Socially Aware Bias Measurements for Hindi Language Representations

Abstract

Trigger warning: This paper contains examples of stereotypes and other harms that could be offensive and triggering to individuals.Language representations are efficient tools used across NLP applications, but they are strife with encoded societal biases. These biases are studied extensively, but with a primary focus on English language representations and biases common in the context of Western society. In this work, we investigate biases present in Hindi language representations with focuses on caste and religion-associated biases. We demonstrate how biases are unique to specific language representations based on the history and culture of the region they are widely spoken in, and how the same societal bias (such as binary gender-associated biases) is encoded by different words and text spans across languages. The discoveries of our work highlight the necessity of culture awareness and linguistic artifacts when modeling language representations, in order to better understand the encoded biases.

Re-contextualizing Fairness in NLP: The Case of India

“…There is recent expanding fairness research to languages like Arabic (Lauscher et al, 2020), Japanese (Takeshita et al, 2020), Hindi, Bengali, Telugu (Pujari et al, 2019Malik et al, 2021). Evidence of cultural biases for different countries have also been recorded (Ghosh et al, 2021) in LMs.…”

“…Religion Religious biases have been studied in NLP (Dev et al, 2020;Nadeem et al, 2020;Abid et al, 2021), however the social disparities and stereotypes about various religious groups differ significantly in India from the West, (Malik et al, 2021). For e.g.…”

“…Generating Candidate Associations: We build the set of candidate association tuples (i, t) using identity terms described in Section 4 for religion and region. We then create a list of tokens based on prior work (Malik et al, 2021;Nangia et al, 2020;Nadeem et al, 2020); including lists of professions, subjects of study (history, science, etc. ), action-verbs, and adjectives for behaviour, socio-economic status, food habits, and clothing preferences.…”

“…Multilingual fairness research: NLP Fairness research relies on bias evaluation resources and while we present such resources for the Indian context, we limited our focus to only English. It is crucial to expand this effort into Indian languages, along the lines of recent work on Hindi, Bengali, and Telugu (Malik et al, 2021;Pujari et al, 2019). This is especially important since biases may manifest differently in data and models for different languages.…”

“…This is especially troubling for India, a pluralistic nation of 1.4 billion people, with fast-growing investments in NLP from the government 3 , and the private sector 4 . There is commendable recent work on NLP fairness in Indian languages like Hindi, Bengali, Telugu (Pujari et al, 2019;Malik et al, 2021;Gupta et al, 2021). But, for a nation with many religions, ethnicities, and cultures, recontextualizing NLP fairness needs to account for the various axes of social disparities in the Indian society, their proxies in language data, the disparate NLP capabilities in Indian languages, and the (lack of) resources for bias evaluation.…”