Proceedings of Machine Learning Research 81:1–14, 2018 Conference on Fairness, Accountability, and Transparency

Analyze, Detect and Remove Gender Stereotyping from

Bollywood Movies

Nishtha Madaan [email protected]

Sameep Mehta sameepmeht[email protected]

IBM Research-INDIA

Taneea S Agrawaal taneea14166@iiitd.ac.in

Vrinda Malhotra [email protected]

Aditi Aggarwal aditi16004@iiitd.ac.in

IIIT-Delhi

aspects of society is a well-known phe-

nomenon. In this paper, we focus on

studying such stereotypes and bias in Hindi

movie industry (Bollywood) and propose an

algorithm to remove these stereotypes from

text. We analyze movie plots and posters

for all movies released since 1970. The

gender bias is detected by semantic model-

ing of plots at sentence and intra-sentence

level. Different features like occupation,

show that such bias is not applicable for

movie posters where females get equal im-

portance even though their character has

little or no impact on the movie plot. The

silver lining is that our system was able to

identify 30 movies over last 3 years where

such stereotypes were broken. The next

step, is to generate debiased stories. The

proposed debiasing algorithm extracts gen-

der biased graphs from unstructured piece

ety. Therefore, we believe movies could act as

the proxy to understand how prevalent gender

bias and stereotypes are in any society. In this

paper, we leverage NLP and image understand-

ing techniques to quantitatively study this bias.

To further motivate the problem we pick a small

section from the plot of a blockbuster movie.

”Rohit is an aspiring singer who works as a

salesman in a car showroom, run by Malik (Dalip

Tahil). One day he meets Sonia Saxena (Amee-

ion:

1. Male (Rohit) is portrayed with a profession

& an aspiration

2. Male (Malik) is a business owner

In contrast, the female role is introduced with

no profession or aspiration. The introduction,

itself, is dependent upon another male character

”daughter of”!

One goal of our work is to analyze and quantify

c

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

a holistic study of the corpus. Also, it helps

us in capturing the characteristics exhibited by

male and female class. We have extracted movies

pages of all the Hindi movies released from 1970-

present from Wikipedia. We also employ deep

image analytics to capture such bias in movie

posters and previews.

1.1. Analysis Tasks

We focus on following tasks to study gender bias

in Bollywood.

appearance? How do the descriptions differ in

both of them? How does that indicate gender

stereotyping?

III) Centrality of Male and Female Char-

acters - What is the role of males and females

in movie plots? How does the amount of male

being central or female being central differ? How

does it present a male or female bias?

IV) Mentions(Image vs Plot) - How many

males and females are the faces of the promo-

VI) Singers - Does the same bias occur in

movie songs? How does the distribution of

singers with gender vary over a period of time

for different movies?

VII) Female-centric Movies- Are the movie

stories and portrayal of females evolving? Have

we seen female-centric movies in the recent past?

VIII) Screen Time - Which gender, if any,

has a greater screen time in movie trailers?

IX) Emotions of Males and Females -

Which emotions are most commonly displayed by

the analysis majorly involves information re-

trieval tasks involving a wide variety of prior

work in this area. (Fast et al., 2016) have worked

on gender stereotypes in English fiction particu-

larly on the Online Fiction Writing Community.

The work deals primarily with the analysis of

how males and females behave and are described

in this online fiction. Furthermore, this work

also presents that males are over-represented and

finds that traditional gender stereotypes are com-

dren books (Gooden and Gooden, 2001) and mu-

sic lyrics (Millar, 2008) which found that men

are portrayed as strong and violent, and on the

other hand, women are associated with home and

are considered to be gentle and less active com-

pared to men. These studies have been very

useful to uncover the trend but the derivation

of these analyses has been done on very small

data sets. In some works, gender drives the de-

cision for being hired in corporate organizations

(Dobbin and Jung, 2012). Not just hiring, it has

by putting down female claims of raise requests.

roles as compared to their male counterparts in

image search results of occupations (Kay et al.,

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

becomes an interesting study.

b) The data-set is large. We analyze 4000

movies which cover all the movies since 1970. So

it becomes a good first step to develop compu-

tational tools to analyze the existence of stereo-

types over a period of time.

c) These movies are a reflection of society. It

Our data-set consists of 880 official movie trailers

riaga, 2017).

tify if the actress has an important role in the

movie or not. In Figure 2 it is observed that,

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

Moreover, there is a consistency of this ratio from

1970 to 2017(for almost 50 years)!

2) Cast Appearance in Movie Plot - We

analyze how male cast and female cast have been

addressed. This essentially involves extracting

verbs and adjectives associated with male cast

and female cast. To extract verbs and adjec-

tives linked to a particular cast, we use Stanford

Dependency Parser (De Marneffe et al., 2006).

In Fig ?? and ?? we present the adjectives and

ful and attractive in movie plots.

3) Cast Introductions in Movie Plot - We

analyze how male cast and female cast have been

introduced in the plot. We use OpenIE (Fader

et al., 2011) to capture such introductions by ex-

tracting relations corresponding to a cast. Fi-

nally, on aggregating the relations by gender, we

find that males are generally introduced with a

profession like as a famous singer, an honest po-

lice officer, a successful scientist and so on while

to be successful.

4) Occupation as a stereotype - We per-

Figure 2: Total Cast Mentions showing mentions

of male and female cast. Female mentions are

presented in pink and Male mentions in blue

females are represented. To perform this analy-

sis, we collated an occupation list from multiple

sources over the web comprising of 350 occupa-

tions. We then extracted an associated ”noun”

tag attached with cast member of the movie using

Stanford Dependency Parser (De Marneffe et al.,

figure it is clearly evident that, males are given

higher level occupations than females. Figure 1

presents a combined plot of percentages of male

and female having the same occupation. This

plot shows that when it comes to occupation like

”teacher” or ”student”, females are high in num-

ber. But for ”lawyer” and ”doctor” the story is

totally opposite.

5) Singers and Gender distribution in

Soundtracks - We perform an analysis on how

page and then aggregated the numbers of songs

sung by males and females over the years. In

Figure 4, we report the aforementioned distribu-

tion for recent years ranging from 2010-2017. We

observe that the gender-gap is almost consistent

over all these years.

Please note that currently this analysis only takes

into account the presence or absence of female

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

presented on X-axis and Male dialogues on Y-axis

construct a word graph for each sentence by

resent the dependencies extracted using depen-

dency parser across all sentences in the movie

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

2. Study of bias using word embeddings

for male and female cast for each movie contains

a set of nodes connected to them. These nodes

of (shoots, violent, scientist, beats). In this fash-

ion we assign a context vector to each cast node.

The main idea behind assigning a context vector

is to analyze the differences between contexts for

male and female.

testing data. We fit the training data using a K-

Nearest Neighbor with varying K. We study the

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

test data. In Figure 5, we show the accuracy val-

ues for varying values of K. While studying bias

using word embeddings by constructing a con-

ciated with males and females. Although we did

an individual analysis for each of the aforemen-

tioned intra-sentence level tasks, but the com-

bined inter-sentence level analysis makes the ar-

gument of existence of bias stronger. Note the

key point is not that the accuracy goes up as the

training data is increased. The key point is that

since the gender bias is high, the small training

data has enough information to classify correctly

60-70% of the cases.

responses having a positive confidence score.

movie is biased towards a female on advertising

material like posters, and have a small or incon-

Figure 10: Percentage of female-centric movies

over the years

While 80% of the movie plots have more male

on posters very prominently. Also, when we look

at Image and Plot mentions, we observe that

in 56% of the movies, female plot mentions are

less than half the male plot mentions while in

posters this number is around 30%. Our sys-

tem detected 289 female-centric movies, where

this stereotype is being broken. To further study

this, we plotted centrality of females and their

mentions in plots over the years for these 289

movies. Figure 10 shows that both plot men-

Analyze, Detect and Remove Gender Stereotyping from Bollywood Movies

the trend for percentage of female-centric movies.

It is enlightening to see that the percentage shows

a rising trend. Our system discovered at least 30

movies in last three years where females play cen-

tral role in plot as well as in posters. We also note

that over time such biases are decreasing - still

far away from being neutral but the trend is en-

movie preview dataset and obtain information re-

garding the presence/absence of a male/female in

the frame. If any person is present in the frame

we then find out the emotion displayed by the

person. The emotion displayed can be one of an-

gry, disgust, fear, happy, neutral, sad, surprise.

Note that there can be more than one person

detected in a single frame, in that instance, emo-

tions of each person is detected. We then aggre-

gate the results to analyze the following tasks on

% compared to 68.5 % of the time for male char-

acters.

we analyze the emotions most commonly exhib-

happy than male characters every year. These re-

sults correspond to the gender stereotypes which

Figure 11: Percentage of screen-on time for males

and females over the years

Figure 12: Year wise distribution of emotions dis-

played by males and females.

for other emotions because we could not see any

proper trend exhibited by them.

4. Algorithm for Bias Removal

System - DeCogTeller

developed a specialized module to handle occupa-

tions. Very often, gender bias shows in assigned

occupation { (Male, Doctor), (Female, Nurse)}

or { (Male, Boss), (Female, Assistant)}.

In Figure 13 we give a holistic view of our sys-

tem DeCogTeller which is described in a detailed