Dear authors, Congratulations! The 5th ACM International Conference on Systems for Built Environments (BuildSys 2018) program committee is delighted to inform you that your submission #74 has been accepted to appear in the conference. Reviews and comments on your paper are appended to this email. The submissions site also has the paper's reviews and comments, as well as more information about review scores. Before the camera-ready deadline -- 20th Sept 2018, 11:59PM GMT, please kindly consider addressing the comments and incorporating the suggested changes. For paper format specifics, please refer to the publication chair's Note (updated just 20th Aug 2018): http://buildsys.acm.org/2018/resources/documents/HowTo.pdf Sincerely, - Marta and Polly, on behalf of BuildSys'18 PC Title: Wordly: Transferrable Semi-Automated Semantic Metadata Normalization using Intermediate Representation Site: https://buildsys18.hotcrp.com/paper/74?cap=074abqkw5Ewzkoo Review #74A =========================================================================== Overall merit ------------- 4. Accept Reviewer expertise ------------------ 2. Some familiarity Paper summary ------------- The paper describes Wordly, a framework to normalise metadata from building information to learn the meaning of tags in building automation systems. Wordly uses a number of learning mechanisms and can be augmented with expert human knowledge to learn from one building and transfer this learned structure to a new target building. This approach has the potential to significantly reduce human effort and can support interoperability among building automation system applications. Strengths --------- The Wordly framework combines a number of knowledge extraction and learning techniques that provide novel flexibility and accuracy compared to existing frameworks. It seems from the evaluation that Wordly has excellent accuracy even when using only a subset of building information. Weaknesses ---------- The approach was evaluated with existing, specific tag information that seems to be based on BACNet technology. BACNet is common un the US but less so in Europe. The questions is how transferable the approach is to the metadata of other building automation system technologies. Comments for author ------------------- Good paper. It would have been nice to see more variation in the evaluation, e.g. using different building automation system technologies and their metadata. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Review #74B =========================================================================== Overall merit ------------- 2. Weak reject Reviewer expertise ------------------ 3. Knowledgeable Paper summary ------------- A meta-data classification model framework is proposed to infer descriptive tags for the points in a building management system. This platform relies solely on a systematic deconstruction of the point labels themselves in combination with text-based feature extraction and modeling. The platform tests time-series features and find that they reduce the precision. Strengths --------- Metadata normalization is useful and applicable to a practical building analytics process. Weaknesses ---------- The paper presents no comparison of the technique with data or techniques from previous studies. As a result, I believe the generated technique may be overfitted to the three buildings tested, especially since there are from only two campuses. Comments for author ------------------- - In Section 5, the proposed technique is compared to the use of time-series feature extraction. It is found that there are significant limitations to the use of these features, especially in poorly represented classes. While this insight is interesting, there hasn’t been a very thorough analysis of the use of time-series data. Only a limited number of temporal features were extracted and imbalanced classification challenges should also impact the use of labels. Perhaps this section should clarify the deficiencies of this comparison (semantic vs. time-series data features) - Wordly would need to be implemented on a much more diverse data set in order to make any conclusions about its applicability * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Review #74C =========================================================================== Overall merit ------------- 3. Weak accept Reviewer expertise ------------------ 2. Some familiarity Paper summary ------------- This paper presents a learning framework called Wordly to map schema from unknown buildings to a target building. Strengths --------- This research is interesting and meaningful. The methodology is sound. Weaknesses ---------- The contributions of the proposed study is built on the BRICK which is not yet a widely accepted by the academia and industry. The validation section is weak. Comments for author ------------------- 1. Please discuss how the proposed methods can be used if other metadata schema other than BRICK is used. 2. More validations with more diverse buildings and building systems, and more comparisons with implementations of other methods in the same datasets, are necessary.