August 2020 Summaries

The upcoming features in Gretel Public Beta include automatic data labeling using Natural Language Processing (NLP) and neural network-based entity recognition for names and addresses, managed regular expressions, and custom extractors. These features enable the discovery of personally identifiable information (PII) such as full names and email addresses in datasets like Lending Club's financial dataset on Kaggle. Gretel helps developers share data more safely by providing workflows to understand and make informed decisions about data safety.

Gretel Synthetics has released new features for its latest version, making it easier to create synthetic data with a batch interface that works directly with Pandas DataFrames. The Batch interface automates manual steps and supports high dimensionality datasets by clustering like columns and training models on subsets of the entire dataset at once. It maintains correlations and statistical relationships between columns, allowing for scaling to highly dimensional datasets with minimal loss in accuracy. Users can create a synthetic dataset interactively using Google Colaboratory's batch training notebook. The Batch module allows validators to be set for each batch, ensuring that the output DataFrame has the same shape as the input DataFrame. Gretel Synthetics plans to release its custom validation package soon, which will automatically learn constraints in data and enforce them during generation.

The Contact Tracing proposal by Apple and Google aims to provide privacy preserving capabilities for informing people when they have been in close proximity to a COVID-19 infected person. The cryptographic portion of the specification has been implemented in Python, with a rudimentary "Life" simulator showing expected results. The main takeaway is that Apple and Google's specification substantially lowers privacy risk for contact tracing by not collecting existing data but generating new custom data specific to the use case. Handsets exchange secure unique bits of data using Low Energy Bluetooth connections, which are only possible when the handsets are near each other. This would be an opt-in program, with voluntary reporting of a positive COVID-19 test. The privacy focus is about ensuring the application and aggregator of the self-identifying subject data cannot identify who you are or who you have interacted with. The biggest risk to privacy will be how application developers implement the extraction and storage of Daily Trace Keys and what additional pieces of information they might request (or take) that are not within the scope of the initial specification.

Gretel is a platform designed by developers who have worked on various projects, including chatbots and cloud-connected kegerators. They realized that managing sensitive data can be challenging for developers working within organizations or collaborating with others. To address this issue, Gretel offers real-time tools to enable safe sharing and collaboration between developers and applications with any kind of data. The platform integrates machine learning, synthetic data, and formal reasoning to provide provable privacy guarantees for data. By incorporating privacy into developer workflows, Gretel aims to unlock siloed data and encourage rapid innovation in various fields such as healthcare, citizen science, and AI development.

FastText word embeddings can be used to quickly understand new datasets and build more consistent labels for structured data such as tables, JSON, or CSV files. The technique involves using a pre-trained FastText model based on schema examples from large collections of data. This approach helps in finding synonyms, abbreviations, and other variations of field headers, which can be useful when designing new table schemas or assessing the joinability of two tables. Additionally, it can aid in enforcing standardization policies across multiple internal data sources by comparing header suggestions with company standards.