Apply for community grant: Academic project (gpu)
Conception Phase: Task 1 Chatbot for booking a hotel room.
Who wouldn’t like to minimize the hassle of hotel booking? Firstly, the client needs to input the dates of check-in and check-out, the hotel's location, and the number of people. How many are adults, and how many are kids? Even when you get the available hotel rooms, the customers must go through filters based on their personal preferences, such as the included breakfast filter. The personal data of the customer is required to complete the booking process. In the end, a chatbot booking assistant aims to provide a quick and convenient service to save time and effort for your hotel bookings.
To save time and costs, a proposed service provider for our hotel booking chatbot is Amazon AWS. AWS Bedrock service offers access to base Large Language Models (LLMs). Additionally, AWS Bedrock provides access to common embedding APIs. (Amazon Bedrock Documentation, n.d.) A base LLM requires a knowledge base to answer questions on a specific domain. To collect data for our hotel assistant, web scrapping techniques, like UnstructuredUrlLoader from Langchain, are helpful. Web scrapping is the retrieval of text data from a website where the URL is provided (Pandya & Holia, 2023; Topsakal & Akinci, 2023). This way allows the collection of data from specific hotel websites to create a knowledge base. An easy and practical way to store the data is using the AWS S3 bucket service. (Amazon S3, n.d.)
Furthermore, an Embedding API is required to translate the data into a mathematical sequence and store them in a vector space. This is one of the critical elements for creating the memory. (Devaraj & Gangrade, n.d., Pandya & Holia, 2023) Since the goal is for the chatbot to remember various information that the customer provided and take specific actions, Langchain Conversation Memory buffer is required. This open-source framework offers state context memory, so the LLM can generate content based on the information that it has already received.(Introduction | Langchain. (n.d.), Topsakal & Akinci, 2023). Also, the Langchain Conversation Chain is required to introduce a state for every chat session. The assistant can now provide different actions from information that it received during the session.
To make the hotel assistant user-friendly, a front-end user interface is necessary. Streamlit is a Python library to create the front-end component of a web application. (Streamlit Docs. (n.d.))The goal is to deploy the chatbot application using AWS EC2 service, which offers a secure and robust computing platform. Scalability is also one of the most beneficial advantages for the future development of the product. (Benefits, n.d.)
References:
1.
Devaraj, P. N., PV, R. T., & Gangrade, A. (2023). Development of a Legal Document AI-Chatbot. arXiv preprint arXiv:2311.12719. https://doi.org/10.48550/arXiv.2311.12719
2.
Pandya, K., & Holia, M. (2023). Automating Customer Service using LangChain: Building custom open-source GPT Chatbot for organizations (arXiv:2310.05421). arXiv. https://doi.org/10.48550/arXiv.2310.05421
3.
Topsakal, O., & Akinci, T. C. (2023). Creating Large Language Model Applications Utilizing LangChain: A Primer on Developing LLM Apps Fast. International Conference on Applied Engineering and Natural Sciences, 1, 1050–1056. https://doi.org/10.59287/icaens.1127
4.
Amazon Bedrock Documentation. (n.d.). Retrieved February 18, 2024, from https://docs.aws.amazon.com/bedrock/
5.
Amazon S3. (n.d.). Amazon Web Services, Inc. Retrieved February 18, 2024, from https://aws.amazon.com/pm/serv-s3/
6.
Benefits. (n.d.). Amazon Web Services, Inc. Retrieved February 18, 2024, from https://aws.amazon.com/application-hosting/benefits/
7.
Introduction | Langchain. (n.d.). Retrieved February 18, 2024, from https://python.langchain.com/docs/get_started/introduction
8.
Streamlit Docs. (n.d.). Retrieved February 18, 2024, from https://docs.streamlit.io/
Figure 1 Own representation of the chatbot architecture.