Video Retrieval Using Knowledge Graph and Word2vec Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for retrieving similar videos are heavily dependent on timely corpus updates, leading to low recall and resource wastage due to frequent model training and computing resource consumption.
Innovation Solution
A method and apparatus that utilize a preset knowledge graph and word2vec model to convert video information into word vectors, combining knowledge graph and text similarity to retrieve similar videos, reducing the frequency of model training and conserving computing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the word2vec model is retrained frequently to maintain retrieval accuracy with updated corpora, then the retrieval accuracy is improved, but the computing resource consumption increases
Solution Approach 1:
The patent pre-trains the word2vec model offline with comprehensive video title corpora before deployment. This preliminary action allows the model to be ready for accurate retrieval without requiring frequent retraining, thus maintaining high retrieval accuracy while minimizing online computing resource consumption.
Solution Approach 2:
The patent uses a pre-trained word2vec model that can be copied and deployed multiple times across different systems. This copying approach eliminates the need for each system to independently retrain models frequently, reducing overall computing resource consumption while maintaining consistent retrieval accuracy across deployments.
2Reliability
If the word2vec model is retrained frequently to adapt to updated corpora, then the retrieval result reliability is improved, but the time consumption increases
Solution Approach 1:
The patent performs comprehensive model training in advance during an offline phase, preparing the word2vec model with up-to-date corpora before the actual retrieval operation. This preliminary action ensures high retrieval result reliability when needed, while avoiding time-consuming retraining during urgent retrieval operations.
Solution Approach 2:
The patent implements periodic offline retraining of the word2vec model at scheduled intervals rather than continuously or on-demand. This periodic action maintains retrieval result reliability by keeping the model updated with new corpora, while minimizing time loss by avoiding frequent or continuous training interruptions.
3Ease of manufacture
If conventional text similarity methods are used for video retrieval, then the implementation simplicity is maintained, but the recall rate decreases
Solution Approach 1:
The patent replaces conventional text similarity calculation methods with a word2vec-based vector space model. This substitution transforms the mechanical string-matching approach into a semantic vector comparison approach, significantly improving the recall rate by capturing semantic relationships while maintaining computational efficiency through optimized vector operations.
Data Source
AI summary
Embodiments of this application disclose a method for retrieving similar videos performed at a computing device. The computing device obtains video information of a video for which similar videos are to be retrieved, the video information including a video tag and a video title, and trains the video information by using a preset text depth representation model, to convert the video information into a word vector. After selecting, from a video library according to a preset knowledge graph, videos matching the video information, to obtain a first candidate video set, the computing device screens, in the video library, videos similar to the video information according to the word vector, to obtain a second candidate video set and then determines a similar video for the video information from the first candidate video set and the second candidate video set.