AI Conversation: Inside Rasa's open source approach

AI Conversation Inside Rasas open source approach

Statement: Rasa is not the only open source of natural language processing, but her large community suggests that she is doing something right.

Image: iStock / metamorworks

Want a fake chat (AI) platform? No problem - you only need to select one. Microsoft (LUIS) has one. So is Google (Dialogflow). AWS? Yep. (Lex.) But don't stop now: There are hundreds of options (from to SAP to Cisco's MindMeld to etc.).

Rasa’s approach may just stand out.

“We believe an infrastructure for a long-term conversational interface will remain open,” said Tyler Dunn, product manager at Rasa. To this end, Rasa, the company, opened up their machine learning framework to automate text and voice conversations. "The goal? To get rid of rules - based hard chat bots to AI that understands the context of what a person is saying.

I am not in a good position to measure the convenience of Rasa code. What interests me is just how much community the project has attracted. This may speak to the effectiveness of Rasa's open source approach, but also to the advent of, or soon to become, mainstream chat AI.


    More on open source

    The Rasa team may be right about the need for conversational AI to be an open source problem, but still wrong in its approach. After all, there are plenty of other open source chat AI platforms. Rasa is not the first to point out that developers prefer open source infrastructure.

    SEE: Managing AI and ML in the 2022 campaign: Technical leaders increase project development and implementation (Premium TechRepublic)

    While GitHub stars are an incomplete (true) measure of project success, they are a sign. Rasa has more than 10,000 stars, while other open source projects such as MindMeld (416), DeepPavlov (4,900), or BotPress (9,000) have fewer. Among this group, Rasa serves a diverse community: The type that wants to dig deep into natural language processing (NLP). In contrast, JavaScript developers like a project like BotPress that may or may not want to go lower in the stack.

    The Rasa community is interested in normalizing NLP. This is one reason it attracts over 10,000 people to the Rasa community forum. That's also why Rasa has over 500 people involved in the project. When I was surprised that there would be a huge crowd of developers with the ability to add meaningful code to something like Rasa, Alan Nichol, co - founder of Rasa and CTO, told me that it is “basically against that ”of what I recommend. No, not all of them will be experts in NLP, he continued, but there may be valuable contributions such as integration with various messaging platforms, or an extension of Rasa's capability to support new APIs. could use chat platforms.

    READ  Sony will start selling the official PlayStation 5 cover next month

    Even for those who do not return, it is important that Rasa is open, Nichol said:

    [C]onversational AI is one of the [areas of software] where you get the most benefit [from open source]. Because you can customize it to do it yourself, even if those changes you are pushing up the river are not, it's invaluable. Much larger than the number of people who could write customize something within MongoDB or something like that. The number of people who could write a custom NLP component to do emotion analysis or some sorting of their users, or just to tweak some hyper-parameters, use words which they applied to the corpus of their own company, all sorts of things. There are many, many ways in which people customize their software.

    The real competition for something like Rasa is customers who could roll their own chat AI bot, perhaps using TensorFlow. Rasa is built on TensorFlow, and for a team with enough skills, they were able to bypass Rasa and operate directly on the low - level TensorFlow. Rasa's bet is that most companies will not have the knowledge or patience to do this.

    They're also looking for something ready for production, other than projects like Uber's Plato or Facebook's ParlAI, which tend to target researchers. For Rasa, it's been important to combine language comprehension and conversation models into one end - to - end system, so that when you have messages you will have messages that do not quickly get into your scheme. not breaking down ("taking down)" is the essence of that user and converting it into a vector of floating point numbers to continuous representation, "is the more geeky explanation offered by Nichols).

    The good news is that you do not have to take my word for it - or the words of Nichols or Dunn. It is open. You can explore GitHub, customize it to meet your needs and hopefully submit a pull request to improve it.

    Disclosure: I work for AWS, but the views expressed herein are mine.

    See also

    Related Posts

    Deja una respuesta

    Tu dirección de correo electrónico no será publicada.


    We use cookies to ensure that we give the best user experience on our website. If you continue to use this site we will assume that you agree. More information