Sentiment Analysis for Indian Languages (Code Mixed) Shared Task

SAIL CodeMixed 2017


Artificial Intelligence



India is a linguistic area with one of the longest histories of contact, influence, use, teaching and learning of English-in-diaspora in the world (Kachru and Nelson, 2006). Thus, a huge number of Indians active on the internet are able in English communication to some degree. India also enjoys huge diversity in language. Apart from Hindi, it has several regional languages that are the primary tongue of people native to the region. This is to the extent that social media including Facebook, WhatsApp, Twitter, etc. contain more than one language, and such phenomena are called code-mixing and code-switching. On the other side, the evolution of sentiments from such social media texts have also created many new opportunities for information access and language technology, but also many new challenges, making it one of the prime present-day research areas. Sentiment analysis in code-mixed data has several real-life applications in opinion mining from social media campaign to feedback analysis.
Linguistic processing of such social media dataset and its sentiment analysis is a difficult task. Till date, most of the experiments have been performed on identifying the languages (Bali et al., 2014; Das and Gamback, 2014), parts-of-speech tagging (Ghosh et al., 2016), etc. Few tasks also have been started on the sentiment analysis of code-mixed data such as Hindi-English (Joshi et al., 2016). Therefore, we believe that it is the best place to bring more research attention towards developing language technologies for identifying sentiments from Indian social media texts.
Main goal of this task is to identify the sentence level sentiment polarity of the code-mixed dataset of Indian languages pairs (Hi-En, Ben-Hi-En) collected from Twitter, Facebook, and WhatsApp. Each of the sentences is annotated with language information as well as polarity at the sentence level. The participants will be provided development, training and test dataset.
Each participating team will be allowed to submit two systems for each of the language pairs, and the best result will be considered as final. The final evaluation will be performed based on the macro-averaged F1-measure. The python code for the evaluation will be provided by the organizers. Initially, each of participating teams will have access to the development and training data. Later, the unlabeled test data will be provided, and the teams have to submit the results within 24 hours. There will be no distinction between constrained and unconstrained systems, but the participants will be asked to report what additional resources they have used for each of their submitted runs.