Helping languange models process spatial data using Langchain

The advent of ChatGPT and other Large Language Models in recent years has caused a surge in popular interest in Artificial Intelligence and its capabilities. Although Large Language Models may seem capable of an endless variety of tasks, there are still areas it struggles in such as the halluc...

Full description

Saved in:
Bibliographic Details
Main Author: Thong, Gareth Jun Hong
Other Authors: Long Cheng
Format: Final Year Project
Language:English
Published: Nanyang Technological University 2024
Subjects:
Online Access:https://hdl.handle.net/10356/175031
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:The advent of ChatGPT and other Large Language Models in recent years has caused a surge in popular interest in Artificial Intelligence and its capabilities. Although Large Language Models may seem capable of an endless variety of tasks, there are still areas it struggles in such as the hallucination problem or in its understanding of non-textual data, like geospatial data. This project seeks to address this issue by exploring methodologies for Large Language Models to interpret and use geospatial data more accurately, and to develop an easily operable workflow that uses PostGIS and LangChain, among other technologies. The workflow will be grounded in theoretical concepts like few-shot learning, which will be elaborated upon in this report. An application user interface incorporating this workflow will be built in Flask, and it is hoped that the results presented in this report will be useful for the future development of software wishing to best utilize Large Language Models for processing geospatial data.