail screen has key accessibility information describing what is on-site regarding ramps, wide entrance door, wheelchair accessible washrooms, guide dogs, automatic doors. In addition to the accessibility information, the system has automatically created suggested tags based on what they had identified to help them to recognize features of the site. Finally, the interface displayed a notification indicating they have successfully bookmarked the site location.

   Fig. 4. Home screen displaying location recommendations based on user- bookmarked accessibility tags.

   Fig. 5 WheelBuddy homepage showing how users receive personal recommendations for accessible locations. Each recommendation is based on previously saved bookmarks and accessibility tags – ramp, elevator, wide entrances, and so on

   – as filters. This allows the user to find a location that meets her accessibility standards, leading to a user-centric and situationally relevant discovery process.

   Fig. 5. Home screen showing location recommendations based on user bookmarked accessibility tags.

4. Physical Integration with QR Code Sharing

QR codes can help bridge the gap between digital services and the physical world. Places can openly display the QR codes connected to each authorized site. The technology can be used in practical situations like hospitals, transit centers, schools, or any other public setting and users can scan a code to easily access information.

wide accessibility data, the platform facilitates navigation for new users.

1. COMPARISON WITH RELATED SYSTEMS

   WheelMap, AccessNow, and AXS Map are the current platforms that provide community based accessibility places. Even though they have been essential in increasing awareness and gathering crowdsourced accessible data, they frequently lack tools for personalization, automation, and structured moderation. Table 1 shows the comparison of features across these systems.

   Feature	Whee lMap	Acces sNow	AXS Map	Wheel Buddy
   Auto- tagging (NLP)	No	No	No	Yes
   Admin moderation	No	No	No	Yes
   Recommen dations	No	No	No	Yes
   QR code access	No	No	No	Yes
   Map-based search	Yes	Yes	Yes	Yes
   User ratings	Yes	Yes/p>	Yes	Yes

   Table. 1. Comparison of key features

2. SYSTEM EVALUATION AND RESULTS

   V. USE CASE SCENARIOS

   WheelBuddy provides useful features including map-based search, QR access, and tailored recommendations to improve accessibility in the real world. The examples that follow demonstrate how applicable it is:

   1. Navigating a Shopping Mall

      At the mall entrance, a consumer scans a QR code to display comprehensive accessibility services including "ramp access" and "elevator." Stores with "wide entrance" tags could be found using the map and filters. Places that are bookmarked generate recommendations for comparable accessible locations.

   2. Campus Accessibility

   Students can quickly view the accessibility of each amenity by scanning the QR codes on campus buildings. In addition to encouraging community contributions to enhance campus-

   The functionality and performance of WheelBuddy were assessed using internal testing and Postman API validation.

   1. Core Functionality Verification

      All devices were able to reliably use key functions as place submission, auto-tagging, moderation, search, bookmarking, and QR access. System stability was confirmed by backend APIs responding appropriately to important operations.

   2. Tagging Performance

      Common tags like "ramp" and "elevator" were correctly identified by the NLP engine, although as can be expected from a rule-based approach, it occasionally overlooked less common terms.

   3. Recommendation Output

      Bookmarked data was successfully used to generate output tag-based recommendations, which provided pertinent results with less processing.

   4. QR Integration

   Real-world usage was supported by consistently scannable location-specific QR codes that were accurately connected to place information.

3. CHALLENGES AND LIMITATIONS

   WheelBuddy has a few major issues that could affect its efficacy and scalability despite its advantages:

   1. Limitations of Rule-Based Auto-Tagging

      The accuracy of tags may be impacted by the existing rule- based NLP approach's lack of contextual awareness and potential inability to recognize accessibility phrases used in regional or diversified languages.

   2. Uneven Geographic Data Coverage

      The majority of user contributions come from urban areas, with rural areas contributing very little. This makes the platform less inclusive and necessitates the use of tactics to promote wider involvement.

   3. Scalability of Manual Moderation

   Although manual inspection guarantees data quality, it could not be scalable as the number of submissions rises. In the absence of community-based review mechanisms or automation, moderation may become a bottleneck.

4. CONCLUSION AND FUTURE WORK

WheelBuddy helps people fill in accessibility gaps by providing real-time information on public spaces and allowing users to participate. NLP-based auto-tagging, admin moderation, customized suggestions, and QR-based access are some of the main features.

Evaluation verified that the system worked, with dependable performance across modules and correct tag generation. The bookmarking and map functionalities were positively engaged by users.

Future research will concentrate on bringing scalable moderation through community feedback, strengthening suggestions using adaptive models, and increasing tag accuracy through machine learning. Increasing geographic reach via collaborations and outreach is also a top objective.

WheelBuddy serves as an example of how lightweight automation and participatory design may promote inclusive mobility in urban settings.

REFERENCES

1. J. L. González and R. Parra, Improving Accessibility for Mobility Impaired People in Smart City using Crowdsourcing, ResearchGate,

   Oct. 2019. [Online]. Available: https://www.researchgate.net/publication/336260036_Improving_Acc essibility_for_Mobility_Impaired_People_in_Smart_City_using_Cro wdsourcing

2. D. Khanna and A. Yadav, Crowdsourcing for Social Good: Framework and Challenges, SSRN Electronic Journal, Apr. 2024. [Online]. Available:

   https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5128436

3. M. Degbelo, A. Granell, S. Trilles, J. Huerta, and S. Kray, Participatory and Open Mapping for Accessible Cities: An Evaluation Framework, Open Geospatial Data, Software and Standards, vol. 2, no. 6, pp. 116, 2017. [Online]. Available: https://opengeospatialdata.springeropen.com/articles/10.1186/s40965- 017-0040-5

4. WheelMap. [Online]. Available: https://wheelmap.org

5. ccessNow. [Online]. Available: https://accessnow.com

6. AXS Map. [Online]. Available: https://www.axsmap.com