The Barcelona Activa Chatbot is a digital assistant created to enhance user interaction on the Barcelona Activa website. Barcelona Activa is the economic development agency of the Barcelona City Council. The goal of this project is to provide users with a seamless and efficient way to access information and resources related to economic development in Barcelona.

Methodology

Data Collection:
Relevant content from the Barcelona Activa website was collected using web scraping techniques. This content served as the basis for training the chatbot.

Content Processing:
The Recursive Character Text Splitter method was employed to organize the scraped content effectively. This process aimed to improve the chatbot’s understanding and responsiveness to user queries.

Content Embedding:
The processed content was embedded into a vector database using Pinecone, a vector database platform. This enabled efficient retrieval and matching of user queries.

Bot Training:
The Langchain LLM framework, along with OpenAI Embeddings, was used to train the chatbot. The bot was programmed to understand user queries and generate appropriate responses by leveraging the embedded content and links.

Images

Links

Demo BCN Activa Chatbot.

Technologies Used

• Python: Python programming language was used for various data processing and scripting tasks.
• Javascript: JavaScript was employed for frontend integration and user interaction components.
• Langchain LLM: This framework was used to structure and train the chatbot’s language capabilities.
• Flowise: Flowise technology was used to streamline user interactions and conversation flow.
• OpenAI Embeddings: OpenAI Embeddings were employed to enhance the chatbot’s understanding of textual content.
• Pinecone Vector Database: used for efficient storage and retrieval of embedded content.