Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Natural language semantic reasoning method

A reasoning method and natural language technology, applied in reasoning methods, semantic analysis, electronic digital data processing, etc., can solve problems such as inability to give answers, uncontrollable answers, grammatical errors in answers, etc., to achieve the best reasoning results and customer experience , Strengthen the effect of practical value and commercial value

Active Publication Date: 2018-04-20
SHANDONG SYNTHESIS ELECTRONICS TECH
View PDF3 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The disadvantage of the evidence reasoning method is that it needs to exhaust all kinds of sentence patterns or questioning methods under natural language conditions as much as possible, and the second is that it is impossible to give answers to questions or situations that are not included in the knowledge base
The disadvantage of this method is that the answers given are not easy to control, unexpected answers are prone to appear, and the answers are prone to grammatical errors
Therefore, the application of such methods in commercial systems is still immature, and most of them are based on academic research.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] A natural language semantic reasoning method, comprising the following steps:

[0027] 1) Receive the dialogue sentence Input input by the user, and the sentence is word-segmented to calculate the user intention information Intent and the entity information set Entity.

[0028] The intent information Intent refers to the user's topic range, such as ordering food, telling stories, and querying the weather. The entity information set Entity can be a simple keyword set obtained through calculation, or it can be keyword information, part-of-speech information and weight information, that is, the entity information set Entity includes word segmentation set, word segmentation part-of-speech and word weight.

[0029] 2) According to the obtained user intent information Intent and entity information set Entity, through the index system, locate the candidate answer set CandiAnswer from the massive database. This indexing system first retrieves knowledge base entries that direct...

Embodiment 2

[0040] In this embodiment, the evidence reasoning best answer set Answer and the deep learning answer DeepAnswer obtain the final feedback information BestAnswer through a decision fusion algorithm in parallel mode. Parallel mode means that the two types of answers, the best answer set for evidence reasoning and the deep learning answer DeepAnswer, are not in sequence, and the optimal confidence answer is obtained through the index evaluation algorithm. A simplest implementation of the parallel mode: calculate the highest confidence based on the inference best answer set Answer and deep learning answer DeepAnswer, and use the corresponding answer as the final BestAnswer value. Others are the same as in Embodiment 1, and will not be repeated here.

Embodiment 3

[0042] In this embodiment, the evidence reasoning best answer set Answer and the deep learning answer DeepAnswer obtain the final feedback information BestAnswer through a mixed-mode decision fusion algorithm. The mixed mode refers to one or more of the two types of answers, the best answer set for evidence reasoning and the deep learning answer DeepAnswer, as the final BestAnswer value. An implementation of the mixed mode: set thresholds Tevid and Tdeep respectively, if the confidence of the best answer in Answer is greater than Tevid, then add the answer corresponding to this confidence to BestAnswer; if the confidence of DeepAnswer is greater than Tdeep, then add this confidence to BestAnswer The corresponding answer is added to BestAnswer. At this time, BestAnswer may have multiple answers. This result can be fed back directly to the customer. Others are the same as in Embodiment 1, and will not be repeated here.

[0043] The set thresholds in Embodiment 1, Embodiment 2...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a natural language semantic reasoning method. The semantic reasoning method is combined with evidential reasoning and deep learning. According to different application scenes,different combination strategies are utilized for combining two different kinds of information, thereby realizing semantic reasoning with higher accuracy and higher client experience. The natural language semantic reasoning method has advantages of effectively preventing defects in two kinds of methods and keeping respective advantages, realizing more accurate semantic reasoning, realizing more accurate and natural client feedback, and realizing better client experience.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, in particular to the fields of artificial intelligence natural language semantic reasoning and robot dialogue, specifically, a natural language semantic reasoning method. Background technique [0002] Robot semantic reasoning can understand customer questions and give reasonable feedback in a specific way, which is the core link of the dialogue system. Semantic reasoning first determines user intent and extracts entity information, and secondly infers the most reasonable answer based on intent information and entity information. The current main semantic reasoning methods mainly include two methods: evidential reasoning and deep learning. [0003] Evidence reasoning first locates candidate answers from massive databases through the index system, and then sorts through simple fuzzy matching or complex Learning to Rank and other algorithms to obtain the most suitable answer. In an e...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06N5/04G06F17/27
CPCG06F40/30G06N5/04
Inventor 张传锋
Owner SHANDONG SYNTHESIS ELECTRONICS TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com