Knowledge graph is widely used to answer questions. Generally, there are three components in a QA system: question interpretor, which is responsible for parsing the questions in natural language, and understanding the intents of questioners; knowledge graph based reasoning engine, which is to query and reason answers to these questions, and; answer generator, which can organize the answers and translate them into natural language.
The common types of questions in Knowledge Graph based QA contains:
- Entity recognition
- Disambiguation
- Relation classification
- Segmentation
- DBpedia
- WikiData
- [Google's Knowledege Graph]
- [Microsoft Bing's Satori Knowledge Base]
- [Yandex' Object Answer]
(From Neural Network-based Question Answering over Knowledge Graphs on Word and Character Level)
- Lexical Gap – surface forms for relations expressed in a question can be quite different from those used in the KG
- Ambiguity – the same word is used for different entities, such as president of a company or a country,
- Unknown knowledge boundaries – QA systems can often hardly decide whether a certain question is answerable at all give a certain knowledge base
| Technique | Question | Algorithm | Novelty |
|---|---|---|---|
| [1] | Entity prediction | Gated Recurrent Unit + RNN | Character- and word-level representations for questions |
| [2] KnowBot |
- Neural Network-based Question Answering over Knowledge Graphs on Word and Character Level, IW3C2 2017
- Learning Knowledge Graphs for Question Answering through Conversational Dialog