Crowdsourcing Speech Data for Low-Resource Languages from Low-Income Workers

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2819–2826 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2819 Crowdsourcing Speech Data for Low-Resource Languages from Low-Income Workers Basil Abraham*, Danish Goel, Divya Siddarth, Kalika Bali, Manu Chopra, Monojit Choudhury, Pratik Joshi, Preethi Jyoti†, Sunayana Sitaram, Vivek Seshadri Microsoft Research India, *Microsoft India Development Center, †IIT Bombay basil.abraham@microsoft.com, danishdgoel@gmail.com, dsiddarth1@gmail.com, kalikab@microsoft.com, mchopra@cs.stanford.edu, monojitc@microsoft.com, pratikmjoshi123@gmail.com, pjyothi@cse.iitb.ac.in, sunayana.sitaram@microsoft.com, visesha@microsoft.com Abstract Voice-based technologies are essential to cater to the hundreds of millions of new smartphone users. However, most of the languages spoken by these new users have little to no labelled speech data. Unfortunately, collecting labelled speech data in any language is an expensive and resource-intensive task. Moreover, existing platforms typically collect speech data only from urban speakers familiar with digital technology whose dialects are often very different from low-income users. In this paper, we explore the possibility of collecting labelled speech data directly from low-income workers. In addition to providing diversity to the speech dataset, we believe this approach can also provide valuable supplemental earning opportunities to these communities. To this end, we conducted a study where we collected labelled speech data in the Marathi language from three different user groups: low-income rural users, low-income urban users, and university students. Overall, we collected 109 hours of data from 36 participants. Our results show that the data collected from low-income participants is of comparable quality to the data collected from university students (who are typically employed to do this work) and that crowdsourcing speech data from low-income rural and urban workers is a viable method of gathering speech data. Keywords: Labelled Speech Data, Low-resource Languages, Rural Communities 1. Introduction In the next 3 years, roughly 400 million Indians are ex- pected to get access to a smartphone for the first time (BusinessStandard, 2019). These new smartphone users speak one of India’s 19,569 different languages (IndiaTo- day, 2019), most of which are not represented in the main- stream. Moreover, given the complexity of existing hier- archical, text-based user interfaces, voice-based technolo- gies have proven to be the best way to reach rural users (Medhi Thies, 2015). Therefore, there is significant value in creating local-language technologies to serve these users. Almost all Indian languages can be considered as “low- resource languages”, i.e. there is little to no speech data available in these languages. Unfortunately, collecting good-quality labelled speech data in any language is an ex- tremely expensive and resource-intensive operation. This makes speech data collection the biggest bottleneck in cre- ating local-language voice technologies. Prior work has looked at crowdsourcing as a way to collect speech data at low cost (Hughes et al., 2010). So far, these platforms have only worked with educated urban partici- pants (such as university students) as they can be reached easily and are typically digitally savvy. However, hiring only urban participants leads to a lack of diversity in the collected dataset. It is important to note that even within the same …

Content truncated to 500 words