Unknown-Word Vector Assignment with Static and Dynamic Dictionaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data search technologies face challenges in accurately assigning vectors to unknown words, leading to low vector accuracy in data retrieval.
Innovation Solution
The proposed solution involves using a static dictionary and a dynamic dictionary to assign vectors to unknown words. The static dictionary defines relationships between static words, codes, and vectors, while the dynamic dictionary dynamically assigns codes and vectors to unknown words based on their similarity to defined unique words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a static dictionary is used to assign vectors to words, then the vector assignment is simple and fast, but the accuracy is low for unknown words
Solution Approach 1:
The patent divides the dictionary system into two segments: a static dictionary for common words and a dynamic dictionary for unknown words. This segmentation allows the system to use the simple static dictionary for most words while handling unknown words through the dynamic dictionary with Poincaré embedding, thus improving overall accuracy without applying complexity uniformly across all words.
Solution Approach 2:
The patent introduces a dynamic dictionary as an intermediary between the static dictionary and the vector assignment process. When a word is not found in the static dictionary, the dynamic dictionary uses Poincaré embedding to generate appropriate vectors, acting as a mediator that bridges the gap between known and unknown words.
2Measurement precision
If Poincaré embedding is used for unknown words, then vector accuracy improves, but processing time and computational complexity increase
Solution Approach 1:
The patent segments the word processing workflow into two paths: common words are quickly assigned vectors from the static dictionary, while only unknown words trigger the more computationally intensive Poincaré embedding process through the dynamic dictionary. This segmentation minimizes the time impact of Poincaré embedding by limiting it to only necessary cases.
Solution Approach 2:
The patent applies Poincaré embedding partially - only when necessary for unknown words - rather than applying it to all words. This partial action approach maintains high accuracy for unknown words while avoiding the excessive computational overhead of applying the same complex process uniformly to all vocabulary.
3Reliability
If a dynamic dictionary is implemented to handle unknown words, then retrieval effectiveness improves, but system complexity increases
Solution Approach 1:
The patent segments the dictionary system into static and dynamic components, where the static dictionary handles the majority of common words with simple lookups, and the dynamic dictionary handles only unknown words through Poincaré embedding. This segmentation improves retrieval effectiveness for unknown words while keeping the overall system complexity manageable by maintaining a simple static foundation.
Data Source
AI summary
A medium storing comprising instructions for: dividing a text into words; specifying an unknown word indicating a word other than a static word in a static dictionary among words in the text based on the static dictionary associating the static word with a vector; determining whether a unique word corresponding to the unknown word is included in a definition, using the definition defining a relationship between the unique word and a vector generated based on a condition that similar vectors are set for similar unique words; and when the unique word corresponding to the unknown word is included in the definition, assigning a vector corresponding to the unique word to the unknown word; and when the unique word corresponding to the unknown word is not included in the definition, assigning a vector to the unknown word using words arranged before and after the unknown word.