Retrospective Layer for Coherent Key Phrase Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for automatic key phrase extraction often produce non-coherent key phrases due to the inclusion of superfluous words, which can affect the accuracy of corpus categorization and search engine optimization.
Innovation Solution
The implementation of a retrospective layer in the machine learning model that modifies predicted probabilities based on the sequence and position of words within the corpus, ensuring that only coherent key phrases are extracted by considering the dependence relationships between words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a machine learning model extracts keywords based on individual word probabilities, then the extraction process is simple and fast, but the resulting key phrases are non-coherent and include superfluous words
Solution Approach 1:
The model is segmented into distinct functional layers: initial keyword probability determination layers and a retrospective layer. Each layer performs a specific function - the initial layers identify potential keywords based on word importance, while the retrospective layer specifically handles coherence verification by analyzing word sequence relationships. This segmentation allows the complex task of coherent key phrase extraction to be divided into manageable, specialized sub-tasks.
Solution Approach 2:
The retrospective layer performs preliminary action by modifying keyword probabilities based on sequence relationships before the final key phrase assembly occurs. By adjusting probabilities in advance based on whether words appear in logical sequences, the model ensures coherence is established before commitment to final key phrase selection, preventing superfluous words from being included.
2Measurement precision
If the model considers word sequence relationships to improve coherence, then key phrase quality improves, but the processing time and computational complexity increase
Solution Approach 1:
The coherence verification function is extracted as a separate retrospective layer distinct from the initial keyword identification process. This extracted layer specifically handles sequence relationship analysis, allowing the main keyword extraction to remain efficient while adding coherence checking as a dedicated post-processing step that modifies only the necessary probability adjustments.
Solution Approach 2:
The model changes parameters dynamically by adjusting keyword probabilities based on sequence relationships. The retrospective layer modifies the initial probability values by applying adjustments that reflect word position and sequence coherence, transforming the raw probability parameters into refined coherence-aware probabilities without fundamentally changing the extraction algorithm.
3Productivity
If superfluous words are included in key phrases, then more words are captured potentially relevant to the corpus, but the categorization accuracy and search engine optimization effectiveness decrease
Solution Approach 1:
The retrospective layer implements feedback by continuously monitoring and adjusting keyword probabilities based on sequence relationship analysis. When words are identified as being in non-coherent sequences, the feedback mechanism reduces their probabilities, preventing them from being included in final key phrases. This feedback loop ensures that only words contributing to coherent phrases are selected, improving categorization accuracy.
Data Source
AI summary
A method including receiving, in a machine learning model (MLM), a corpus including words. The MLM includes layers configured to extract keywords from the corpus, plus a retrospective layer. A first keyword and a second keyword from the corpus are identified in the layers. The first and second keywords are assigned first and second probabilities. Each probability is a likelihood that a keyword is to be included in a key phrase. A determination is made, in the retrospective layer, of a first probability modifier that modifies the first probability based on a first dependence relationship between the second keyword being placed after the first keyword. The first probability is modified using the first probability modifier. The first modified probability is used to determine whether the first keyword and the second keyword together form the key phrase. The key phrase is stored in a non-transitory computer readable storage medium.