17-09-2014, 03:43 PM
Natural language processing (NLP) is one of the most explored and highly technical fields of artificial intelligence. Its notable applications include question answering (QA), machine translation (MT), text summarization, information extraction (IE) and many more. Moreover research efforts have enabled NLP to have multi-lingual span. Representation learning is one type of attempt whereas deep learning is another approach. Deep learning is contextual. In deep learning various factors are organized into multiple levels corresponding to different levels of abstraction or composition. Current NLP systems are incredibly fragile because of their atomic symbol representation. Feature learning needs multiple levels of representations. Multiple level of latent variable allows combinatorial sharing of statistical strength but insufficient model depth can exponentially be inefficient. Successive model layers learn deeper intermediate representation. Sentence is made up phrases of words and grammatical structure. Available NLP packages can perform tasks like name entity recognition (NER), part-of-speech (POS) tagging, phrase structure, dependency graphs, etc. But for contextuality, other approaches are required. Linguistic modality is one such region where the perception is made depending on the context. In this study attempt is made to summarize logical approaches that explored in context to describe modality and the performance of selected software with respect to specific modality term discrimination was compared. This study puts light on human cognitive functions with context to deep learning and performance of artificial intelligence.