ISA-bEL: Intelligent Search Algorithm based on Entity Linking
This addresses a specific issue in natural language processing for industries using text- or voice-based interfaces, but appears incremental as it builds on existing entity linking and knowledge graph methods.
The paper tackles the problem of information loss in user input processing by proposing a natural language processing pipeline that emphasizes entities, linking them to a knowledge graph to improve search accuracy, though no concrete numbers are provided.
Nowadays, the way in which the people interact with computers has changed. Text- or voice-based interfaces are being widely applied in different industries. Among the most used ways of processing the user input are those based on intents or retrieval algorithms. In these solutions, important information of the user could be lost in the process. For the proposed natural language processing pipeline the entities are going to take a principal role, under the assumption that entities are where the purpose of the user resides. Entities fed with context will be projected to a specific domain supported by a knowledge graph, resulting in what has been named as linked entities. These linked entities serve then as a key for searching a top level aggregation concept within our knowledge graph.