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

DSMM (Deep semantic match model) entity linking method based on multi-granularity LSTM (long short term memory) network

A semantic matching, multi-granularity technology, applied in the field of information processing

Inactive Publication Date: 2018-07-10
BEIJING UNIV OF POSTS & TELECOMM
View PDF5 Cites 23 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to solve the existing technical problems, the present invention provides a deep semantic matching entity linking method based on a multi-granularity LSTM network

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
  • DSMM (Deep semantic match model) entity linking method based on multi-granularity LSTM (long short term memory) network
  • DSMM (Deep semantic match model) entity linking method based on multi-granularity LSTM (long short term memory) network
  • DSMM (Deep semantic match model) entity linking method based on multi-granularity LSTM (long short term memory) network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0011] Next, the implementation method of the present invention will be described in more detail.

[0012] figure 1 It is a network structure diagram of a deep semantic matching entity linking system based on a multi-granularity LSTM network provided by the present invention, including:

[0013] Step S1: Surface form matching

[0014] Step S2: Context Semantic Matching

[0015] Step S3: Similarity measure

[0016] figure 2 The structure diagram of char / word-LSTM is given.

[0017] Each step will be described in detail below:

[0018] Step S1: surface form matching. Since the common lengths of entity references and candidate entities are very short, the present invention uses a character-level bidirectional LSTM network (char-LSTM) to extract the surface form feature representations of the two. char-LSTM is more robust, able to accept character errors due to some printing, tense or other spelling reasons, and can contain a certain degree of semantic information of the w...

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 discloses a DSMM (deep semantic match model) entity linking method based on a multi-granularity LSTM (long short term memory) network, and belongs to the field of information processing.The method is characterized by comprising steps as follows: firstly, extracting surface form feature representations of an entity reference and a candidate entity by using a character-level bidirectional LSTM network; then, encoding the sentence in which the entity reference is located by adopting a word-level bidirectional LSTM network, outputting the sentence as a context semantic feature vector of the entity reference, and learning the context semantic feature vector of the candidate entity based on information of a structured knowledge map; finally, calculating similarity scores of the surface form and the semantics for the entity reference, the surface form of the candidate entity and the context semantic feature vector respectively, and combining the similarity scores as a final score of the entity reference and candidate entity pair. The entity link effect is improved by combining the multi-granularity LSTM network and the knowledge representation learning method.

Description

technical field [0001] The invention relates to the field of information processing, in particular to a deep semantic matching (Deep Semantic Match Model, DSMM) entity linking method based on a multi-granularity LSTM network. Background technique [0002] Entity linking is the basic link in various application fields of natural language processing. Its goal is to link entity references in free text to entities corresponding to the target knowledge graph, so as to solve the problem of ambiguity between entities. The core of entity linking research is how to sort the candidate entity set to pick out the correct mapped entity. The quality of entity linking will directly affect the upper-level tasks, such as information retrieval and automatic question answering. [0003] Most of the traditional entity linking algorithms generally use unstructured knowledge graphs, and manually extract the feature vectors of the entity reference and the context text of the candidate entity. Ho...

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
IPC IPC(8): G06F17/30
CPCG06F16/36
Inventor 高升罗安根王新怡徐雅静李思
Owner BEIJING UNIV OF POSTS & TELECOMM