Named entity recognition in Malayalam using fuzzy support vector machine
G. Lakshmi, Janu R. Panicker, S. Meera · 2016
Named entities in a text are the atomic elements that represent the name of something, and the name can be a person name, name of an organization, name of a place or location etc. In the field of information extraction the identification and classification of named entities are quite an important task. The identification and classification of the named entities in a text into some pre-defined classes is known as named entity recognition. The commonly used pre-defined classes are the person name, place, name of organization, date, numericals, measurements, and so on. It is a subtask of information extraction and many other natural language processing like text summarization, text categorization, question answering etc. In the case of language processing of Malayalam documents no effective tools are readily available. Through this paper a named entity recognition for Malayalam language is presented. The system proposed follows machine learning approach using support vector machine integrated with fuzzy module for improving the performance. The design is a kind of One-Against-All-Multi classification technique to solve the ambiguity caused by traditional SVM classifier. The system is based on contextual semantic rules and linguistic grammar rules. Malayalam NER is a challenging work as there is no specific feature for identifying named entities like capitalization feature in English. Also no named entity tagged corpus for Malayalam language is available for training the system. The system defines four primary named entity classes, i.e, Name, Organization, Place and Date.