|
|
|
Lexical Knowledge Networks are resources which are critical for most NLP applications. This is evidenced by the success of the endeavour on wordnets which are now regarded as indispensable for work on semantic search, knowledge based machine translation, summarization, question answering and such other high end NLP applications.
In this presentation, we delve into the world of lexcial networks which include wordnets in different languages, conceptnet, mindnet, hownet, verbnet etc. We elucidate the principles behind these networks and also describe the their design and implementation. This helps us draw critical comparisons between these networks. This discussion is of great importance to any NLP researcher. We then move onto theoretical studies on wordnets- focusing on their combinatorial properties which are inspired by studies on social networks and small world properties. We observe similarities on these dimensions in wordnets of English, Hindi and Marathi. The final part will deal with interesting approaches to automatic validation of the wordnet synsets using web evidences. The web has the potential to provide validatory evidences for myriad language phenomena. This general stream of research on using the web as the validator of language processing theories will be discussed, which should be highly beneficial to students. We have recently obtained interesting results on automatic synset validation, whcih will be presented. In summary, this lecture should equip the student with knowledge of lexical networks- their properties, comparisons, design and implementations and their automatic validation. |