Unlock instant, AI-driven research and patent intelligence for your innovation.

Course knowledge relationship extraction method and system based on sentence bag attention remote supervision

A remote supervision and relationship extraction technology, applied in the field of relationship extraction, can solve problems affecting test results and difficult evaluation of model performance

Pending Publication Date: 2020-11-10
HUBEI UNIV OF TECH
View PDF6 Cites 21 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Another type of mislabeling is a false negative example caused by an incomplete knowledge base, that is, a certain entity pair does indicate a certain relationship in a sentence, but because the information does not exist in the knowledge base, the sentence is labeled by the machine. no relationship
This problem occurs in both the training set and the test set, and will greatly affect the test results
Both types of mislabeling problems are quite challenging, and it is difficult to evaluate the performance of the model

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 more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Course knowledge relationship extraction method and system based on sentence bag attention remote supervision
  • Course knowledge relationship extraction method and system based on sentence bag attention remote supervision
  • Course knowledge relationship extraction method and system based on sentence bag attention remote supervision

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0209] Example 1: Course Knowledge Entity Extraction Experiment

[0210] The present invention at first extracts knowledge entity by three kinds of keyword extraction algorithms of TF-IDF, TextRank, Word2vec word clustering, comprises following steps:

[0211] Step 1, preprocessing the text;

[0212] Step 2, extract knowledge entities through three keyword extraction algorithms: TF-IDF, TextRank, and Word2vec word clustering;

[0213] In the present invention, the text is preprocessed in the step 1, including: transcoding the text format, removing redundant symbols and stop words, and word segmentation operations, and the step 1 further includes:

[0214] Step 1.1, merge the files in the folder and write them into one file, which is convenient for subsequent operations, and modify the encoding of the files to UTF-8, which is more conducive to subsequent program processing after modification;

[0215] Step 1.2, remove the redundant symbols in the text, because these texts nee...

Embodiment 2

[0242] Example 2: Noise reduction experiment of remote supervised extraction

[0243] The second embodiment of the present invention is based on artificially marked knowledge triplets, using the remote supervision method to automatically obtain the training corpus from the course teaching text; then use PCNN to extract sentence features, and use the sentence bag attention mechanism to mark the remote supervision method The large amount of noise existing in the data is denoised. Then step 2 in embodiment 1, concrete steps are as follows:

[0244] Step 3: Manually labeled knowledge triplets;

[0245] Step 4: Automatically obtain the training corpus from the course teaching text using the distance supervision method and the triplet of step 3;

[0246] Step 5: Use PCNN to extract sentence features, and use the sentence bag attention mechanism to denoise a large amount of noise existing in the data marked by the remote supervision method;

[0247] In the present invention, the a...

Embodiment 3

[0300] Example 3: Experiment of course knowledge point relationship extraction

[0301] The third embodiment of the present invention is the extraction of course knowledge point relations. The contextual semantic information is captured through the word vector with attention, and the location information and type information of the entity are fused to construct entity features, which are input into the Bi_LSTM model to obtain knowledge point relation extraction.

[0302] Then step 5 of the embodiment, in the step 6, the Bi_LSTM model obtains knowledge point relationship extraction, and the steps include:

[0303] Step 6: Capture the contextual semantic information through the word vector with attention, and integrate the location information and type information of the entity to construct entity features, and input the Bi_LSTM model to obtain knowledge point relationship extraction.

[0304] The three features of entity-aware attention include: (1) Bi_LSTM hidden layer state H...

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

No PUM Login to View More

Abstract

The invention belongs to the technical field of relationship extraction, and discloses a course knowledge relationship extraction method and system based on sentence bag attention remote supervision,and the method comprises the steps: extracting a knowledge entity through three types of keyword extraction algorithms: TFIDF, TextRank and Word2vec word clustering; automatically obtaining a trainingcorpus from the course teaching text by using a remote supervision method based on the knowledge triple of manual annotation; extracting sentence features by using a PCNN, and denoising a large amount of noise existing in the data labeled by the remote supervision method by using a sentence bag attention mechanism; capturing context semantic information through a word vector with attention, fusing the position information and the type information of the entity to construct entity features, and inputting the entity features into a BiLSTM model to obtain knowledge point relation extraction. Heavy manual labeling work is not needed, the work of manual feature construction is reduced, the method can be applied to course teaching of different subjects, and a good result can be obtained for knowledge relationship extraction in courses.

Description

technical field [0001] The invention belongs to the technical field of relation extraction, and in particular relates to a method and system for extracting course knowledge relations based on bag-of-sentence attention remote supervision. Background technique [0002] At present, with the wide application of artificial intelligence technology in various fields of society, using information extraction technology to extract key information from course teaching materials to construct a knowledge map of course learning is a research hotspot in the current course information construction. Among them, relational extraction is an important part of information extraction technology. It refers to the automatic extraction of semantic relations between entity pairs and effective semantic knowledge by modeling text information, which is a very critical part in the construction of knowledge graphs. [0003] In recent years, the development of deep learning has provided strong support for ...

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
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F40/295G06F40/211G06F16/35G06N3/04G06N3/08
CPCG06F40/295G06F40/211G06F16/35G06N3/049G06N3/08G06N3/047G06N3/044G06N3/045Y02D10/00
Inventor 陈建峡张水晶陈煜张杰
Owner HUBEI UNIV OF TECH