İsimli varlık tanıma araçlarının biyografik makalelere uygulanarak karşılaştırılması ve birleştirilmesi

In natural language processing domain, there are many named entity recognition tools with several different properties. This makes it difficult to select an appropriate NER tool for a specific situation. In this work, we try to answer this question in the context of biographic texts. For this matter, we first correct, clean and complete the corpus constituted by B. Kupelioglu [1]. We then select 4 publicly available, well known and free for research NER tools for comparison: Stanford NER, Illinois NET, OpenCalais NER WS and Alias-i LingPipe. We take advantage of the framework developed by Yasa Akbulut to compare Stanford, Illinois and OpenCalais, and complete it so that it can also handle LingPipe, too. We also add to this platform a new way of evaluating NER performance. We then compared the tools' performances. When considering overall performances, a clear hierarchy emerges: Stanford has best results, followed by LingPipe, Illionois and OpenCalais. However, a more detailed evaluation, considering entity types and article categories, highlights that performances are diversely influenced by those factors. This complementarity brings us to the definition of a combination method in order to improve the overall performance, using Support Vector Machine (SVM) trained on our corpus. We also manually define a set of rules to perform the same operation, in order to have a baseline when assessing the performance of our combination tool. We have found that these rules are better at performing full detection of entities, but that the SVM classifier is better at performing partial detection.