– In this talk, Graype co-founder, Sanjaya, will share:

1. our transition from working with closed source to open source; 
2. the journey in developing winkNLP, our developer-friendly, JavaScript library for Natural Language Processing (NLP); 
3. the challenges we have mitigated and continue to address as the library grows and becomes more robust.

– Background: our modest start with small packages — mostly around NLP, and what encouraged us to take a plunge to develop a comprehensive NLP tool:

1. User feedback and community encouragement
2. Application of our packages by a range of projects like Social Analyzer and Trustroots, among others
3. Community role in transitioning licence to MIT from GPL
4. Our focus on standards, quality and documentation

– Beginning: in a landscape dominated by NLTK, spaCy and coreNLP, we began developing an integrated NLP product in late 2018 with the following objectives:

1. light weight
2. developer friendly
3. balance between performance and accuracy
4. support for pure browser side and server side

– Journey: Sanjaya will highlight how with a team of 2½ people we met our objectives and addressed the roadblocks we faced:

1. the difficulty in finding labelled data with permissive licences for training models;
2. and our alternate approach, where we:
3.  created language models using hierarchical finite state machines
4.  used limited open source data and complemented them with manual efforts
5. developing a declarative syntax to make winkNLP developer friendly
6. meeting performance and accuracy benchmarks
7. outcome — quick demo on web browser
8. learning from user feedback

– Finally, Sanjaya will talk about where we are today and headed next, including our:

1. Current challenges: sustaining an open-source project, and keeping pace with a fast-changing technological landscape; and
2. Future plans: to extend winkNLP to more languages that use the Latin script, as well as Indian languages; and our efforts to nurture a community around the project.