Browse Prior Art Database

System and Method for Filtering Sensitive Words Using Speech Recognition and Metadata Injections

IP.com Disclosure Number: IPCOM000247137D
Publication Date: 2016-Aug-10
Document File: 8 page(s) / 1M

Publishing Venue

The IP.com Prior Art Database

Abstract

This article describes an idea of distinguish sensitive words from original voice by using speech recognition technology, through sensitive words matching to record the time when these words appear. Then automatically block it , without human shield. Grade weight of each sensitive words using Metadata injections and set a threshold to determine whether it is necessary to process the original voice.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 8

System and Method for Filtering Sensitive Words Using Speech Recognition and Metadata Injections
Background
1. Network language is relatively new and spread quickly. But some network language contains sensitive words are not suitable for citizens who are under eighteen years of age. Some sensitive words are harmful to teenagers' mental growth.

2. Some sensitive words in network language generally include politics, porno, fraud, etc. Distinguish these words in voice by manually is time-consuming and expensive. And different people has his own rule to distinguish sensitive words, level of sensitivity is subjective, so usually team need a discussion to decide it.

Invention key idea


1. Using speech recognition technology to distinguish sensitive words from original voice.

2. Through sensitive words matching to record the time when these words appear. Then automatically block it , without human shield.

3. Grade each sensitive words using Metadata injections.

4. Metadata file can be customized or can be downloaded from Cloud server. The Metadata data on Cloud server is based on data analytics.
5. Set a threshold to determine whether it is necessary to process the original voice.

Advantage &Value


1. This disclosure provide a way to make filtering sensitive words in voice more simple, more flexible, more cheap.

2. This disclosure provide a way to filter the sensitive words during the conversations flexibly. For example, As parents, can grade sensitive words using metadata by themselves, to create a suitable network environment for their children.

3. This disclosure provide a way to reduce the communication time about sensitive degree for a word. For example, if a user found a sensitive word, he will not argue with other users to discuss the sensitive degree for a word. What he needs to do is to input the weight of the word in metadata from his angle or refer to the value which Cloud Server provided.

Description of solution


1. This is architecture of the application described in this disclosure.

1



Page 02 of 8


2. Module 1 contains 2 components C11 and C12. Through C11 Speech Recognition System, to distinguish words from original voice.The function of C12 is to record the time of these words appear.

2



Page 03 of 8


3. Module 2 contains 2 components C21 and C22. The function of C21 is provided to add sensitive words to C22 component. The func...