Luca Scagliarini: With the approach we chose, which originally was thought to be something that could not be implemented in the real world, we reached a good level of understanding. I think there’s more room in terms of being able to process content that comes from different domains in a much more effective way. For example, if I have speech-to-text, having something to understand the text can also improve the speech. I think that a combination of pure machine learning with semantic understanding is where this improvement will be visible to the average user.
Sramana Mitra: What do you mean by a combination of pure machine learning?
Luca Scagliarini: The pure machine learning approach is where you take a text and you don’t understand the meaning, then you just process the quantity and you try to understand the pattern and a lot of other things that are not really related to the meaning. If you merge that with something that can help you understand the meaning in context, I think the combination of the two will bring us to the next level of understanding. However, most of the companies have chosen the pure machine learning approach. They don’t have the semantic understanding. We already have the core technology. We hope to be able to show the market that this is doable and that the difference is significant. It’s the combination of the two approaches that count on the deep understanding of content. This can happen just by adding a semantic layer on top of traditional search engine.
Having a semantic layer that works on top of an existing search engine can give you a certain level of improvement. We think the direction is going to be there. The second direction is that there will be vertical applications. We see a significant improvement in the voice recognition. Adding a semantic layer there can even bring that to a higher level. In general, the combination of models that put together structured and unstructured data will bring us to being able to predict more and better. We hope to be able to contribute to the unstructured piece.
Sramana Mitra: In terms of industry vocabulary and industry-specific terminologies, once you work all of that in, there are industry-specific solutions that you can build that can be universally applicable within that industry. You can actually productize versus selling it as a custom software.
Luca Scagliarini: Absolutely. We want to make sure that we provide the technology that helps you in understanding as much as possible. Once you have that, it can be used by the domain expert to develop something that could be highly relevant to this specific domain. Where this can become big for us or for anybody who’s in unstrucutred Big Data is being able to provide the core piece that is then taken by the expert and design processes and tools around specific sections of processes. I think there’s a lot that can be done there. Everything starts from enabling a machine to understand content similar to how people understand.