New NLP people at Convertelligence

Eli Rugaard and Emanuele Lapponi

If you want a chatbot that can talk to users in a way that feels natural and helpful, you’ll need to use machine learning. We’ve been working with machine learning since day one, and last month we welcomed two new people who will be focusing their efforts on improving our machine learning models, making our chatbots smarter than ever.

Emanuele Lapponi will be working as a data scientist. He is currently finishing up his Ph.D. in language technology at the University of Oslo, where he’s been for the past 6 years researching everything from machine learning-based semantic analysis to classification of political party affiliation.

Emanuele Lapponi

He witnessed Convertelligence grow from the labs at the Department of Informatics to the company it is today, so accepting a job offer as a machine learning expert was not a difficult decision to make.

He will be working with anything that is NLP-related (Natural Language Processing) in Kindly together with the machine learning team. They are mainly concerned with the core technology in Kindly, which is the things that make the bot understand more than just what’s written, letter by letter. That’s the part where machine learning comes in. Machine learning is often referred to as artificial intelligence, although that tends to make it sound scarier than it really is.

Our new linguist, Eli Rugaard, chimes in, saying that a power struggle with robots is definitely not in our near future.

Eli will be working in our copywriting team, but she’ll cooperate more closely with the machine learning team, bridging the gap between writers and developers.

She spent the last two years working as a research scientist in Germany. Her previous employer, Nuance, is one of the world’s leading companies working with machine learning, especially within speech recognition and language modelling. Although she initially had an interest in natural languages, she also took some classes in computational linguistics during her master’s degree in linguistics. This combination makes her a great intermediary for our product and copywriting teams.

Her tasks will include, among other things, annotating data, evaluating chatbots and working on extracting linguistic features from the messages to get the most out of every message.

“NLP is an interdisciplinary field, so it has room for people who are more code-oriented, but also those with a background in linguistic theory, and that’s worth its weight in gold for us,” Eman says. “It’s rare to find a workplace like this, where front-end developers, machine learning people and writers are all sitting together. And it’s amazing to have that person right in the middle with a background in linguistics.”

Eli emphasizes that with a new field such as this, the only option is to hire people with lots of different qualifications and experiences and try to cover as many useful bases as possible.

“You just have to stay updated, pay attention to what’s happening, and constantly learn new things.”
— Eli Rugaard

Eli and Eman will collaborate in different ways, depending on what features and models they will implement in Kindly.

“Different kinds of models require different preprocessing steps. Eli’s
linguistic chops will come in handy as we discuss, say, which part-of-speech
tag-sets or syntactic theories are better fits for different scenarios, and
she has already started compiling and curating linguistic resources for our
next-gen models,” Eman says.

Eli explains that machine learning is really just trying something out and seeing what happens. And then you’ll try something different and tweak it a bit.

“In theory, it’s math. In practice, it’s more just trying something out, and then maybe trying something else. And then you just try, fail and develop it further.”

“It’s a process,” Eman agrees.

Eli Rugaard

Eli didn’t plan for a job in machine learning, but she knew she wanted to use her degree doing something useful. So when she was offered the position as a research scientist in natural language understanding (NLU for short) for German, she knew it was the right choice. It was a demanding job with challenging customers and a lot of things going on all at once. But she still found it to be a fun environment where she learned a lot and got to approach her tasks in a practical way.

“You’re told to find a solution within a certain time frame, and how you get there is just part of the process,” she says.

When it comes to the future of artificial intelligence, Eli is sure that the development will move fast forward. But this also depends on how skeptical people are of this technology. It’s just math and statistics combined with large amounts of data.

“It’s a helpful tool and not something we should be scared of,” she says.

Eman, on the other hand, is skeptical of other people’s perceptions of what artificial intelligence actually is. He says there are so many different views on the matter, that it’s difficult to even discuss what it’s supposed to be.

“If artificial intelligence is a machine mapping an input to an output, then we’ve already achieved it. This will only get better and better in our lifespan. If we’re talking about achieving the kind of artificial intelligence that is portrayed in Hollywood’s sci-fi movies, then I’ll leave that question to someone else because I find it completely uninteresting,” he says with a shrug.

Eli is very optimistic about the future of Convertelligence as well, stating that we’ve found a place in a market that is rapidly growing, and where we have a lot to offer.

“One of the things I’ve really noticed since coming here is how open everyone is to new suggestions. Everyone gets to voice their opinions, and everyone is involved in the big decisions. I think this really helps us to move forward,” she says.

Eman agrees and adds: “We’ve got so many great minds working together in the same place, yet we all have different backgrounds. That’s really unique for a business, but also very fitting for a place working with an interdisciplinary field like NLP.”

He says that just these factors are good indicators for our potential, but in addition to this, we’ve also got new models, we’re growing and we’re hiring a lot of new people. The models and the product can only get better.

“The future is rosy.”