Unlocking a Knowledgeable Chatbot Using Wikipedia and GPT-3
Written on
Chapter 1: Introduction to the Chatbot
Explore the capabilities of a chatbot that draws its knowledge from Wikipedia and operates seamlessly online. This tool offers responses to inquiries in natural language, including voice interactions, all at no cost and without the need for registration.
Here is a powerful chatbot that allows conversational interaction, utilizing the advanced GPT-3 technology:
Go Talk With this Chatbot Online, Now!
A web application that integrates GPT-3 with speech recognition and synthesis, all accessible via the web.
Chapter 1.1: The Rise of the Knowledgeable Bot
Although GPT-3 has been trained on a wealth of data, its knowledge is not all-encompassing. To enhance its capabilities, I’ve developed a new version that taps into Wikipedia for information, making it significantly more informed.
You can access the bot here if you have an API key (available with free tokens from the OpenAI website for GPT-3):
(I recommend using Chrome for the best experience with speech recognition and synthesis.)
Section 1.1.1: Understanding How the Wikibot Operates
The technical workings behind this "Wikibot" are complex, but I break down the process in detail (including how to obtain the source code) in a separate article:
A Question-Answering Bot Powered by Wikipedia Coupled to GPT-3
Still captivated by the potential of GPT-3 when paired with Wikipedia.
towardsdatascience.com
The bot functions by querying Wikipedia for pertinent articles and employing the content of these articles to enhance its learning for the question at hand. In a nutshell, it processes the user's question by filtering out insignificant words, queries Wikipedia with the refined query, and retrieves a selection of articles that likely contain relevant information.
The bot then appends the original question to each article's content and utilizes GPT-3 to generate plausible answers based on the retrieved texts. Notably, the bot provides multiple answers—one for each article accessed—and ultimately summarizes these responses. Additionally, each answer includes a link to the corresponding Wikipedia article, allowing users to verify the information, given that GPT-3 may not fully grasp the accuracy of its responses.
You can also check out a more sophisticated version of the tool here:
Section 1.2: Expanding on GPT-3 Capabilities
While the tool is not without flaws, it performs admirably based on my testing. For further insights into GPT-3 and how few-shot learning enhances its responses, consider reading my articles:
All my articles on GPT-3 as of October 2022
Exploring my favorite language model and its applications across various online platforms using pure JavaScript.
GPT-3-like models with extended training could be the future 24/7 tutors for biology students
Despite mixed results in other subjects, GPT-3 excels in biology, especially with…
towardsdatascience.com
Control web apps via natural language by casting speech to commands with GPT-3
An article detailing practical uses of GPT-3, including comprehensive explanations of workflows.
towardsdatascience.com
Coupling GPT-3 with speech recognition and synthesis to achieve a fully talking chatbot that runs…
A guide on how I developed this web app, enabling natural conversations with GPT-3 on any topic, all online.
towardsdatascience.com
Chapter 2: Video Demonstrations
This video demonstrates how to create a chatbot that utilizes Wikipedia, guiding you through the process step by step.
In this video, learn how to develop a question-and-answer chatbot that searches through wikis, showcasing its practical applications.