Question Answering Neural Network Vocabulary Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing reading comprehension technologies generate inaccurate answers when encountering words not in the standard vocabulary, such as proper nouns or technical terms, as they treat these words as unknown entities.
Innovation Solution
A question answering apparatus that uses a learned neural network to generate answer sentences by incorporating both a general vocabulary and a vocabulary specific to the document and question, adjusting the probability of selecting words from the specific vocabulary to include relevant words in the answer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a standard vocabulary is used for generating answer sentences, then the system is simple and easy to operate, but words not included in the vocabulary (such as proper nouns or technical terms) are treated as unknown words, reducing answer accuracy
Solution Approach 1:
The patent combines the standard vocabulary (first vocabulary) with the specific vocabulary extracted from the document and question (second vocabulary) to form a unified vocabulary set. This merging allows the system to leverage both the general knowledge from the standard vocabulary and the context-specific terms from the document, thereby improving answer accuracy without requiring manual management of an excessively large vocabulary database.
Solution Approach 2:
The system performs preliminary extraction of specific vocabulary from the document and question before generating the answer sentence. By pre-identifying and incorporating terms that appear in the input text, the system prepares the necessary context-specific vocabulary in advance, ensuring that proper nouns and technical terms are available when generating the answer, thus improving reliability.
2Reliability
If a learned neural network adjusts word selection probability based on vocabulary inclusion, then answer accuracy improves, but the processing time and computational complexity increase
Solution Approach 1:
The learned neural network applies different selection probabilities to different words based on their local characteristics - specifically whether they belong to the standard vocabulary or the specific vocabulary extracted from the document and question. This local differentiation allows the system to focus computational resources on adjusting probabilities for context-specific terms while maintaining efficient processing for common words, thereby improving answer accuracy with minimized additional processing time.
Data Source
AI summary
This disclosure is provided, in which an answer generation unit configured to receive a document and a question as inputs, and execute processing of generating an answer sentence for the question by a learned model by using a word included in a union of a predetermined first vocabulary and a second vocabulary composed of words included in the document and the question, in which the learned model includes a learned neural network that has been learned in advance whether word included in the answer sentence is included in the second vocabulary, and increases or decreases a probability at which a word included in the second vocabulary is selected as the word included in the answer sentence at the time of generating the answer sentence by the learned neural network.