Sight-Machine-Factory-Namespace-Manager
Version: 1
This is a tool to help manufacturers create corporate standard data dictionaries of machine sensor data. With many generations of equipment and sensors, similar data fields often have different names, adding to the complexity of discovering and analyzing data. This model helps with translation from the original set of machine sensor names to a new corporate standard. It will attempt to understand the rules behind legacy naming schemes and map them all to a new enterprise-wide naming convention. This tool uses a fine-tuned Phi-3 model. By providing it with information on the new naming convention and a list of data fields to be renamed, it will provide a new name for those fields and a new confidence score on the potential name.
If you need to access to the model artifacts, please contact info@sightmachine.com
Primary Use Cases
Tag Namespace Manager is intended to be used when a manufacturer has a large number of machine sensor tags that need to be named or renamed according to a new corporate standard. Manually renaming tags can be a tedious and time-consuming process and requires extensive knowledge on the nuances of both the legacy and the target naming scheme. This tool will help automate and speed up the process. Having a uniform naming scheme allows the tags to be more understandable and usable by process experts on and off the line, as well as data analysts/scientists. When renaming a wide variety of uncommon asset types, or for highly complex corporate naming conventions, the language model may struggle to find matching patterns. Also, input variable names that do not contain clues as to the data payload (e.g. all numbers or names not based on English-words) may not map easily to the new naming convention. Subject matter experts should review variable name predictions.Out-of-Scope Use Cases
This model is trained to predict new variable names for factory plant floor data. Uses other than data dictionary management and applications other than plant floor data are out of scope.Responsible AI Considerations
Our language model is specifically designed for maintaining data dictionaries, a relatively low-risk activity. We have tested its accuracy using a variety of string distance metrics to ensure reliable performance. We are committed to continuously improving the model; however, we recommend that all results are reviewed by data and subject matter experts to ensure the highest quality and accuracy. We respect our customers’ privacy and confidentiality, and data entered into this model will not be used to train future iterations of the model. We've carefully designed and tested our fine-tuned language model, ensuring its responsible use through rigorous red teaming processes. This has enhanced the model's safety and minimized the risk of generating inappropriate content. Our commitment to Responsible AI principles drives us to continually improve our systems for secure and ethical deployment.Training Data
The model is trained on real and synthetic data from a variety of plant floor data sources in a variety of industries.Model Specifications
LicenseCustom
Last UpdatedNovember 2024
PublisherSight Machine
Languages1 Language