Multi-scenario application of intelligent response in telecom industry operation and maintenance field

Inhalt:
With the expansion and development of telecom business, maintenance in telecom is facing more pressure and challenges, which needs digital transformation urgently. Work order is an important part of maintenance in telecom, mainly involving two types of scenarios: work order dispatching and work order interaction. Work order dispatching system faces the challenge of accurately identifying work orders and dispatching them to the corresponding departments in time, while work order interaction system needs to respond quickly according to the feedback from provincial companies and on-site operators to avoid the problem of delayed processing progress. Recently, artificial intelligence algorithms have shown superiority in solving the problems of these two scenarios. In this paper, we propose solutions for two scenarios. For work order dispatching, on the one hand, we propose a text classification algorithm based on work order pre-training model to identify work orders and dispatch them; on the other hand, we propose a Bilateral Branching Network in the field of text recognition(T-BBN) to build a work order dispatching system for the case of data showing long-tail distribution. For the work order interaction scenario, we perform model fusion based on four deep learning models, use contextual semantic modeling to achieve intelligent interaction with provincial companies or on-site feedback content, and innovatively incorporate shallow features according to business reality. The methods proposed in this paper all prove their effectiveness through experiments and can better accomplish the work order tasks in the actual telecom business.