Title:
Inducing a Domain-Independent Sentiment Lexicon in Malay

Speaker:
ZAINUDIN, Suhaila (UKM) on behalf of OMAR, Nazlia (UKM)

Abstract:
Sentiment analysis (SA) is a discipline that involves the detection of user sentiment, emotion and opinion within natural language text. Lexicon-based SA models make use of a sentiment lexicon for SA tasks, which is a linguistic resource that comprises a priori knowledge about subjective words tagged with their underlying sentiment polarity. A sentiment lexicon and greatly contributes to SA tasks. This is evident in the emergence of the large number of research works that have aimed to develop automated sentiment lexicon induction algorithms. However, most works primarily consider the English language; this is attributable to the availability of a sufficient amount resources and tools for this language. On the other hand, this is not the case for low-resource languages such as Malay. Research focused on sentiment lexicon induction algorithms in particular, and SA in general, in the Malay language, is lacking. This has brought up the motivation to develop a sentiment lexicon induction algorithm for the Malay language. We first map WordNet Bahasa onto the English WordNet to construct a multilingual word network, and then use a dictionary-based approach and a supervised classifier for classifying words with their sentiment polarities. The algorithm was evaluated against the General Inquirer lexicon, demonstrating that it performs with accuracy that is comparable to human accuracy.

Extended Abstract:
pdf