Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/rimo02/gender-bias


https://github.com/rimo02/gender-bias

Last synced: 3 days ago
JSON representation

Awesome Lists containing this project

README 

        
# Problem Statement - 

Textual similarity is an important problem in NLP.  The broad goal of textual similarity is to measure the extent of similarity between a given pair of text fragment (sentence/paragraph etc) based on a specific aspect/criterion (such as topic, sentiment, etc).

In this problem, you will be given a set of pairs of text and the end goal is to determine if they are similar or dissimilar based on some form of gender bias present in the text. The forms of the gender bias could be: i) firstness: where a gender (male/female) is always mentioned first, ii) stereotyping of a particular gender, iii) subordination: where the text reflects a gender is subordinate compared to other.  



Training data format: The training data will contain a set of text pairs (p1 p2) along with

their labels (0 or 1), where 0 indicates p1 and p2 can be both biased or both unbiased,

similarly 1 indicates if one is biased but the other is unbiased. There are two files for training

data. The first file (name: text-and-id) contains the text (2 nd column) and its unique id (1 st

column) in each line, while the second file (name: pairs-label-training) contains the ids of

the two text fragments and the corresponding label (0/1) in each line.



Desired output: Given a text pair (p1 p2), your goal would be to mark this as 0 or 1. The

meaning of 0 and 1 is the same as in the training data.



Test data: Test data contains a set of text pairs and you have to produce the 0/1 tag for

each pair in the test data.