Semi Supervised Agriculture Information Extraction and Knowledge Graph Creation Model using Weighted Distributional Semantics Syntactic Dependencies

Abstract

Information Extraction (IE) is a field of Natural Language Processing (NLP) that involves newlineautomatically extracting useful information from text data. Developing accurate information newlineextraction models requires overcoming several challenges, domain specific vocabulary, data newlineintegration challenges, dynamic data, and the need for domain expertise. It involves newlineidentifying and extracting specific pieces of information, such as entities, relationships, newlineevents, and concepts, from unstructured or semistructured data, such as text documents, newlineweb pages, or social media posts. It has applications in industries such as finance, healthcare, newlineagriculture, and social media analysis. IE systems and Knowledge Graphs (KGs) are newlineinterconnected because the former is used to extract information from unstructured data, newlinewhile the latter stores this information in an organized and easily accessible manner. newlineThe knowledge graph facilitates easy querying and analysis of information. The knowledge newlinegraph can be accessed by users in an organization to gain insights, make decisions, and newlineautomate processes. For the past few decades, there has been significant research activity in newlinethe field of automatic knowledge graph creation. Among the various tasks involved, triplet newlineextraction, which involves identifying entities and their relationships, has proven to be newlineparticularly challenging. Supervised approaches demand an extensive corpus of annotated newlinetraining data, comprising entities and relationships. This training data is employed to newlinetrain a classifier, which in turn, is utilized to extract relationships from the test data. newlineThough supervised models outperform unsupervised models, they are constrained by the newlineneed of labelled data for the triplet extraction task. In our study, the main focus is on newlineautomatically extracting triplets from agricultural text documents and constructing an newlineagricultural knowledge graph.A major contribution of this thesis is an unsupervised weighted distributional semantics approach for entity labeling in the agricultural...