Data Extraction from Natural Language using Universal Networking Language

摘要

Data extraction, which falls under the area of Natural Language Processing (UNL), finds specific data from unstructured data. This research paves the way to introduce a unique technique on data extraction - providing the user with exactly what is asked without any mimicry of unsolicited data. The proposal sets logical and symmetrical relation between the search criteria and operational data. Since the data is unstructured and volume can be relatively high, we have emphasized highly on putting the data under categories - defined and used by the researchers for further exploitation of data. Universal Networking Language (UNL) is efficiently used to compare data and merge. A new approach of machine learning is presented herein that essentially augments efficiency of Natural Language Computing (NLC) and Cognitive Computing (CC). This proposed approach uses UNL relationship and successful test data shows much improved results and efficient generalization. Existing machine learning approaches are widely used on numeric data which are producing expected results but one key contention is the limitation of data type that can be handled. Current models fail to properly train on the semantics, logical consistency; many natural language properties are either ignored or prove too much of a task. Consequently, the approach presented herein this paper carries further positive points in producing meaningful and worthwhile result. Moreover, complex data that are consisted of alphanumeric data, sequence and resulting criteria can be executed correctly.