Advanced research capabilities

• →

  Research workers now ask questions: if a research goal is now imprecise, or the need for more clarity is realized during runtime, flows will now pause and ask the user for more input.

• →

  Improved handling of data types: a new approach to entity inference has reduced instances of data type conflation.

• →

  Improved coreference resolution: information about entities in text is more comprehensively and completely captured as structured triples.

• →

  Multi-source entites: workers are now far more capable of stitching together information about entities from multiple sources, combining references into single output entities.

• →

  Fewer duplicate entities: improved duplicate detection results in fewer redundant entities being created as a result of research jobs the first place.

• →

  Inferred entities now include provenance information: PDFs encountered during the process are also saved to the user's graph and can be retrospected at any time.

• →

  Research tasks can now infer entities client-side: using the cookies from a user's browser, workers can now scrape authenticated sites (e.g. paywalled news websites, or platforms like LinkedIn/Twitter).

• →

  More informative “Activity Log”: more items now appear in the activity log while a flow is running, providing greater realtime insight into ongoing research jobs