Demo of SemIndex: Semantic-Aware Inverted Index on Text

Processing keyword-based queries is a central problem in Information Retrieval (IR), where several studies have been done to develop effective keyword-based search techniques [1, 2]. A standard containment keyword-based query, which retrieves textual identities that contain a set of keywords, is generally supported by a full-text index. The inverted index is considered as one of the most useful full-text indexing techniques for large textual collections [3], supported by many RDBMSs. It is also increasingly used on semi-structured [2] and unstructured data [1] to support keyword-based queries. Besides the standard containment keyword-based query, the so-called semantic-aware or knowledge-aware (keyword) query has emerged as a natural extension, encouraged by (nonexpert) user demand. In semantic-aware queries, some knowledge needs to be taken into consideration while performing query processing. To illustrate this, let’s assume having data from a movie database , as shown in Table 1. Each movie in , identified with an id, is described with some text, including movie title , year and plot. For queries “sound of music”, “Maria nun” and “sound Maria”, the query result is movie O3. However, if the user wants to search for a movie but cannot recall the exact movie title, or the exact plot description, it is natural to assume that she may modify the query terms to some semantically similar terms, for example, “voice of music”.