An Entity Retrieval system is a fundamental task of Information Retrieval that provides direct answer to an information need of user. Prior work of entity retrieval utilizes either the Knowledge Graph fields or the text relevant to the query via pseudo-relevance feedback to improve the performance. Recently, Knowledge Graph embeddings or other entity representations, which capture the entity information from a knowledge graph are shown to be beneficial for entity retrieval. However, such embeddings are query-agnostic. In this dissertation work, we aim to improve entity retrieval by exploring the pseudo-relevance feedback to generate entity representations that capture query-aware entity information to determine the relevance of entities. We study the effectiveness of pseudo-relevance feedback against Knowledge Graph fields and investigate the efficacy of the Knowledge Graph embeddings for entity retrieval. We aim to understand the importance of utilization of query-aware signals and modeling of such signals with Knowledge Graph embeddings. Our results show that pseudo-relevance feedback is more effective than the Knowledge Graph fields by 30%.

We provide a resource for automatically harvesting relevance benchmarks from Wikipedia -- which we refer to as "Wikimarks" to differentiate them from manually created benchmarks. Unlike simulated benchmarks, they are based on manual annotations of Wikipedia authors. Studies on the TREC Complex Answer Retrieval track demonstrated that leaderboards under Wikimarks and manually annotated benchmarks are very similar. Because of their availability, Wikimarks can fill an important need for Information Retrieval research.