Motivation, Participation and Organisation in Crowdsourcing Translation – A Case Study on Yeeyan

Crowdsourcing, i.e. engaging online communities in productive activities, has become a commonly used practice in content-generating areas, including translation. It has fostered alternative translation models, such as volunteer or paid crowdsourcing by professional or bilinguals translating different text types with various purposes. Scholars embarked on the investigations of crowdsourcing translation practices in the West (Dolmaya, 2012; Jiménez-Crespo, 2013a; O’Brien and Schäler, 2010; Olohan, 2014, 2012; Orrego-Carmona, 2015). However, little focus has been paid to crowdsourcing translation in China. This study explores the mechanism of crowdsourcing translation drawing from one of the largest Chinese amateurs’ communities – Yeeyan, seeking to raise awareness of the particular collaborative model both in the Language Services Industry and in Translation Studies. The study examines collaborative translation projects in Yeeyan, aiming to find out: 1. the motivating and de-motivating factors that influence participation; 2. the individuals’ contribution to crowdsourcing translation projects; 3. the quality of crowdsourcing outputs and the effectiveness of quality control methods; 4. the texts types that are feasible to translate via crowdsourcing; 5. the peer-learning process taking place in the crowdsourced environment. Questionnaire survey and observation were the main methods for data collection. The result of the questionnaire survey suggested that individuals are usually driven by multiple reasons both intrinsically and extrinsically. However, extrinsic motivations are perceived to be more important than intrinsic motivations. Moreover, it revealed that the individual’s behaviour (i.e. the time and effort that one invests in crowdsourcing translation) is influenced by these motivations. Empirical data of collaborative translation projects in the chosen environment was collected through Observation. Workflow Mapping demonstrated how projects progress, and identified two types of workflow that cater for different translation purposes: • the user-driven flexible workflow mainly depending on the participants’ spontaneity; • the waterfall workflow strictly managed and monitored by the project manager. Furthermore, the study highlighted the role of communication in facilitating collaboration. Dialogue Act Analysis was conducted to find out the frequency and category of interaction, and the learning process embedded in online peer-to-peer interaction. Social Network Analysis was then used to visualise the communication patterns, to characterise the power relationship between participants, and to illustrate the individuals’ contribution to the projects. An Error-based Quality Assessment dealt with the outputs of crowdsourcing translation at different stages (translation, revision, finalisation). The result revealed that the crowd can produce ‘fit-for-purpose’ translation through collaboration. It proved that it is feasible to translate both specialised texts and literary texts through crowdsourcing. The assessment focused on the effectiveness of quality control methods and showed that ‘peer-revision’ is the most efficient approach in improving translation quality. In addition, the study identified the significant error types which the crowd makes most often. More importantly, a peer-learning process was recognised in both the interaction between participants and the feedback provided during collaboration. The study produces a comprehensive evaluation of collaborative translation projects in the crowdsourced environment. The outcome is expected to bridge the gap between this relatively new phenomenon in translation practices and Translation Studies, to offer insights into how crowdsourcing translation projects can be organised more effectively, and to offer suggestions regarding setting up enhanced crowdsourcing initiatives for both translation production and translation training.