A Pilot Study on the Collection and Computational Analysis of Linguistic Differences Amongst Men and Women in a Kuwaiti Arabic WhatsApp Dataset

Hesah Aldihan, Robert J. Gaizauskas, Susan Fitzmaurice

Introduction

Sociolinguistic research has long explored how gender influences communication, yet much of this work focuses on Western contexts and the English language. In the Arab world, Gulf Arabic dialects—specifically Kuwaiti Arabic (KA)—remain under-researched and are often classified as "low-resource" in the field of Natural Language Processing (NLP). This study addresses this gap by analyzing a newly compiled dataset of mixed-gender WhatsApp book club conversations in Kuwait, examining both the volume and the stylistic nuances of digital communication.

Methodology and Dataset

The researchers archived nine months of chat history from three Kuwaiti WhatsApp book clubs, providing a naturally occurring yet relatively formal conversational environment. The raw data was preprocessed using CAMeL tools for tokenization and normalization, and participants were anonymized and labeled by gender.

The resulting dataset consists of 4,479 "turns" (individual messages or sequences). As shown in the descriptive statistics below, the dataset includes participation from 28 women and 14 men, providing a robust foundation for comparing linguistic habits.

Descriptive statistics of the Kuwaiti Arabic WhatsApp dataset, categorized by gender.

To analyze this data, the team employed a two-pronged approach:

  1. Quantitative Analysis: Using the Mann-Whitney U test to identify statistically significant differences in message frequency and length.
  2. Qualitative Analysis: Investigating specific features such as emoji selection and "word lengthening" (e.g., repeating characters for emphasis).

Quantitative Findings: Parity in Volume

A key finding of the study is that men and women in these book clubs communicate with a similar volume. Statistical analysis revealed no significant difference (p > 0.05) in the number of turns, word counts, or the total number of emojis used. For instance, the p-value for word count was 0.350, and for the number of emojis, it was 0.779. This suggests that in this specific digital "majlis," both genders participate with equal frequency and verbosity.

Qualitative Insights: The Gendered "Flavor" of Communication

While the amount of communication is similar, the style is markedly different. The study identified two primary areas of divergence: emoji usage and lexical "stretching."

Emoji Preferences

Although both genders use emojis at similar rates, their specific choices are highly gendered. Women utilized a significantly wider variety of emoji categories, including various flowers, hearts, and animals. In contrast, men tended to stick to a more limited set of standard smileys and male-character emojis.

Top 10 emojis used by women and men, showing distinct stylistic preferences.

As illustrated in the ranking above, women frequently used expressive emojis like clapping hands and praying hands, while men’s top choices were dominated by the thumbs-up and laughing faces. Notably, 68% of women used heart emojis compared to only 29% of men. Women also utilized a vast array of plant and flower emojis, whereas men used only two types.

Word Lengthening and Lexical Taboos

Women were 1.5 times more likely than men to use lengthened words (e.g., "hiii" instead of "hi") to convey emotion or emphasis. While men primarily reserved this "stretching" for laughter (hhhhh), women applied it to a broader range of greetings and expressive terms. Furthermore, the study noted the presence of "lexical taboos"—words like eyanen ("amazing") are socially coded as feminine in KA, and their use by men is rare due to social perceptions.

Conclusion

This pilot study provides the first publicly available dataset for Kuwaiti Arabic conversational data and highlights that gender differences in digital communication are more qualitative than quantitative. While the volume of participation remains balanced, the "flavor" of the language—expressed through emoji variety and orthographic stretching—reveals distinct gendered patterns. Future work will leverage these features to develop automated gender-classification systems tailored for the Kuwaiti dialect.

Summary updated

Abstract

This study focuses on the collection and computational analysis of Kuwaiti Arabic (KA), which is considered a low resource dialect, to test different sociolinguistic hypotheses related to gendered language use. In this paper, we describe the collection and analysis of a corpus of WhatsApp Group chats with mixed gender Kuwaiti participants. This corpus, which we are making publicly available, is the first corpus of KA conversational data. We analyse different interactional and linguistic features to get insights about features that may be indicative of gender to inform the development of a gender classification system for KA in an upcoming study. Statistical analysis of our data shows that there is insufficient evidence to claim that there are significant differences amongst men and women with respect to number of turns, length of turns and number of emojis. However, qualitative analysis shows that men and women differ substantially in the types of emojis they use and in their use of lengthened words.
BibTeX
@inproceedings{aldihan2022pilot,
  author = {Hesah Aldihan and Robert J. Gaizauskas and Susan Fitzmaurice},
  title = {{A Pilot Study on the Collection and Computational Analysis of Linguistic Differences Amongst Men and Women in a Kuwaiti Arabic WhatsApp Dataset}},
  booktitle = {WANLP @ EMNLP},
  year = {2022},
  doi = {10.18653/v1/2022.wanlp-1.35},
}
Figure