An official website of the United States government
As natural language processing methods are increasingly deployed in real-world scenarios such as healthcare, legal systems, and social science, it becomes necessary to recognize the role they potentially play in shaping social biases and stereotypes. Previous work has revealed the presence of social biases in widely used word embeddings involving gender, race, religion, and other social constructs. While some methods were proposed to debias these word-level embeddings, there is a need to perform debiasing at the sentence-level given the recent shift towards new contextualized sentence representations such as ELMo and BERT. In this paper, we investigate the presence of social biases in sentence-level representations and propose a new method, Sent-Debias, to reduce these biases. We show that Sent-Debias is effective in removing biases, and at the same time, preserves performance on sentence-level downstream tasks such as sentiment analysis, linguistic acceptability, and natural language understanding. We hope that our work will inspire future research on characterizing and removing social biases from widely adopted sentence representations for fairer NLP.
more »
« less
When do Word Embeddings Accurately Reflect Surveys on our Beliefs About People?
Social biases are encoded in word embeddings. This presents a unique opportunity to study society historically and at scale, and a unique danger when embeddings are used in downstream applications. Here, we investigate the extent to which publicly-available word embeddings accurately reflect beliefs about certain kinds of people as measured via traditional survey methods. We find that biases found in word embeddings do, on average, closely mirror survey data across seventeen dimensions of social meaning. However, we also find that biases in embeddings are much more reflective of survey data for some dimensions of meaning (e.g. gender) than others (e.g. race), and that we can be highly confident that embedding-based measures reflect survey data only for the most salient biases.
more »
« less
- Award ID(s):
- 1939579
- PAR ID:
- 10206193
- Date Published:
- Journal Name:
- Association for Computational Linguistics
- Volume:
- Proceedings of the 58th Annual Meeting of the Association for Comput
- Page Range / eLocation ID:
- 4392 to 4415
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
While contextualized word representations have improved state-of-the-art benchmarks in many NLP tasks, their potential usefulness for social-oriented tasks remains largely unexplored. We show how contextualized word embeddings can be used to capture affect dimensions in portrayals of people. We evaluate our methodology quantitatively, on held-out affect lexicons, and qualitatively, through case examples. We find that contextualized word representations do encode meaningful affect information, but they are heavily biased towards their training data, which limits their usefulness to in-domain analyses. We ultimately use our method to examine differences in portrayals of men and women.more » « less
-
With the increase of natural disasters all over the world, we are in crucial need of innovative solutions with inexpensive implementations to assist the emergency response systems. Information collected through conventional sources (e.g., incident reports, 911 calls, physical volunteers, etc.) are proving to be insufficient [1]. Responsible organizations are now leaning towards research grounds that explore digital human connectivity and freely available sources of information. U.S. Geological Survey and Federal Emergency Management Agency (FEMA) introduced Critical Lifeline (CLL) s which identifies the most significant areas that require immediate attention in case of natural disasters. These organizations applied crowdsourcing by connecting digital volunteer networks to collect data on the critical lifelines from data sources including social media [3], [4], [5]. In the past couple of years, during some of the deadly hurricanes (e.g., Harvey, IRMA, Maria, Michael, Florence, etc.), people took on different social media platforms like never seen before, in search of help for rescue, shelter, and relief. Their posts reflect crisis updates and their real-time observations on the devastation that they witness. In this paper, we propose a methodology to build and analyze time-frequency features of words on social media to assist the volunteer networks in identifying the context before, during and after a natural disaster and distinguishing contexts connected to the critical lifelines. We employ Continuous Wavelet Transform to help create word features and propose two ways to reduce the dimensions which we use to create word clusters to identify themes of conversations associated with stages of a disaster and these lifelines. We compare two different methodologies of wavelet features and word clusters both qualitatively and quantitatively, to show that wavelet features can identify and separate context without using semantic information as inputs.more » « less
-
Contextual word embeddings such as BERT have achieved state of the art performance in numerous NLP tasks. Since they are optimized to capture the statistical properties of training data, they tend to pick up on and amplify social stereotypes present in the data as well. In this study, we (1) propose a template-based method to quantify bias in BERT; (2) show that this method obtains more consistent results in capturing social biases than the traditional cosine based method; and (3) conduct a case study, evaluating gender bias in a downstream task of Gender Pronoun Resolution. Although our case study focuses on gender bias, the proposed technique is generalizable to unveiling other biases, including in multiclass settings, such as racial and religious biases.more » « less
-
VERB: Visualizing and Interpreting Bias Mitigation Techniques Geometrically for Word RepresentationsWord vector embeddings have been shown to contain and amplify biases in the data they are extracted from. Consequently, many techniques have been proposed to identify, mitigate, and attenuate these biases in word representations. In this paper, we utilize interactive visualization to increase the interpretability and accessibility of a collection of state-of-the-art debiasing techniques. To aid this, we present the Visualization of Embedding Representations for deBiasing (“VERB”) system, an open-source web-based visualization tool that helps users gain a technical understanding and visual intuition of the inner workings of debiasing techniques, with a focus on their geometric properties. In particular, VERB offers easy-to-follow examples that explore the effects of these debiasing techniques on the geometry of high-dimensional word vectors. To help understand how various debiasing techniques change the underlying geometry, VERB decomposes each technique into interpretable sequences of primitive transformations and highlights their effect on the word vectors using dimensionality reduction and interactive visual exploration. VERB is designed to target natural language processing (NLP) practitioners who are designing decision-making systems on top of word embeddings, and also researchers working with the fairness and ethics of machine learning systems in NLP. It can also serve as a visual medium for education, which helps an NLP novice understand and mitigate biases in word embeddings.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.18653/v1/2020.acl-main.405
