Annotate Against Bias
Create a language model that detects bias in German job ads.
Are you interested in how bias is expressed in the German language, and would you like to teach a language model how to write inclusively? You might be the one we are looking for!
Develop Diverse has built an AI-based software platform that promotes inclusive writing to attract more qualified candidates. The Platform detects implicit and explicit stereotypic language in texts and suggests neutral alternatives in real-time. This way, customers like Danske Bank, Vestas, DSB, and TDC have attracted more qualified candidates regardless of gender, ethnicity, age, neurodiversity, and physical disabilities while saving time on recruitment.
We already have an English, Danish, and Swedish language model covered and are now looking for someone to work on a German model. You will gain insights into how research is implemented and experience handling linguistic data.
The position is a part-time position, 20 hours a week. Remote work can be arranged after training and on-boarding. The position's duration is tied to the project on developing our German model, which we assume will be 4-5 months, but might be prolonged.
What You Will Do:
You will be reviewing research on how bias is expressed in German and transferring that knowledge into the language model within a proven annotation framework. One of the questions you will find an answer to is: How do we deal with grammatical gender in job titles? Another one could be: which German idioms fall into our predefined bias categories?
Additionally, you will be writing our German annotation guidelines, where you set up rules for which words and phrases should be detected and why.
You'll spend most of your time annotating our German text corpus of job ads, classifying words and phrases from job ads into overarching bias categories, thereby building the data foundation for our German language model.
You will not start from scratch but get a thorough introduction to our current framework and annotation style.
Later, you will analyze the model's performance to improve it by adding data or correcting inconsistencies.
What You Bring:
About Develop Diverse
We are a small company comprising 16 colleagues who enjoy each other's company and have a big team lunch every day. You will be part of our Data Team and work closely with our Linguist, who has been working on the English, Danish, and Swedish models.
What You'll Get
Apply
Apply by sending us your CV and writing a cover letter describing how you have worked with German culture or language in the past and how you can benefit from that in this position. What makes you good at detecting bias in a German context? Max. 350 words.
If you have any questions about the position, please contact our Linguistic Data Specialist, Linea at [email protected].
Evaluation process
We will evaluate applications on an ongoing basis. We always use blind CVs to make sure we minimize our own biases.
Accommodation Requests
If you need assistance during the application process, please let us know and we will make sure to accommodate.
Husk at skrive i din ansøgning, at du så jobbet hos Ofir