Functionality

We have worked hard to implement all the key and additional requirements:

• Test user knowledge on the PRIDAR standard
• Illustrate how prepared user’s projects are for NHS deployment
• Help users understand the PRIDAR requirements
• EU GDPR compliance
• Allow for easy management by clinical professionals
• Lightweight and easily distributable software
• Provide visualisations of data
• Establish a national average for scores

While we are proud of this accomplishment, some features could be improved upon and deepened further with more time.

For instance, the test mode could be adjusted to allow users to test for a single topic or custom range of topics. Similarly, the view resources feature could allow the same kind of customization instead of listing all the resources.

Stability & Efficiency

The PRIDAR chatbot functions consistently well, much in part thanks to Voiceflow hosting the project, and the chosen backend integrations being reliable and giving fast API responses. The same is true of the bot’s Telegram deployment, which we are very proud of.

Those strengths however could rarely turn into weaknesses, for instance if one of our service providers is down for maintenance, features cannot work, and if this happens to be Voiceflow then the bot will be completely down. That being said, this is a rare occurrence and can be kept track of easily through the appropriate provider’s website.

Also, the internal algorithms used to sort content and search users are linear. While this works just fine for now, in the case of the bot becoming congested with content and users, this could slow it down and even time requests to the bot. The reason for these linear algorithms is a lack of time and difficulty implementing data structures in Voiceflow’s coding environment, which lags with complicated code, provides no syntax highlighting, and is restricted to a very small view with no scroll bar, making it hard to use.

Availability

We believe that the PRIDAR chatbot is extremely accessible to users, and was one of the key goals we kept in mind during development. This was achieved by using reliable services which are rarely ever down or not functioning as intended.

Our text-based approach and lightweight deployment played an arguably bigger part in this however, as users could interact with the bot regardless of their hardware, through either Telegram or Voiceflow’s web app, requiring just a url to visit.

Maintainability

A huge consideration going into this project was how clinicians would both update the bot’s content, as well as potentially extend it’s functionality. This drove our decisions in technology, and we believe it paid off.

The bot is entirely manageable through the databases spreadsheet interface, which is intuitive, and can be shared easily. This data can also be transported elsewhere, managed, and integrated with extra functionality (such as automatically emailing or running scripts on).

The Voiceflow project is very accessible due to it’s easily comprehensible block design, and the attention we paid to commenting on blocks and code thoroughly. In addition to this, we have created a step by step walkthrough of the components in the Voiceflow project, with highlighted examples and notes explaining the most intricate of details in order to help clinicians who aren’t familiar with the project or Voiceflow edit it down the line without issue.

Project Management

Our team allocated roles efficiently, based on both availability and ability. We always worked using version history (both for the website and Voiceflow project) in a collaborative manner.

As much as possible, we followed an AGILE approach to development, meeting our client Dr Connor every Thursday, before discussing feedback and creating a plan for the coming week together in a separate meeting. We would then have meetings at the start of the week to check our progress and evaluate our direction.

Our main mode of communication was through Whatsapp, which was extremely convenient and made us more accessible to one another. We conducted team meetings through Zoom at least once a week, and presented through MS Teams. We also used MS office to share documents and collaborate, particularly towards assessed deadlines.

Bugs

Thanks to our comprehensive testing approach, our bot has no bugs present.

However, certain limitations could cause issues in the future, such as database GET requests only returning a certain number of records, databases only storing a finite number of records, or linear algorithms taking too long causing the bot to time out. Additionally, the bot currently does not scale very well with topics. This means that the resources table and averages table must be updated accordingly by hand when editing the number of topics in the bot.

Other issues with the chatbot, such as lack of ability to comprehend different inputs, are attributed to it’s limited scope rather than unintended bugs in behaviour.