New “AI in Action” Podcast Episode with Mark Kelly (Alldus) and Ivan Bilan (TrustYou)
Recently, Alldus International’s Chief Customer Officer Mark Kelly and I have set down together for a new episode of the AI in Action Podcast which you can listen to here.
This blog post will give you more insights into what we have discussed and also give you more background on some of the topics mentioned.
The episode covers the following topics:
- Introduction to TrustYou’s NLP solutions
- My academic and industry background
- How the NLP industry and TrustYou had to adapt to the pandemic
- The current state of NLP and best practices for building NLP projects
- Open source in NLP
Adapting NLP Solutions to the Pandemic
Last year or so has been tough on everyone. It has changed all of our lives, it has also impacted how we work and has reorganized a lot of our lives into new forms and shapes. Such a societal shift also had a considerable impact on the language itself and as a result also the existing NLP solutions. There have been a lot of new words introduced to our vocabulary and many existing words have changed in meaning. The most extreme example is probably the word corona, which coincidentally was also a brand name of a beer. As such, all existing NLP models need to be retrained with the new language data generated from 2020 onwards, as a lot of word meanings and associations that these AI models have learned need to be relearned and adjusted.
In addition to that, a lot of NLP research has been done to work with Covid-19 related data. TrustYou is also no exception.
In 2020 we have introduced a new product line that allows our NLP-based solutions to extract and analyze everything the travelers are talking about in their reviews related to public health safety. Our solution can extract insights related to how health safety standards are applied throughout hotels, whether the staff and guests are using PPE, how often and how rigorous the hotels clean their facilities, and a lot more. Find out more about our Hotel Relief Pack on our official product page.
Overall, we provide analytics insights that we extract from hotel and restaurant reviews to our travel industry clients, and these insights are automatically extracted and analyzed from millions of reviews. Our NLP solutions support 23 languages, including languages like English, Spanish, Traditional and Simplified Chinese for which there is an abundance of resources and training data, as well as many low-resource languages for which there are few NLP solutions and data available.
Best Practices for Building NLP Projects
In the podcast, we also discuss the best practices for building NLP projects. There has been a lot of work done in advancing the NLP ecosystem with solutions that can help build more refined and less biased AI models. Some great recent tools for more rigorous NLP model testing are checklist and TextAttack. There has also been a lot of work done in advancing MLOps practices and applying them to NLP.
If you are interested in the topic, make sure to check out my In Search of Best Practices for NLP Projects talk from PyData Pune 2020.
Open Source in NLP
The NLP community thrives on open source. A lot of amazing NLP frameworks are in free access, for instance, the transformers framework from HuggingFace is about to reach its 50k GitHub stars mark. There are also a lot of great open-source resources available for learning about NLP.
For example, I am the maintainer of the NLP Pandect, which is an organized list of all possible NLP resources per various NLP tasks. It also includes references to many other NLP resources, learning material, academic, as well as industry topics, and a lot more. If you are looking for anything NLP-related, you most probably will find it there. It’s also open-source, and any contributions to the content from the community are welcome.
About the Author
My name is Ivan Bilan, I am an Engineering Manager with academic and industry experience in NLP, AI as well as Data Engineering. Besides academic experience in Computational Linguistics and General Linguistics, I have also studied Technology Management and as a result, spend a lot of time within the fascinating start-up culture here in Munich.
Thank you for reading! Enjoy the podcast episode, and if you want to see more content like this make sure to follow:
