What non-IBM technologies are being used to enhance the IBM Watson ecosystem?
What non-IBM technologies are being used to enhance the IBM Watson ecosystem? by Vince Kasten
Answer by Vince Kasten:
IBM Watson offers some great capability and has the benefit of being a real product backed by a global company. The same goes for, say, the Ford Mustang. I just did a quick google search “ford mustang aftermarket” and it came up with 41 million results (in .46 seconds btw 🙂 ). Point being that there are ways of adding cool features to anything.
So, to the question:
Developers use other NLP technologies such as Stanford NLP (look for Stanford coreNLP) or DARPA NLP to augment or replace the Watson NLP. NLP pipelines have strengths and weaknesses that can be exploited for different uses. And NLP is often augmented with old reliable RegEx, which can be really useful in specialised language sets where there are easily identifiable and semantically rich keywords.
Also, I’ve seen solutions built that use Watson’s DeepQA capability, and then store pre-digested questions and answers in an external Solr database – it’s a good choice because it’s free, provides lots of open access methods, provides strong text-based searching, and it’s free. Also, it’s free.
The style seems to be that enterprise users who want to leverage cognitive to improve customer engagement or backoffice operations use Watson as a product more or less out of the box. Entrepreneurs who are building value added services with lots of embedded IP augment Watson APIs and capabilities to match their value prop – in this case the added technical debt is outweighed by the market value of what they’re building. Data scientists might use Watson’s Alchemy APIs in tandem with open source machine learning code (think projects like NuPIC or Weka) for purpose-built data ingestion.