Broccoli AI at its best 🥦

Unstructured Data

Bengsoon Chuah highlights the untapped potential of unstructured data in the energy sector, particularly in safety reports. He notes that while structured data is often prioritized, unstructured data can reveal deeper insights into operational safety conditions 1. This data, often stored in platforms like Microsoft SharePoint, includes comments and observations that are not captured in structured formats. Chuah explains the challenge of extracting insights from this data, which is often overlooked but holds valuable information 2.

You have all this data that's ready for you, that has all the insight, that has all the information that is locked up waiting to be on earth.

--- Bengsoon Chuah

By leveraging machine learning, companies can scale the analysis of unstructured data, unlocking insights that were previously inaccessible.

   

NLP Bootstrapping

Bootstrapping data for NLP applications presents unique challenges, especially when dealing with nuanced and sensitive data. Bengsoon Chuah shares his experience with labeling processes, emphasizing the importance of understanding the context and nuances of the data 3. He describes the iterative process of working closely with users to develop a set of instructions for labeling, which involved a voting system to resolve contentions 4.

I noticed no matter how tightly knit your labelers or your team are, there will always be contention and you just got to work around with it.

--- Bengsoon Chuah

Chuah utilized tools like Argila to enhance the user experience and facilitate the labeling process, ultimately achieving a scale of 1800 to 2000 labels to train the first model 5.