Question answering system based on multisource heterogeneous data for medical field and implementing method thereof
A multi-source heterogeneous data, question answering system technology, applied in medical reference, digital data processing, healthcare informatics, etc., can solve the problems of less semantic information, single algorithm, single data source, etc., to improve the accuracy and reliability. Richness, alleviating the effect of single data source and effective utilization
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
specific Embodiment approach 1
[0038] Specific implementation mode 1. Combination figure 1 This embodiment is described. In this embodiment, a question answering system based on multi-source heterogeneous data for the medical field is described. The question answering system includes a client and a server. The client initiates a request for an answer to the server. The medical and health question-and-answer data and the open knowledge graph obtain the answers, process and synthesize the answers and return them to the client.
[0039] In this embodiment, the client includes a web terminal and a mobile terminal.
specific Embodiment approach 2
[0040] Specific embodiment two, combine figure 2 This embodiment is described. In this embodiment, an implementation method of a question answering system based on multi-source heterogeneous data for the medical field is described. The specific process of the implementation method of the question answering system is as follows:
[0041] Step 1, collect Chinese corpus;
[0042] Collect medical and health question-and-answer data;
[0043] Collect medical and health text data with structured features, and transform the medical and health text data into a structured database;
[0044] Step 2, perform word vector training on the corpus data in the Chinese corpus collected in step 1, and save the model as the pre-trained word vector of the deep learning model;
[0045]Step 3. Use the pre-trained word vector in step 2 to train the sequence-to-sequence network with an attention mechanism on the medical and health question-and-answer data collected in step 1, and the model generate...
specific Embodiment approach 3
[0062] Specific implementation mode three, this implementation mode is to further explain specific implementation mode two, the specific process of performing sequence-to-sequence network training with attention mechanism on data described in step 3 is:
[0063] Step 3-1, using the pre-trained word vector in step 2 to initialize all the words in the medical health question and answer data;
[0064] Step 3-2, using the memory network model to encode the user question, and obtain the encoded output value and hidden state;
[0065] Step 3-3, using the memory network model, combined with the attention mechanism, to decode the encoded output value, hidden state and input value of the user question, and obtain the predicted answer;
[0066] Step 3-4, calculate the loss function according to the decoded answer and the real answer;
[0067] Step 3-5, judge whether the loss parameter is converged, if not, execute step 3-6, if yes, execute step 3-7;
[0068] Step 3-6, perform backprop...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com