Trace2trace-A Feasibility Study on Neural Machine Translation Applied to Human Motion Trajectories

Area of Science:

• Computational Linguistics
• Human Mobility Analysis
• Deep Learning Applications

Background:

• Neural machine translation (NMT) excels at text translation using deep learning.
• Human mobility analysis involves understanding movement patterns from location data.
• A gap exists in applying NMT techniques to trajectory data.

Purpose of the Study:

• To assess the feasibility of transferring NMT to human mobility and trajectory analysis.
• To translate individual trajectories between different user categories, specifically tourist nationalities.
• To explore novel methods for mobility information disclosure.

Main Methods:

• Utilized a sequence-to-sequence (seq2seq) model with an encoder-decoder architecture.
• Employed long short-term memory (LSTM) neural networks and neural embeddings.
• Mapped sentence translation concepts to location sequence translation.

Main Results:

• Successfully translated motion behavior between different tourist nationalities.
• Demonstrated effective motion transformation between distinct entities using real-world data.
• Validated the framework on a large-scale dataset.

Conclusions:

• Neural machine translation principles can be effectively applied to human mobility analysis.
• The proposed framework offers a novel approach to understanding and transforming trajectory data.
• This method has significant potential for crowd management and smart city applications.