In this article, our main concern is the question and answer based on Chinese disease knowledge base, which is a limited domain question. For this type of question and answer, there are mainly three steps: (1) First, identify the entity and attribute of the question raised by the user; (2) Then, turn the entity and attribute into a structured query; (3) Finally use the structure Inquiries can be found in the disease knowledge base. According to the above three steps, the problems we need to solve are: (1) Missing Chinese question database about diseases asked by users; (2) Chinese disease knowledge base is missing; (3) There is a cascading error in the current mainstream pipeline method. For questions 1 and 2, we used crawler technology to crawl relative disease questions and structured disease knowledge bases on disease question and answer websites and disease websites, respectively. We named the disease question database and knowledge database as DieaseQuestion and DieaseBase respectively. For problem 3, we propose a joint model to identify the entities and attributes in the question. In this article, we prove the effectiveness of our model on the DieaseQuestion dataset.
[1]
Wen Wang,et al.
BERT for Joint Intent Classification and Slot Filling
,
2019,
ArXiv.
[2]
Hae-Chang Rim,et al.
Joint Relational Embeddings for Knowledge-based Question Answering
,
2014,
EMNLP.
[3]
Tsuyoshi Murata,et al.
{m
,
1934,
ACML.
[4]
Andrew Chou,et al.
Semantic Parsing on Freebase from Question-Answer Pairs
,
2013,
EMNLP.
[5]
Ken Yano,et al.
Neural Disease Named Entity Extraction with Character-based BiLSTM+CRF in Japanese Medical Text
,
2018,
ArXiv.
[6]
Zhi Jin,et al.
Classifying Relations via Long Short Term Memory Networks along Shortest Dependency Paths
,
2015,
EMNLP.
[7]
Xuchen Yao,et al.
Information Extraction over Structured Data: Question Answering with Freebase
,
2014,
ACL.