Recurrent neural network learning for text routing

This paper describes new recurrent plausibility networks with internal recurrent hysteresis connections. These recurrent connections in multiple layers encode the sequential context of word sequences. We show how these networks can support text routing of noisy newswire titles according to different given categories. We demonstrate the potential of these networks using an 82339 word corpus from the Reuters newswire, reaching recall and precision rates above 92%. In addition, we carefully analyze the internal representation using cluster analysis and output representations using a new surface error technique. In general, based on the current recall and precision performance, as well as the detailed analysis, we show that recurrent plausibility networks hold a lot of potential for developing learning and robust newswire agents for the internet.