Research On Task Distribution Method Of Crowdsourcing Translation Based On Worker Feature Model

In recent years,with the gradual developing of the Internet,computer performance and natural language processing capabilities,through the global Internet platform,assisted with the analysis of algorithms and automatic control of processes,bringing workers join the task as many as possible and achieving crowdsourcing tasks at a lower cost and achieving better results become a research hotspot.This is also the key research content of group intelligence collaborative computing.In crowdsourcing collaborative translation,screening crowdsourcing workers and quality controling is the key to achieve better translation results and control costs.How to assign tasks quickly and accurately and how to control the quality of translated texts will be the focus of research.Taking Thai and Chinese translation as an example,this project studied the construction method of crowdsourcing translators based on worker characteristics,designed and completed a crowdsourcing translation system model and framework based on collaborative computing,implemented the key technologies and modules and made the feasibility analysis of screening model and quality evaluation by experimental testing.In terms of task assignment,this project constructed a worker screening model by analytic hierarchy process(AHP)and found the most suitable dynamic matching of workers and tasks based on this.Compared with the distribution method based on historical information,the problem of"cold start" has been solved.In the initial stage,the model can objectively assign between workers and tasks according to the initial data.In the later stage of the model,with the continuous feedback of workers'information,the distribution effect is more significant.In terms of translation quality control.In this project,an improved method based on BLEU algorithm is used to evaluate Thai automatically.Through word segmentation technique and synonym substitution,the flexibility and accuracy of BLEU algorithm in evaluating Thai text have been greatly improved.In the end,the application of worker screening model and automatic translation evaluation technology in crowdsourcing translation system will help to quickly identify workers and obtain high-quality results,solving the problem that the cost of professional translation is too high and the quality of machine translation is too low.