The Civic Opinion Express Platform in a certain district consolidates various channels for public opinion events, including 12345, Yue Zhengyi, District Mayor's Mailbox, Secretary's Mailbox, Integrated Coordination Platform, Internet of Things Perception Platform, and SMS Platform. It involves 1842 classification criteria for event grading. The primary method of classification relies on manual efforts by platform dispatchers, handling a daily workload of approximately 800 cases per person, with peak volumes reaching 2000 cases per person. This demands substantial human resources.
The project utilizes the Rock system to govern a massive volume of historical public opinion event data. Through in-depth data analysis, it involves the refinement of classification relationship mapping standards and the integration of data labeling. Simultaneously, employing advanced semantic matching technology, the project constructs an intelligent routing engine. This engine facilitates the automatic grading and classification of events, assisting users in enhancing the efficiency of event processing.
Multiple channels for event reporting, with several business systems employing the same manual classification process.
Reliance on individual experience to determine the category of lengthy event description texts, resulting in slow processing efficiency.
New personnel require training and extended practice, leading to a high rate of event rejections.
Fusion and mapping of annotations from multiple data acquisition channels, integrating analysis and exploring the inherent relationships between events and standards.
Empowering long-text contextual recognition through core algorithms such as rule input, text Named Entity Recognition (NER), and deep semantic recognition.
After training and validation using historical data, integrating vast corpus data for self-learning to enhance and improve
Full coverage of event classification, achieving a 100% event routing rate.
Classification speed for individual events is less than 0.8 seconds.
Event classification accuracy exceeds 86%.
Grandhoo