Matt Porritt, a Moodler, former air traffic controller, and Solutions Architect at New Zealand Moodle Partner Catalyst IT has taken a concrete first step into Moodle Chatbot territory.
His development, showcased as part of his “Find All The Things!” presentation, describes how the chatbot can be useful for general search tasks. Using our “Chatbot ingredients list,” let’s take a look at the choices made by Porritt in his exciting attempt:
Use case: General Productivity. The chatbot, which according to its maker could be “easy to integrate” across Moodle pages, could start as a “Help Center” offering global search results. Over time, it could incorporate more factors in its response (for example, Moodle Contexts) and evolve towards more robust “Natural Language Understanding,” offering more relevant answers to less standardized questions. It could even get closer to recognizing, and perhaps also eliciting, intent. Offering support for all users would also mean it would be able to reach a wider number of language learning interactions.
Business model: As a product of a Catalyst IT engineer, the company is financing its deployment, which so far has remained open and available on Porritt’s personal GitHub. We’ll have to wait and see if the Moodle Partner keeps it free, turns it into a plugin, and allows access to the learning data set. At this point, the end game is uncertain (possibly for Catalyst, too).
Development platforms: Porritt took advantage of several services offered by Amazon Web Services (AWS), starting with Lex, which is the basis of Amazon’s voice-basedassistant Alexa. Like other platforms, Lex can power chatbots in several clients, including Facebook Messenger and Slack. It is hard to tell if the chatbot would have been better through any other platform, but Lex’s advantages include being free for AWS users, seamless integration to the Services ecosystem, simplicity of use, and the ability to eventually turn the chatbot into a “skill” for Alexa.
Engines: In addition to the NLU model, Porritt adopted other tools to enhance the quality of service, including:
- Elastic Search is the service chosen to power Moodle’s Global Search. It is maintained by Catalyst and is compatible with AWS.
- Rekognition by AWS offers image analysis, allowing results to include images that correspond with the search terms within a statistical level of confidence. (See video.) The Elastic Search plugin can run Rekognition to include image analysis data into search results.
- A possible next step might involve Sentiment Analysis to try to identify the user’s emotions and respond accordingly. Porritt shows an analysis of sentiments made on a Moodle Forum using IBM Watson.
Community: According to the repository, Porritt has been the only developer since the project started about two and a half months ago. Community involvement, and user feedback in particular, is the basic ingredient needed for the chatbot to grow. It would also help define the priorities for future development, among which Porritt lists:
- Sentiment as an Intervention Trigger
- Full AI Search Engine
- Faceted (filtered) and Contextual search
- Validation and research for further scalability.
Click here for more about Moodle Chat Bot or “Moodle Lex” at GitHub (in development).