Named Entities as a Metadata Resource for Indexing and Searching Information

Named Entity Recognition (NER) is an essential task in Natural Language Processing (NLP). By using NER, it is possible to create associations in a text to recognize real-world entities. The data indexing process is also considered a vital resource, as it makes it easier to find texts in a set of documents. When we analyze a search engine, we aim at the ease of the user's search process. Indexing recognized entities could help the search engine find data with a high semantic index, therefore, more accurate. This study aims to investigate the automatic transformation of annotated entities as indexes for a search engine. The recognition of entities used the hybrid model CRF+LG. Search engines usually work with keyword localization (tokens). However, this research aimed to use a semantic search, as it improves the quality of the results by understanding the user's intention using enricher meta factors besides the keyword. We performed ten experiments using P@{5, 10, and 20} and the search engine with a high semantic index achieved accuracy of 100%, correctly returning all results. The search engine without NER was confused when producing results for person and organization categories, mainly.