Pattern Learning and Active Feature Selection for Word Sense Disambiguation

Pattern Learning and Active Feature Selection for Word Sense Disambiguation Rada F. MIHALCEA Southern Methodist University Dallas, Texas, 75275-0122 rada@seas.smu.edu Abstract We present here the main ideas of the algo- rithm employed in the SMUls and SMU aw sys- tems. These systems have participated in the SENSEVAL-2 competition attaining the best per- formance for both English all words and En- glish lexical sample tasks1. The algorithm has two main components (1) pattern learning from available sense tagged corpora (SemCor) and dic- tionary definitions (WordNet), and (2) instance based learning with active feature selection, when training data is available for a particular word. 1 Introduction It is well known that WSD constitutes one of the hardest problems in Natural Language Pro- cessing, yet is a necessary step in a large range of applications including machine translation, knowledge acquisition, coreference, information retrieval and others. This motivates a continu- ously increasing number of researchers to develop WSD systems and devote time to finding solu- tions for this challenging problem. The system presented here was initially de- signed for the semantic disambiguation of all words in open text. The SENSEVAL competitions created a good environment for supervised sys- tems and this encouraged us to improve our sys- tem with the capability of incorporating larger training data sets when provided. There are two important modules in this sys- tem. The first one uses pattern learning relying on large sense tagged corpora to tag all words in open text. The second module is triggered only for the words with large training data, as was the case with the words from the lexical sample tasks. It uses an instance based learning algorithm with active feature selection. 1This is in conformity with the original ranking, fol- lowing the evaluation of systems answers submitted before deadline. 127 Dan I. MOLDOVAN University of Texas at Dallas Richardson, Texas, 75083-0688 moldovan@utdallas.edu To our knowledge, both pattern learning and active feature selection are novel approaches in the WSD field, and they led to very good results during the SENSEVAL-2 evaluation exercise. 2 System description The WSD algorithm used in this system has the capability of tagging words when no specific sense tagged corpora is available, automatically scaling up to larger training data2 when provided. Due to space constraints, we will not be able to give a detailed description of the system. How- ever we try to gain space and replace one thou- sand words with a picture: Figure 1 shows an overview of the system architecture. It illustrates the two main components, namely pattern learn- ing from available sense tagged corpora and dic- tionary definitions and instance based learning with active feature selection. The two modules are preceded by a preprocessing phase which in- cludes compound concept identification, and fol- lowed by a default phase that assigns the most frequent sense as a last resort, when no other previous methods could be applied. The shaded areas in Figure 1 are specific for the case when larger training data sets are available. During …

Content truncated to 500 words