Deep learning supported location-aware keyword query
In this project report, I explore the integration of deep learning techniques with location-aware keyword queries for improved information retrieval in geospatial contexts. It addresses the challenge of efficiently managing high-dimensional geo-textual data and processing queries over large datasets...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project (FYP) |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/175672 |
| _version_ | 1826115384241029120 |
|---|---|
| author | Xia, Tianyi |
| author2 | Gao Cong |
| author_facet | Gao Cong Xia, Tianyi |
| author_sort | Xia, Tianyi |
| collection | NTU |
| description | In this project report, I explore the integration of deep learning techniques with location-aware keyword queries for improved information retrieval in geospatial contexts. It addresses the challenge of efficiently managing high-dimensional geo-textual data and processing queries over large datasets, considering the computational expense of existing similarity metrics. The project implements novel methods for enhancing computation in this domain and adopts MGeo, a multi-modal geographic language model that incorporates geographic context as a modality. Additionally, a large-scale benchmark dataset, GeoTES, is used to facilitate further research in query-POI matching. Extensive experiments are conducted to evaluate the proposed methods against state-of-the-art pre-trained language models (PLMs), demonstrating significant performance improvements. Despite challenges such as the absence of user geolocation or query geographic context, MGeo exhibits the capability to model correlations between text, geographic context, and their combinations. The report concludes by suggesting future research directions, including the exploration of additional modalities such as POI images and the development of inventive geographic encoders to further enhance performance across geography-related tasks. |
| first_indexed | 2024-10-01T03:54:16Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/175672 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:54:16Z |
| publishDate | 2024 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1756722024-05-03T15:38:41Z Deep learning supported location-aware keyword query Xia, Tianyi Gao Cong School of Computer Science and Engineering gaocong@ntu.edu.sg Computer and Information Science In this project report, I explore the integration of deep learning techniques with location-aware keyword queries for improved information retrieval in geospatial contexts. It addresses the challenge of efficiently managing high-dimensional geo-textual data and processing queries over large datasets, considering the computational expense of existing similarity metrics. The project implements novel methods for enhancing computation in this domain and adopts MGeo, a multi-modal geographic language model that incorporates geographic context as a modality. Additionally, a large-scale benchmark dataset, GeoTES, is used to facilitate further research in query-POI matching. Extensive experiments are conducted to evaluate the proposed methods against state-of-the-art pre-trained language models (PLMs), demonstrating significant performance improvements. Despite challenges such as the absence of user geolocation or query geographic context, MGeo exhibits the capability to model correlations between text, geographic context, and their combinations. The report concludes by suggesting future research directions, including the exploration of additional modalities such as POI images and the development of inventive geographic encoders to further enhance performance across geography-related tasks. Bachelor's degree 2024-05-03T00:38:10Z 2024-05-03T00:38:10Z 2024 Final Year Project (FYP) Xia, T. (2024). Deep learning supported location-aware keyword query. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/175672 https://hdl.handle.net/10356/175672 en SCSE23-0605 application/pdf Nanyang Technological University |
| spellingShingle | Computer and Information Science Xia, Tianyi Deep learning supported location-aware keyword query |
| title | Deep learning supported location-aware keyword query |
| title_full | Deep learning supported location-aware keyword query |
| title_fullStr | Deep learning supported location-aware keyword query |
| title_full_unstemmed | Deep learning supported location-aware keyword query |
| title_short | Deep learning supported location-aware keyword query |
| title_sort | deep learning supported location aware keyword query |
| topic | Computer and Information Science |
| url | https://hdl.handle.net/10356/175672 |
| work_keys_str_mv | AT xiatianyi deeplearningsupportedlocationawarekeywordquery |