Obfuscating Gender in Social Media Writing

@inproceedings{Reddy2016ObfuscatingGI,
  title={Obfuscating Gender in Social Media Writing},
  author={Sravana Reddy and Kevin Knight},
  booktitle={NLP+CSS@EMNLP},
  year={2016}
}
The vast availability of textual data on social media has led to an interest in algorithms to predict user attributes such as gender based on the user’s writing. These methods are valuable for social science research as well as targeted advertising and profiling, but also compromise the privacy of users who may not realize that their personal idiolects can give away their demographic identities. Can we automatically modify a text so that the author is classified as a certain target gender… 

Tables from this paper

Profiling Without the Writer ’ s Text

TLDR
It is shown that linguistic cues in public comments directed at a user are sufficient for an accurate inference of that user’s gender, age, religion, diet, and even personality traits.

Writer Profiling Without the Writer's Text

TLDR
It is shown that linguistic cues in public comments directed at a user are sufficient for an accurate inference of that user’s gender, age, religion, diet, and even personality traits.

How well can machine learning predict demographics of social media users

TLDR
Knowing the demographics in a data sample can aid in addressing issues of bias and population representation, so that existing societal inequalities are not exacerbated.

Generating Clues for Gender based Occupation De-biasing in Text

TLDR
This paper presents the first system that discovers the possibility that a given text portrays a gender stereotype associated with an occupation, and offers counter-evidences of opposite gender also being associated with the same occupation in the context of user-provided geography and timespan.

Text Style Transfer for Bias Mitigation using Masked Language Modeling

TLDR
This paper presents a text-style transfer model that can be trained on non-parallel data and be used to automatically mitigate bias in textual data and shows that this technique produces better content preservation whilst maintaining good style transfer accuracy.

Controlling Human Perception of Basic User Traits

TLDR
This pilot study measures the extent to which human perception of basic user trait information – gender and age – is controllable through text and shows that it is practically feasible for multiple applications such as text generation, text summarization or machine translation to be tailored to specific traits and perceived as such.

Stylistic Language Dataset with Multiple Personas

TLDR
PASTEL, the parallel and annotated stylistic language dataset, that contains 41K parallel sentences (8.3K parallel stories) annotated across different personas, is released and a simple supervised model with the authors' parallel text outperforms the unsupervised models using nonparallel text in style transfer.

Jibes & Delights: A Dataset of Targeted Insults and Compliments to Tackle Online Abuse

TLDR
A Reddit-based dataset, consisting of 68,159 insults and 51,102 compliments targeted at individuals instead of targeting a particular community or race, is contributed and it is concluded that the transfer task is challenging.

Towards A Friendly Online Community: An Unsupervised Style Transfer Framework for Profanity Redaction

TLDR
This work designs a Retrieve, Generate and Edit unsupervised style transfer pipeline to redact the offensive comments in a word-restricted manner while maintaining a high level of fluency and preserving the content of the original text.

It’s going to be okay: Measuring Access to Support in Online Communities

TLDR
A new dataset and method for identifying supportive replies and new methods for inferring gender from text and name are created to create a new massive corpus of 102M online interactions with gender-labeled users, each rated by degree of supportiveness.
...

References

SHOWING 1-10 OF 47 REFERENCES

Discriminating Gender on Twitter

TLDR
The construction of a large, multilingual dataset labeled with gender is described and statistical models for determining the gender of uncharacterized Twitter users are investigated, and several different classifier types are explored.

Why Gender and Age Prediction from Tweets is Hard: Lessons from a Crowdsourcing Experiment

TLDR
In this paper, insights from sociolinguistics and data collected through an online game are combined to underline the importance of approaching age and gender as social variables rather than static biological variables.

Gender identity and lexical variation in social media

TLDR
Pairing computational methods and social theory offers a new perspective on how gender emerges as individuals position themselves relative to audiences, topics, and mainstream gender norms.

What's in a Name? Using First Names as Features for Gender Inference in Twitter

TLDR
A thorough investigation of the link between gender and first name in English tweets is performed and a novel way of obtaining gender-labels for Twitter users that does not require analysis of the user’s profile or textual content is developed.

Weakly Supervised User Profile Extraction from Twitter

TLDR
This paper presents a weaklysupervised approach to user profile extraction from Twitter that takes into account network information, a unique opportunity offered by social media.

A Machine Learning Approach to Twitter User Classification

TLDR
This paper automatically infer the values of user attributes such as political orientation or ethnicity by leveraging observable information such as the user behavior, network structure and the linguistic content of the user’s Twitter feed through a machine learning approach.

Inferring Latent User Properties from Texts Published in Social Media

TLDR
This work examines individual tweets to detect emotions and opinions emanating from them, and analyzes all the tweets published by a user to infer latent traits of that individual, focusing on Ekman’s six emotions.

Classifying latent user attributes in twitter

TLDR
A novel investigation of stacked-SVM-based classification algorithms over a rich set of original features, applied to classifying these four user attributes, as distinct from the other primarily spoken genres previously studied in the user-property classification literature.

Discovering User Attribute Stylistic Differences via Paraphrasing

TLDR
This study aims to find linguistic style distinctions across three different user attributes: gender, age and occupational class and shows their predictive power in user profiling, conformity with human perception and psycholinguistic hypotheses, and potential use in generating natural language tailored to specific user traits.

Improving Gender Classification of Blog Authors

TLDR
Empirical evaluation using a real-life blog data set shows that these two techniques improve the classification accuracy of the current state-of-the-art methods significantly.