Social Graph Speech Recognition for Contact Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice recognition systems face inaccuracies and confusion when audio quality is poor or users have heavy accents, especially when trying to identify names or relationships from directories, as they rely solely on linear searches rather than contextual relationships.
Innovation Solution
The integration of social graph analysis within speech recognition systems to identify called parties by utilizing relationships from social networks, allowing for more accurate and efficient searches based on the caller's social connections, even with poor audio quality or heavy accents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If voice recognition technology is used to retrieve names from directories, then users can access contact information more conveniently, but accuracy decreases when audio quality is poor or users have heavy accents
Solution Approach 1:
The patent transitions from traditional linear directory searching to multi-dimensional social graph-based searching. Instead of only searching by name strings, the system incorporates relationship dimensions (friend, family, colleague), interaction history dimensions (call frequency, message exchange), and contextual dimensions (location, time) to identify the intended contact, thereby improving recognition accuracy without compromising convenience
Solution Approach 2:
The patent introduces social relationship data as an intermediary layer between the user's voice input and the directory database. This intermediary layer provides contextual information about the user's connections, allowing the system to disambiguate between similarly named contacts and accurately identify the intended recipient even when audio quality is poor
2Device complexity
If traditional linear search methods are used in directories, then the system structure remains simple, but search efficiency decreases as directories grow larger
Solution Approach 1:
The patent segments the large directory database into smaller social graph communities based on user relationships. Instead of searching through the entire directory linearly, the system divides contacts into relevant social circles (friends, family, colleagues) and searches only within the appropriate segment, dramatically improving search speed while maintaining manageable data structures
Solution Approach 2:
The patent performs preliminary organization of contact data into social graph structures before search operations. By pre-establishing relationship networks and interaction patterns, the system prepares the data in advance for rapid querying, allowing users to locate contacts quickly without requiring complex search algorithms to run in real-time
3Device complexity
If voice recognition relies solely on spoken names, then the system remains simple to implement, but accuracy decreases when users speak relationships instead of names
Solution Approach 1:
The patent creates a universal recognition system that handles multiple input types (spoken names, relationships, combinations) through a single integrated social graph search interface. The system can process various forms of user input and translate them all into effective search queries against the social graph, making the system both versatile and accurate without requiring separate specialized systems for different input types
Data Source
AI summary
In an example embodiment, there is disclosed an apparatus comprising an audio interface configured to receive an audio signal, a data interface is configured to communicate with at least one social graph, and logic is coupled to the audio interface and the data interface. The logic is configured to identify a calling party. The logic is further configured to acquire data representative of a called party from the audio signal. The logic is configured to initiate a search of the at least one social graph for the data representative of the called party to identify the called party responsive to acquiring the data representative of the called party.