Dual-Channel Sentiment Preference Recognition for Swapped Object Order
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sentiment preference recognition models are sensitive to the order of comparison objects in sentences, leading to decreased discrimination accuracy when the positions of the objects are exchanged, and they lack intelligence and accuracy for practical applications in comparative opinion recognition.
Innovation Solution
A dual-channel sentiment preference recognition model is developed, utilizing a pre-trained language model with an original and mirror channel to generate text templates with swapped comparison objects, and a comparison module to distinguish logical relationships, enhancing robustness and accuracy by minimizing the impact of object order changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a traditional sentiment preference recognition model is used, then the model structure is simple, but the discrimination accuracy decreases when the positions of comparison objects are exchanged
Solution Approach 1:
The patent applies inversion by creating a mirror channel that processes the inverted version of the original text template. Specifically, when the original template is 'Product A handles better than product B', the mirror channel processes 'Product B handles worse than product A'. This allows the model to learn consistent sentiment representations regardless of object order, resolving the contradiction between discrimination accuracy and robustness to position changes.
Solution Approach 2:
The patent introduces a new dimension by adding the mirror channel as a parallel processing path to the original channel. This dimensional expansion allows the model to simultaneously process both original and inverted text templates, enabling it to capture sentiment preferences that are invariant to object ordering while maintaining high discrimination accuracy.
2Measurement precision
If the model processes both original and mirrored text templates, then the sentiment preference recognition accuracy improves, but the device complexity increases
Solution Approach 1:
The patent merges the original channel and mirror channel through shared parameter mechanisms. The comparison module integrates outputs from both channels and uses a shared parameter model to generate the final sentiment preference result. This merging approach allows the system to benefit from dual-channel processing while reducing overall complexity through parameter sharing and integrated output generation.
Solution Approach 2:
The comparison module serves multiple functions: it processes outputs from both original and mirror channels, performs sentiment comparison, and generates final predictions. This multi-functional design reduces the need for separate dedicated components for each processing stage, thereby improving accuracy while controlling device complexity.
3Adaptability or versatility
If more training data with varied object positions is used, then the model robustness improves, but the loss of time for data preparation increases
Solution Approach 1:
The patent applies preliminary action by automatically generating mirror text templates from the original training data during the data preparation phase. Instead of manually creating varied position examples, the system programmatically generates inverted versions of each template (e.g., converting 'A is better than B' to 'B is worse than A'). This preliminary generation of diverse training examples significantly reduces data preparation time while maintaining model robustness.
Data Source
AI summary
The present application discloses a method of training a sentiment preference recognition model for comment information, a recognition method and device thereof, which belong to the field of natural language processing technology. The method of training a sentiment preference recognition model for comment information includes training and optimizing, based on each piece of comment information, and corresponding original text template and mirrored text template thereof, a sentiment preference recognition model in the original channel and the mirror channel, so that the sentiment preference recognition model is used to output a sentiment preference result for the two comparison objects in the comment information in terms of the attribute.