Sentiment parsing method, electronic device, and storage medium

Patent metadata

Field	Value
Publication number	US-12307199-B2
Application number	US-202217942116-A
Country	US
Kind code	B2
Filing date	Sep 10, 2022
Priority date	Feb 25, 2022
Publication date	May 20, 2025
Grant date	May 20, 2025

Abstract

There is provided a sentiment parsing method and apparatus, an electronic device, and a storage medium, which relates to the technical field of artificial intelligence such as machine learning and natural language processing. A specific implementation solution involves: identifying a role of a sentiment parsing object in a specified statement; trimming the specified statement based on the role of the sentiment parsing object to acquire pruning result information after the trimming; and parsing sentiment of the sentiment parsing object in the specified statement based on the pruning result information.

First claim

What is claimed is:

1. A computer-implemented sentiment parsing method, comprising:
selecting a sentiment parsing object from a specified statement; 1
identifying a role of the sentiment parsing object in the specified statement;
trimming the specified statement based on the role of the sentiment parsing object to remove noise irrelevant to the sentiment parsing object in the specified statement to acquire pruning result information after the trimming, wherein the pruning result information is simplified information relevant to the role of the sentiment parsing object in the specified statement which does not include any other irrelevant noise; and
inputting the pruning result information into a pre-trained sentiment parsing model for parsing sentiment of the sentiment parsing object in the specified statement wherein the pre-trained sentiment parsing model is a neural network model that classifies the pruning result information into a plurality of levels of emotions, 2
wherein the trimming the specified statement based on the role of the sentiment parsing object to acquire pruning result information after the trimming comprises:
extracting, if the role of the sentiment parsing object is a subject, the subject, a predicate corresponding to the subject, a modifier of the predicate, an object corresponding to the subject, and a modifier of the object from the specified statement as the pruning result information;
extracting, if the role of the sentiment parsing object is an object, the object, a modifier of the object, a subject corresponding to the object, a modifier of the subject, a predicate corresponding to the object, and a modifier of the predicate from the specified statement as the pruning result information; or
extracting, if the role of the sentiment parsing object is a predicate, the predicate, and a preposition and a complement corresponding to the predicate from the specified statement as the pruning result information,
wherein the role of the sentiment parsing object is a subject or an object, the subject, a predicate corresponding to the subject, a modifier of the predicate, an object corresponding to the subject, and a modifier of the object from the specified statement as the pruning result information, or the predicate, and a preposition and a complement corresponding to the predicate from the specified statement as the pruning result information,
after the extracting, the method further comprises:
detecting whether a conjugate role exists in the predicate corresponding to the sentiment parsing object in the specified statement;
detecting, if yes, whether the sentiment parsing object is the nearest subject or object corresponding to the predicate;
deleting, if no, the modifier of the predicate extracted from the pruning result information;
extracting a modifier corresponding to the conjugate role of the predicate; and
adding the modifier corresponding to the conjugate role of the predicate to the pruning result information, and
prior to the extracting, the method further comprises:
detecting whether a conjugate role exists in the predicate corresponding to the sentiment parsing object in the specified statement;
identifying, if yes, the predicate of the sentiment parsing object by reverse acknowledgment based on the sentiment parsing object; and
marking the predicate of the sentiment parsing object, and
the method further comprises:
detecting whether the pruning result information is clause information in the specified statement;
additionally extracting, if yes, main clause information corresponding to the clause information from the specified statement; and
adding the main clause information to the pruning result information.

2. The method of claim 1 , wherein, if the role of the sentiment parsing object is a subject or an object, the method further comprises:
performing coreference parsing on the sentiment parsing object, and detecting whether a coreference object having the same reference as the sentiment parsing object exists in the specified statement;
extracting, if yes, relevant information of the coreference object from the specified statement based on the coreference object; and
adding the relevant information of the coreference object to the pruning result information.

3. The method of claim 1 , wherein the identifying a role of a sentiment parsing object in a specified statement comprises:
identifying the role of the sentiment parsing object in the specified statement by using a dependency syntactic parsing tool; or
identifying the role of the sentiment parsing object in the specified statement by using a pre-trained role identification model.

4. The method of claim 1 , wherein the parsing sentiment of the sentiment parsing object in the specified statement based on the pruning result information comprises:
parsing the sentiment of the sentiment parsing object in the specified statement based on the pruning result information by using a pre-trained sentiment parsing model.

5. The method of claim 1 , wherein the identifying a role of a sentiment parsing object in a specified statement comprises:
identifying the role of the sentiment parsing object in the specified statement by using a dependency syntactic parsing tool; or
identifying the role of the sentiment parsing object in the specified statement by using a pre-trained role identification model.

6. The method of claim 1 , wherein the parsing sentiment of the sentiment parsing object in the specified statement based on the pruning result information comprises:
parsing the sentiment of the sentiment parsing object in the specified statement based on the pruning result information by using a pre-trained sentiment parsing model.

7. An electronic device, comprising:
at least one processor; and
a memory communicatively connected with the at least one processor;
wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a sentiment parsing method, wherein the sentiment parsing method comprises:
selecting a sentiment parsing object from a specified statement; 3
identifying a role of the sentiment parsing object in the specified statement;
trimming the specified statement based on the role of the sentiment parsing object to remove noise irrelevant to the sentiment parsing object in the specified statement to acquire pruning result information after the trimming, wherein the pruning result information is simplified information relevant to the role of the sentiment parsing object in the specified statement which does not include any other irrelevant noise; and
inputting the pruning result information into a pre-trained sentiment parsing model for parsing sentiment of the sentiment parsing object in the specified statement wherein the pre-trained sentiment parsing model is a neural network model that classifies the pruning result information into a plurality of levels of emotions, 4
wherein the trimming the specified statement based on the role of the sentiment parsing object to acquire pruning result information after the trimming comprises:
extracting, if the role of the sentiment parsing object is a subject, the subject, a predicate corresponding to the subject, a modifier of the predicate, an object corresponding to the subject, and a modifier of the object from the specified statement as the pruning result information;
extracting, if the role of the sentiment parsing object is an object, the object, a modifier of the object, a subject corresponding to the object, a modifier of the subject, a predicate corresponding to the object, and a modifier of the predicate from the specified statement as the pruning res

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Peng Minlong
• Sun Mingming
• Li Ping

Classifications

• G06F40/268

  Morphological analysis · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06F40/289

  Phrasal analysis, e.g. finite state techniques or chunking · CPC title

• G06F40/279

  Recognition of textual entities · CPC title

• G06F40/253

  Grammatical analysis; Style critique · CPC title