The following are the learnings from the podcast:

  • Research Areas :

    • NLP for text analysis. Topic Classification, Entity Extraction, market impact, topic detection
    • Recommender systems- You are reading this- You might be also interested in
    • Pricing illiquid securities based on reference data, pricing data from proxy instruments, discrepancies - move from consensus price to creating features to explain the discrepancy . Unlike individual contributors, we see a lot of data and hence can train better models
    • Anamoly detection and other analytics from social media data
    • Question answering - Watson for finance

  • Topic classification models are different from topic modeling . Classify news in to various topics for all financial data. 5000 hierarchy of topics - mixture of manually curated and machine generated

  • Measure of precision: standard measures are insufficient. They are to be optimised to the specific usecase- usecase is search for filtering content .

  • Hire from masters programs

  • Optimisation skills are Super important

  • Spark conference- lots of Bloomberg folks attend it

  • Financial indicators from unstructured data

  • SVR - Large margin methods for at sentiment analysis

  • Broad trend is quantitative methods are making in roads to finance - based on Spark . Adaptable to finance.

  • Looking for empiricist mindset

  • Experience with building models