Question answer extraction method based on multi-layer perception and electronic device
A multi-layer perception and answer extraction technology, applied in the fields of electronic digital data processing, instruments, unstructured text data retrieval, etc., can solve the problems of complex reasoning process, separate processing of reasoning types, etc.
Active Publication Date: 2021-02-19
INST OF INFORMATION ENG CHINESE ACAD OF SCI
View PDF6 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0007] However, most of the current models do not deal with different types of inference types separately, and the models are mostly complicated in order to model the general inference process
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0038] In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail in conjunction with the accompanying drawings.
[0039] The present invention classifies inference categories in the HotpotQA dataset. There are two main classes, bridging entity classes and comparison classes. If the question is a bridging entity class reasoning type, the model will process the answer prediction process into two hierarchically stacked subtasks. The first layer is to find intermediate bridging entities by looking for modules, and the second layer is to output the first layer through the conversion module. The bridging entity representation and the question and context to lock the final answer; if the question is a comparative reasoning type, the model will process the answer prediction process into a two-layer stacked subtask, the first layer is to find it through two search modules The tw...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention provides a question answer extraction method based on multi-layer perception, which comprises the following steps: splicing a question with a plurality of target documents, inputting thespliced question into a pre-training language model to obtain a representation Q of the question and a context representation P of the target documents, and interacting the representation Q with thecontext representation P to obtain a question answer; obtaining a problem representation u related to the document and a document representation h fusing the problem information; performing multi-layer perception classification on the question representation u to obtain an inference type of the question, and obtaining answer attention distribution of the question in the target document according to the inference type, the question representation u, the document representation h and a sub-question ct generated through representation Q, where t is the number of times of generating the sub-question; and obtaining an answer prediction result of the question according to the answer attention distribution. According to the invention, questions are answered progressively in a sub-question splitting mode, an inference category classifier is introduced to control splitting, answers to the questions are shared, and the inference reading understanding effect is improved.
Description
technical field [0001] The invention belongs to the field of natural language processing, and in particular relates to a method for extracting answers to questions based on multi-layer perception and an electronic device. Background technique [0002] Inferential reading comprehension is to find multiple related documents for a user's question, and find the answer to the question and related evidence sentences from the documents. Reasoning reading comprehension questions requires the model to combine questions, reason about the semantic meaning of the text, and find relevant evidence sentences and final answers to the questions. Inferential reading models can be generally divided into three categories of methods. One is the memory network method, which simulates the reasoning process by continuously iteratively updating the reasoning state; the other is based on the graph neural network method, which performs reasoning through the update of the graph neural network; there a...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06F16/332G06F16/36G06N3/04
CPCG06F16/3329G06F16/36G06N3/049G06N3/044
Inventor 林政付鹏刘欢王伟平孟丹
Owner INST OF INFORMATION ENG CHINESE ACAD OF SCI