TL;DR
This paper investigates how politicians in India use code-mixing (CM) and code-switching (CS) between Telugu dialects and other languages (English, Hindi/Urdu) in their speeches. The authors introduce a novel annotated corpus of political speeches and propose guidelines for analyzing these linguistic phenomena, revealing how audience, ideology, and context influence a speaker's language choices for political gain.
What Problem Does This Paper Address?
Political discourse is a rich area for studying power dynamics, especially in multilingual societies like India. Politicians strategically employ language to connect with diverse audiences and build relationships. However, understanding the nuances of how they use multiple languages and dialects (code-mixing and code-switching) within their speeches, particularly in a specific regional context like Telugu, has been underexplored. Existing research often focuses on social media or specific language pairs (like Hindi-English), leaving a gap in analyzing these phenomena in formal political settings and across different Telugu dialects. This paper aims to fill this gap by analyzing CM and CS in Telugu political speeches to understand the factors driving their usage and their impact on political communication.
Methods
The authors developed a methodology for analyzing code-mixing and code-switching in political speeches. They manually transcribed political speeches and annotated them at the word level for dialectal code-mixing (DCM), language level code-mixing (LCM), and code-switching (CS). They established specific guidelines for handling dialectal variations, categorizing them into vowel and consonant rules (addition, deletion, replacement) and syllable rules. For language level code-mixing, they focused on instances where English or Hindi/Urdu root words were integrated with Telugu morphology. Code-switching was identified at the sentence or phrase level when non-Telugu languages were used. This systematic annotation process allowed them to quantify the occurrences of CM and CS across different communicative contexts.
Datasets
The paper introduces a newly created corpus of political speeches in Telugu. This corpus consists of 6 speeches from two prominent politicians: Mr. K. Chandrasekhar Rao (KCR) and Mr. Chandra Babu Naidu (CBN). The speeches were collected from three different social settings: public meetings in Telangana, felicitations of dignitaries, and discussions on capital development. The total corpus comprises 1134 sentences and approximately 10,000 words, transcribed from YouTube videos. This dataset is unique as it specifically targets political discourse and includes dialectal variations within Telugu.
Key Insights and Ideas
- Strategic Linguistic Choices: CM and CS are not random but are strategic tools used by politicians to connect with specific audiences, convey social identities, and achieve political objectives.
- Audience Matters: The composition of the audience significantly influences the level of dialectal code-mixing. When addressing local Telangana residents, speakers tend to use more of the Telangana dialect to foster a sense of connection.
- Ideology Plays a Role: A speaker's political ideology can also impact their language choices. For instance, KCR's use of Hindi/Urdu in speeches, which are often integrated into the Telangana dialect, might be to establish a better connection with an audience influenced by these languages.
- Context is King: The communicative event or context dictates the linguistic strategies employed. For national-level discussions or when addressing officials who might not be fluent in the local language, a shift towards Modern Standard Telugu (MST) or English is observed.
- Dialectal Nuances are Key: The paper highlights specific phonetic and morphological differences between Standard Telugu and the Telangana dialect, providing a framework for analyzing these subtle but important variations.
Key Results
The study found varying levels of dialectal code-mixing, language level code-mixing, and code-switching across the analyzed speeches and communicative events.
- Communicative Event 1 (Public Meetings in Telangana): Both speakers exhibited relatively higher levels of Telangana dialect usage to connect with the local audience. KCR used more Telangana dialect than CBN, reflecting his role in fighting for a separate Telangana state.
- Communicative Event 2 (Felicitations): In this context, Modern Standard Telugu (MST) was more prevalent, possibly due to the presence of MLAs and parliament members from different regions. National topics also led to lower levels of language mixing and switching.
- Communicative Event 3 (Capital Development): English usage was higher in this event, likely because government officials and collectors might not be fully conversant in the local language. However, KCR's speech still showed prominent Telangana dialect usage due to the presence of local politicians.
Limitations and Future Work
The authors acknowledge that the corpus size, while sufficient for their rule-based and statistical observations, could be expanded for more robust analysis. They also note that their work focuses on a specific set of rules for dialectal variations and that further research could explore these in greater depth.
Future work suggested includes:
- Developing dialectal speech recognition systems using the annotated corpus.
- Extending the analysis of dialectal variations to other Dravidian languages like Tamil and Kannada.
- Developing Machine Translation systems that can handle multiple dialects within a language.
- Expanding the dataset and examining additional factors influencing code-switching and code-mixing.
Summary updated