Artificial Intelligence & Natural Learning Processing
DataHex harnesses AI by leveraging the latest advances in NLP and ML for faster discovery, predictive capability and to detect irregularities in both structured and unstructured data sets.
Leveraging behavioral data to develop market Indices
- PEXA, a world-first digital settlement property platform helps homebuyers and sellers track their settlement progress in real-time.
- To provide greater insight into the mortgage refinancing industry PEXA wanted to develop a Refinance Index enabling mortgage lenders, mortgage securitization holders and government agencies to understand trends, benchmark market share by sector and understand emerging industry risks.
- The Refinance index is calculated against the total market activity at many geographic levels i.e. national, state, metro/regional, local areas etc.
- The index provided PEXA with a unique market insights and benchmarking data product to license to industry participants as well as industry observers looking for lead economic indicators.
- RoZetta data scientists leveraged behavioral data to develop an effective residential refinance index for PEXA to measure the level of mortgage refinance by geographic market.
- RoZetta designed, developed and delivered the Refinance Index through a automated cloud solution with monthly data updates and the production of indices at all levels of geography (i.e. State, metro/regional, local area markets.)
- The final product seamlessly incorporates into the PEXA web portal to provide subscribers with access to the index history and updates.
- The index tracks the refinance activity value and volume historically and allows for comparison between geographic levels providing the user with invaluable benchmarking insights.
- The Refinance Index provides data-driven insights into property data crucial for governments, consumer lenders and holders of securitized portfolios. See the PEXA refinance index tool here.
PEXA was able to accelerate the speed to market in launching the Refinance Index by partnering with RoZetta Technology and accessing a unique blend of industry knowledge, data science and cloud capabilities
The Index enables industry participants to benchmark and compare their mortgage portfolio changes compared to the market - clearly highlighting real gains or losses in market share
Surface any underlying portfolio risks and help with growth and retention resource allocation
The Index creates a new revenue opportunity for PEXA generating value from their rich property market data by providing accurate and timely insights into market performance
As a portfolio management tool, the index is crucial to provide a foundation for future indices and metrics to provide transparency and encourage competition in the highly competitive property market
The Index is a market first in terms of providing quantitative time-series trends at all levels of geography
Utilizing NLP to enhance discovery for unstructured data
- Third Bridge is a market-leading global investment research provider for human-led insights to support capital markets firms with their decision-making process.
- Third Bridge engaged RoZetta to provide an innovative solution to automatically tag forum transcripts with companies from the reference data set previously mentioned in interviews.
- Customers needed a better search experience and wanted to discover relevant content more efficiently.
- Tagging the transcripts to provide a strong foundation for additional enhancements.
- RoZetta’s data science experts, and DataHex platform, mapped entities within 23,000 transcripts, from which over 2.9 million entities were mentioned, identifying 125 entity mentions per transcript on average.
- Enabled linking to additional data sources such as company fundamentals, news, data from other providers and alternative sources of unstructured text.
- Tracked sentiment of entities over time.
- Automated summarization of transcripts.
- Additional entities such as People, Locations and Industries were tagged.
- Developed an objective relevance measure for the identified companies mentions, initially based on transcript content with the ability to further enhance customer activity insights.
RoZetta was able to successfully tag over 152,000 entities by leveraging NLP methods. This was ten (10) times more tags than previously identified by the client
RoZetta’s models achieved discoverability of entity mentions by 98.3% versus 37.4% current state, resulting in a substantially better search and transcript filtering experience, generating more relevant results and watch list notifications
Automated summarization reduced manual processes and improved efficiency
Additional entity tags allowed the client to enhance its search and discoverability
Linking various datasets meant the ability to extract relationships between entities
Sentiment Index efficiently generated insights into the perception of the market
Improved customer interface increased client engagement, reducing attrition, propelling customer growth, and lifting revenue