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

Alessandro Crivellari,Euro Beinat

doi:10.3390/s20123503

Abstract

Neural machine translation is a prominent field in the computational linguistics domain. By leveraging the recent developments of deep learning, it gave birth to powerful algorithms for translating text from one language to another. This study aims to assess the feasibility of transferring the neural machine translation approach into a completely different context, namely human mobility and trajectory analysis. Building a conceptual parallelism between sentences (sequences of words) and motion traces (sequences of locations), we aspire to translate individual trajectories generated by a certain category of users into the corresponding mobility traces potentially generated by a different category of users. The experiment is inserted in the background of tourist mobility analysis, with the goal of translating the motion behavior of tourists belonging to a specific nationality into the motion behavior of tourists belonging to a different nationality. The model adopted is based on the seq2seq approach and consists of an encoder–decoder architecture based on long short-term memory (LSTM) neural networks and neural embeddings. The encoder turns an input location sequence into a corresponding hidden vector; the decoder reverses the process, turning the vector into an output location sequence. The proposed framework, tested on a real-world large-scale dataset, explores an effective attempt of motion transformation between different entities, arising as a potentially powerful source of mobility information disclosure, especially in the context of crowd management and smart city services.

Highlights

Trajectory recordings are gaining a continuously growing interest, allowing for a deep investigation of human mobility patterns [1,2]
Trajectory pre-processing, which describes how the original traces, continuous in time and space, are transformed into discrete location sequences; Trajectory linking, which reports how the input and target sequences used for training the deep learning model are generated and paired; Seq2seq model building, which explains how to train a model that learns the relationship between input and output sequences; Trajectory translation inference, which focuses on the evaluation phase characterized by the generation of new output trajectories as a result of corresponding input sequences
Inspired by neural machine translation approaches in computational linguistics, we presented a deep learning-based methodology to mine human mobility patterns and convert motion traces of a group of users into motion traces of a different group of users

Summary

Introduction

Trajectory recordings are gaining a continuously growing interest, allowing for a deep investigation of human mobility patterns [1,2]. The rise of data availability has boosted the interest in human mobility [3,4], paving the way to various data mining approaches for motion behavior analysis and trajectory-related studies [5,6]. Whereas a number of different tasks have been performed on mobility traces (e.g., trajectory prediction [7,8,9,10], trajectory classification [11,12,13], motion flow modeling [14,15,16], activity recognition [17,18]), the potential uses of such data expand over a multitude of new evolving use cases. The way that NLP developed its powerful methodologies and processing tools represents a source of information and analytical procedures for sequence-related problems

Objectives

Methods

Results

Discussion

Conclusion