SeNsER: Learning Cross-Building Sensor Metadata Tagger

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 950–960 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 950 SeNsER: Learning Cross-Building Sensor Metadata Tagger Yang Jiao†, Jiacheng Li†, Jiaman Wu, Dezhi Hong, Rajesh Gupta, Jingbo Shang University of California, San Diego {yajiao, j9li, j4wu, dehong, gupta, jshang}@eng.ucsd.edu † Equal Contribution Abstract Sensor metadata tagging, akin to the named en- tity recognition task, provides key contextual information (e.g., measurement type and lo- cation) about sensors for running smart build- ing applications. Unfortunately, sensor meta- data in different buildings often follows dis- tinct naming conventions. Therefore, learning a tagger currently requires extensive annota- tions on a per building basis. In this work, we propose a novel framework, SeNsER, which learns a sensor metadata tagger for a new build- ing based on its raw metadata and some ex- isting fully annotated building. It leverages the commonality between different buildings: At the character level, it employs bidirectional neural language models to capture the shared underlying patterns between two buildings and thus regularizes the feature learning process; At the word level, it leverages as features the k-mers existing in the fully annotated build- ing. During inference, we further incorporate the information obtained from sources such as Wikipedia as prior knowledge. As a result, SeNsER shows promising results in extensive experiments on multiple real-world buildings. 1 Introduction Sensor metadata tagging aims at understanding the context (e.g., sensor function and location) of a sen- sor from its name, which is essential to any smart building technologies (Wang et al., 2018). As il- lustrated in Figure 1, sensor metadata is typically a concatenation of esoteric abbreviations, each en- coding specific information about the sensor, in- cluding what they measure/control, where they are located, how they are related to each other, etc. For example, a sensor name SODA1R430 ART con- veys: the building name (SOD), air conditioning equipment ID (A1), room ID (R430), and the mea- surement type, which is area room temperature (ART). Running any application would require such SODA1R430__ART building name equipment identifier room id measurement type: room temperature sensor metadata ( or sensor name): label: AP&M.RM-B215..ZN-T building name room id measurement type: zone temperature Figure 1: Example sensor metadata from two buildings. Sensor name tagging aims to partition a sensor name into segments (shown in color) that encode key con- textual information about sensors. Different buildings adopt distinct vocabularies and naming conventions. contextual information; for example, to detect over- cooled rooms, one needs the temperature and the target temperature set for each room. Currently, learning a metadata tagger for a new building in practice requires extensive human an- notations, thus remaining a bottleneck in deploy- ing smart building techniques widely and effi- ciently (Wang et al., 2018). This is due to the fact that sensor metadata is curated by building-specific vendors, and that their naming conventions vary drastically across buildings, as shown in Figure 1. Anecdotally, annotating one sensor name may cost several hundred dollars, and it takes weeks to do so …

Content truncated to 500 words