A scalable automatic service discovery approach based on probabilistic topic model

Current service discovery approaches mainly focus on syntax matchmaking, which contains little semantic information to discover services automatically. This paper proposes a scalable automatic service discovery approach based on probabilistic topic model. Specifically, a novel service description model PTWSDM is proposed. With this model, heterogeneous service descriptions can be represented in a topic vector form on the same homogeneous plane. For the scarcity of word co-occurrence patterns in service functional descriptions, Biterm topic model is introduced to extract latent topics. Finally, a stream algorithm for topic model updating is introduced in order that the proposed approach is scalable and adaptable for large-scale dynamic registry. Experimental results confirm that the proposed approach outperforms the state-of-the-art solutions in terms of precision and normalised discounted cumulative gain values. It also has good time performance and scalability.