Using SVM with uneven margins to extract acronym-expansions

Extracting acronyms and their expansions from plain text is an important problem in text mining. Previous research work shows that it can be solved using machine learning approaches. That is, converting the problem of acronym extraction into a binary classification one. We investigated the classification problem and found that the classes are highly unbalanced. So we try to tackle the problem using SVM with uneven margins. Experimental results showed that our approach can get better results than baseline methods of using rules and conventional SVM models. Experimental results also showed how uneven margins classifier making the tradeoff between the extracting precision and recall.