Why should we care about code-switching when analyzing social listening data?

Code-switching (see [1]) refers to the ability to switch between different linguistic and cultural styles in communication, often depending on the context and the audience. We all code-switch to adapt our communication in different contexts.

Linguistic code-switching is mostly used within bilingual and multilingual communities, and there are?many reasons?to use this method, such as the need to fit in with a group, as a force of habit, or to convey thoughts and concepts that might be easier to explain in a specific language (see [2]).

Examples of code-switching include:?A Spanish and English speaker saying, "Pero WHY do I have to go a la casa?" Or, "But WHY do I have to go home?"?or a speaker of Black Vernacular English saying, "We finna get to school on time today," instead of "we are going to get to school on time today." (see [1])

When analyzing conversations, especially online, we must be aware of codeswitching, especially when trying to understand concerns and narratives within and around minority groups and communities. Code-switching can serve several purposes (see [3-4]), including:

1. To fit in with the majority culture: Members of minority communities may switch to the dominant language or cultural norms in order to fit in and be accepted by the majority culture.
2. To maintain cultural identity: Code-switching can also be used to maintain cultural identity, especially when members of a minority community are surrounded by the majority culture. By switching to their own cultural mode of communication, they are able to express their unique cultural identity and maintain a sense of belonging to their community.
3. To navigate social hierarchies: Code-switching can also be used to navigate social hierarchies, especially in professional and educational settings. Members of minority communities may switch to the dominant language or cultural norms in order to appear more professional, knowledgeable or competent, or to gain respect and credibility.
4. To communicate with different audiences: Code-switching can also be used to communicate effectively with different audiences, such as family, friends, coworkers, or customers. By switching to different linguistic or cultural modes of communication, members of minority communities are able to effectively communicate and connect with different people, even across cultural and linguistic barriers.

How people use codeswitching online is also dependent on the culture of the communities the users belong to, and has been studied and described as early as the first IRC and usenet groups were popular (hint: since the early 2000s!) (see [5-6]).

Code-switching and misinformation

In the context of the concerns, questions, misinformation and the spread of narratives online, code-switching can be used to promote false or misleading information and narratives in a subtle way.

• For example, individuals who are spreading misinformation may code-switch between different language styles, depending on the audience they are trying to reach. They may use technical language to appeal to experts, or emotional language to appeal to the general public. By code-switching, they can make their false or misleading information appear more credible and appealing to different groups of people.
• Code-switching can also be used to spread false or misleading narratives in a disguised way. For example, individuals who are spreading false or misleading narratives may code-switch between different cultural styles, depending on the audience they are trying to reach. They may use language and symbols that are familiar to certain groups of people, while appearing to be discussing mainstream topics. (see [7]).

Codeswitching and machine learning applications

In the context of AI-aided social media analysis, code-switching refers to the fluid alternation between multiple languages in a single post. This poses a challenge to analyze narratives and conversations in bilingual or multilingual spaces and communities (see [8]). It basically poses a challenge across a variety of analytical tasks in social media analysis:

1. Difficulty in detecting code-switching: Code-switching can be subtle and disguised, making it difficult to detect and identify. This can result in false positive or false negative results in social media analysis, leading to inaccurate conclusions about the spread of information and narratives.
2. Inaccurate sentiment analysis: Sentiment analysis algorithms may struggle to accurately identify the sentiment of code-switched content, especially if the switch is from a language or style that is unfamiliar to the algorithm. This can result in incorrect sentiment scores, leading to inaccurate conclusions about the attitudes and opinions of the audience.
3. Challenges in categorizing content: Code-switching can make it difficult to categorize content accurately, especially if the content is switching between different topics, themes or genres. This can result in incorrect categorization of content, leading to inaccurate conclusions about the spread of information and narratives. (see [9]).
4. Limitations in topic modeling: Topic modeling algorithms may struggle to identify topics in code-switched content, especially if the switch is from a language or style that is unfamiliar to the algorithm. This can result in incorrect topic scores, leading to inaccurate conclusions about the spread of information and narratives.

Why do infodemic managers and social media analysts need to understand code-switching?

Code-switched speech makes up a relatively large portion of bilingual discourse in general – around 20% (see [10]), and therefore must be considered by an analyst that is working to analyze narratives and sentiment in bilingual and multilingual spaces:

1. To accurately interpret and analyze social media data: Code-switching can impact the interpretation of social media data, especially when analyzing content posted by members of minority communities. By understanding the context and motivations behind code-switching, social media analysts can gain a deeper understanding of the messages and narratives being communicated, and avoid making incorrect assumptions or conclusions.
2. To identify and mitigate potential biases in analysis: Code-switching can also introduce biases in social media analysis, especially if the data is being analyzed in a language or cultural mode that is different from the original language or cultural mode used by the community being studied. By being aware of code-switching and its potential effects, social media analysts can take steps to mitigate these biases, such as using multiple languages or cultural modes of analysis, or considering the cultural context of the data.
3. To enhance the accuracy of sentiment analysis: Code-switching can also impact the accuracy of sentiment analysis, especially when analyzing social media data posted by members of minority communities. By understanding the context and motivations behind code-switching, social media analysts can gain a deeper understanding of the emotions and attitudes being expressed, and improve the accuracy of sentiment analysis.
4. To understand the complexity of online communication: Code-switching is just one aspect of the complex and nuanced nature of online communication, especially among minority communities. By being aware of code-switching and its potential effects, social media analysts can gain a deeper appreciation for the complexities of online communication and the challenges involved in accurately analyzing and interpreting social media data.

Further reading

1. Code-switching on study.com
2. Linguistic code-switching: what it is and why it happens
3. The Costs of Code-Switching on Harvard Business Review.
4. Weeks R, White S, Hartner AM, Littlepage S, Wolf J, Masten K, Tingey L. COVID-19 Messaging on Social Media for American Indian and Alaska Native Communities: Thematic Analysis of Audience Reach and Web Behavior. JMIR infodemiology. 2022 Nov 25;2(2):e38441.
5. Cárdenas-Claros MS, Isharyanti N. Code-switching and code-mixing in internet chatting: Between’yes,” ya,’and’si’-a case study. The Jalt Call Journal. 2009;5(3):67-78.
6. Paolillo JC. “Conversational” codeswitching on usenet and internet relay chat. Language@ Internet. 2011 Dec 24;8(3).
7. Yingchen K, Montefrio MJ, Tandoc E. Navigating discipline and indulgence: the performance of contradiction on Instagram food posts in the Philippines. Food, Culture & Society. 2022 Aug 28:1-21.
8. Co NA, Estuar MR, Tan HC, Tan AS, Abao R, Aureus J. Development of Bilingual Sentiment and Emotion Text Classification Models from COVID-19 Vaccination Tweets in the Philippines. InSocial Computing and Social Media: Design, User Experience and Impact: 14th International Conference, SCSM 2022, Held as Part of the 24th HCI International Conference, HCII 2022, Virtual Event, June 26–July 1, 2022, Proceedings, Part I 2022 Jun 16 (pp. 247-266). Cham: Springer International Publishing.
9. Ahmad GI, Singla J, Anis A, Reshi AA, Salameh AA. Machine Learning Techniques for Sentiment Analysis of Code-Mixed and Switched Indian Social Media Text Corpus-A Comprehensive Review. International Journal of Advanced Computer Science and Applications. 2022;13(2).
10. Tomi? A, Kroff JR. Expecting the unexpected: Code-switching as a facilitatory cue in online sentence processing. Bilingualism: Language and Cognition. 2022 Jan;25(1):81-92.