Currently, the applicability of recognition approaches is limited to only native (known) appliances for which training data is available. It means the appliance with no training instances appears as an alien to the approaches. An alien (new) appliance may introduce as household any time by the electricity consumer. The central focus on this paper is on building an appliance recognition approach that can accurately identify both native and alien appliances by leveraging semantic information. This work also collects electricity usage data by deploying smart meters in an apartment complex, for experimental evaluation. The initial accuracy results are satisfactory and validating the effectiveness of our approach for alien appliances.