To Haskell or not to Haskell
Sometimes a programming language’s reputation can hold it back despite it being the best solution in the marketplace. Haskell – a standardised, general-purpose functional programming language – is a case in point. (It was named after the logician Haskell Curry.)
Programs in Haskell are a series of high-level generalisable functions that define what the program is intended to do, letting lower layers (compiler, runtime, and libraries) handle mundane low-level details such as iteration. This means the programmer can focus more on results rather than process.
But its commercial adoption has been held back by its reputation for being hard, too time consuming to assemble, as well as a lack of a commercial vendor to push its adoption.
For Daisee, the advantage of Haskell programmed software was obvious. It’s less bug prone, easily maintainable and scalable.
For Daisee’s principal software engineer Christian Marie, the fact Haskell is purely functional means the team are forced to explicitly model state and side-effects (such as input and output).
“This is a trade off,” Marie says. “By paying the cognitive tax of building a framework for our state and side-effects, we create for ourselves tools to reason about these things far more effectively than if we’d just begun typing without thinking.
“This alternate view of the world forces us into an architecture which naturally exhibits desirable attributes such as composability, modularity and referential transparency. These attributes reliably allow me to deconstruct now-explicit complexity in a way that scales with the chaos of ceaselessly evolving requirements.”
We’re now seeing a rapid adoption of Haskell by many data-intensive industries such as financial services, big pharma and biotech, and online consumer companies.
Daisee releases opensource libraries for transcription
Daisee has created opensource libraries for Haskell for transcription purposes.
The main feature is that files are uploaded to VoiceBase or Speechmatics to get processed and the result is a transcription.
This was created using the VoiceBase REST API http://voicebase.readthedocs.io/en/v2-beta/ and Speechmatics API https://app.speechmatics.com/api-details
The daisee VoiceBase library can be accessed at https://bitbucket.org/daisee/voicebase/src/master/
and the daisee Speechmatics library can be accessed at https://bitbucket.org/daisee/speechmatics-api-client
These libraries can be used by both data scientists and programmers for research purposes.