Unlock instant, AI-driven research and patent intelligence for your innovation.
An End-to-End Aspect-Level Sentiment Analysis Method Combining Reconstructed Syntactic Information
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A sentiment analysis, end-aspect-level technology, applied in the field of end-to-end aspect-level sentiment analysis, can solve the problems of not considering the influence of emotional polarity, losing long-distance dependencies, etc.
Active Publication Date: 2022-05-17
KUNMING UNIV OF SCI & TECH
View PDF8 Cites 0 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
However, the above model only considers the aspect words themselves, and does not take into account the influence of opinion words in comment sentences on the emotional polarity of aspect words. The judgment of emotional polarity is based on the current aspect word information, which will lose the long-distance useful for judging emotions. reliance on
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
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0033] Embodiment 1: as Figure 1-2 As shown, an end-to-end aspect-level sentiment analysis method combined with reconstructed syntactic information, the specific steps of the method are as follows:
[0034] Step 1. Encode the text through the Bert pre-training model to obtain a word vector representation with contextual representation;
[0035] As a further solution of the present invention, in the step 1, the Bert embedding layer uses the sentence as input, and utilizes the information of the whole sentence to calculate the word-level feature, and the word vector representation is through the formula H l =Transformer l (H l-1 ) calculation, where H l Table the feature representation of layer l.
[0036] Step 2, use the AllenNLP tool to design the Biaffine model, and use the biaffine model Biaffine to obtain the initial syntax tree;
[0057] Embodiment 2: as Figure 1-2 As shown, an end-to-end aspect-level sentiment analysis method combined with reconstructed syntactic information,
[0058] Step 1: Input text to bert layer encoding. Enter field-specific comment text w n , expressed as w=(w 1 ,...,w N ), where N is the length of the sentence. Then pack the input features into H 0 =(x 1 ,...,x T ), where x n (t ∈ [1, N]) is the input w n The corresponding combination of word embedding, position embedding and segment embedding;
[0059] Step 2: refine word-level features. Obtain the packaged text through step 1, refine the word-level features layer by layer through the l-layer transformer layer, and calculate the context representation corresponding to W where dim h Indicates the dimensionality of the representation vector.
[0060] The feature representation of layer L is computed as follows:
[0062] Step 3: Predict the label, set H l as contextual rep...
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 relates to an end-to-end aspect-level sentiment analysis method combined with reconstructed syntactic information, belonging to the technical field of natural languageprocessing. The present invention includes the steps of: encoding the text through the Bert pre-training model to obtain a word vector representation with contextual representation; using the biaffine model Biaffine to obtain the initial syntax tree; reshaping and pruning the initial syntax tree through syntax rules to obtain new dependencies tree; use the attention neural network to encode the new dependency tree to obtain reconstructed syntactic features; use the respectively obtained word vector representation and reconstructed syntactic features for feature splicing and fusion, and then input them to the downstream sequence labeling model; through the downstream sequence labeling model The output of is the result of aspect-level sentiment analysis. The present invention can obtain effective syntactic dependencies between aspect words and opinion words, and improve the performance of aspect-level sentiment analysis tasks.
Description
technical field [0001] The invention relates to an end-to-end aspect-level sentiment analysis method combined with reconstructed syntactic information, belonging to the technical field of natural languageprocessing. Background technique [0002] Traditionally, the task of aspect-based sentiment analysis (ABSA) can be divided into two subtasks, namely, the task of aspect word extraction and the task of aspect-level sentiment analysis. The purpose of aspect word extraction is to detect viewpoint targets mentioned in review texts, and to study them extensively. The purpose of aspect-level sentiment classification is to find the corresponding opinion expression words through aspect words, which can help the model predict the emotional polarity of a given aspect target. Most of the current work on ABSA aims to solve one of the subtasks. In order to apply these existing methods in practical settings, i.e., not only extracting evaluation objects but also predicting their sentime...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to View More 
Login to View More