Jeanne Le Garrec: Advancing Natural Language Processing

Hi Jeanne, what do you do and what brought you to Singapore?

What brought us (my husband and our 2 year old) to Singapore nearly 2 years ago(#newbie) is simply curiosity! 
You cannot really understand the culture of a country without living there. I’ve lived most of my life in France, but after spending 3 years in England and 5 in the United States, Asia was the place to be. We wanted to learn, discover, and see what was happening on the other side of the world. We had never been to Singapore before and chose to settle down here anyway. Singapore was the obvious choice to fulfill our career ambitions, our passion for new cultures, and our desire to discover a vibrant innovation hub.

I am lucky enough to be working for Basis Technology. When I told them I was moving to Singapore, they asked me, what I needed to do my job? “An internet connection” I replied. The adventure began! They embraced the opportunity to have someone they trusted developing business and marketing activities.
Basis Technology has a suite of AI tools extracting insights from large volumes of unstructured text. Our most common use cases are:

• Identity resolution
• Entreprise Search
• AML/KYC enhancement
• Social media analysis
• Threat and crime detection

My official title is International Operations Manager, meaning I run operations in South East Asia while supporting our partners around the world. I’ve recently been assigned the management of our Startup Program to help find, collaborate and grow with this innovation ecosystem. 

How is the Singaporean market different from France or the US when it comes to business? 

Prospecting in Asia is definitely different than in the US or France. Getting a first meeting to inform customers or partners is easier in Asia or in the US as opposed to Europe.

Don’t get me wrong, it doesn’t mean it’s easier to sell! Networking seems more culturally spread in the US and Singapore. People want to know what are the newest applications they could use. Europe seems to be more interested in implementing tools that have been proven successful before and need more track records.

When it comes to specific projects (national security or threat detection for instance), governments are eager to learn from one another. The US intelligence community has been at the top of the game for several years, and foreign governments don’t want to waste time re-inventing. If they know that their counterparts in another country have successfully solved a problem they’re facing, they’d rather learn from it and use the same solution than creating something new that would be time consuming to develop and implement. There would be some innovation and improvements of cours.  In a way, that’s quite reassuring. 

I love the fact that the Singaporean government is betting on Artificial Intelligence and Data Science. They’re investing in motivating their population to be educated in AI. You have all these free programs for Singaporean to get trained like AI for everyone. These are great initiatives and really says something about which future they’re betting in. Same thing in China, we know the government is investing tens of millions of dollars to master AI. They joined the race later but they will for sure win it, placing their investments where the future lies. I strongly believe that Artificial Intelligence is shaping Asia’s future.

You work on a specific type of AI called Natural Language Processing, what is it exactly?

Natural Language Processing (NLP) is the analysis of natural language. It is one of the capabilities contained within the broad area of AI and one of the core elements used in enterprise search and information discovery. In order for the “machine” to give you the results you want, it needs to understand what you’re asking for and which document is most likely to answer your query. 
Linguistics knowledge, machine learning and statistical models combined enables the software to understand language “produced” by a human being. Natural Language Processing also includes speech recognition, text analytics, text generation, machine translation (like the Google Translate tool you might be using)…

The issues we’re facing now are how to sort and prioritize these results, how to reduce false positives and have access only to the relevant information. When making a query, you could get hundreds, thousands of results. But, how many are actually helpful and spot on?

I understand we need “better search” but why text analytics? Isn’t Google enough? 

We produce more text data than it is possible for someone to read. Just take yourself for example: think about how many emails you send, how many Social Media posts you share, how many documents you work on, how many articles you read a day, maybe you have your own blog… think about the amount of text data you consume per day, and times it by 365, times the number of people producing and accessing digital content. Getting dizzy, right?

So how are people going to navigate through this information or their company internal data? It always comes to the same struggle: what is relevant? where is the information I need? And also, how do you search when you can’t use a keyword search? What if the information you seek is available but in a different language?
The famous search engines that we all know (Google, Yahoo, Bing…) are easy to use and great to navigate around. The issue is that they don’t help search on your internal data, they don’t analyze your data; plus it’s keyword based, which is limiting… This is why, for quite some time now, companies have been looking to implement “smart, Google-like” search engines to navigate through all the data they have, but they also need the analysis tools for accuracy! This is called Entreprise Search.

I’m sure you have heard about Data Analytics and how companies have to use it to make more insight-driven decisions to be more competitive. How do you make the “right” decisions if you do not have all the information in hand? What I find really exciting about Entreprise Search, NLP and AI is that their applications are only limited by our imagination! Your business, your reputation, your security depends not just on how much information you have available but the value you can get out of it on the spot!

Keyword search is only one way to go, limiting the exploration of available information! On the contrary text analytics can sort documents to be filtered by topic, concept, language, category, feelings, emotions, entity type… As we become more data driven, as we automate steps, operate more critical tasks, evaluate risk using AI, we need to go beyond the conventional keyword search to uncover the accurate and relevant information. The goal is to go one step further in accelerating decision making and maybe even try to predict the future.

You are working mainly with banks and government agencies around the world. What are their main concerns? What will the trend be? 

They all want the same thing, getting quicker and more reliable answers to their questions. Tools used before by the intelligence community to detect threats and stop terrorist attacks are now being implemented in the financial world to fight fraud and money laundering. Law firms or private companies also want faster data search to capture key information.

Whether we talk to American, French, Australian or Singaporean banks: automation of processes is their focus, more precisely “digitization and automation” is what we’re hearing over and over again. Financial institutions are looking for the right tools to accelerate screening and onboarding. They need to comply with all the newest and even stricter AML/KYC international regulations without slowing down their business.  They turn to AI and technology to reduce risk and be more efficient.

Some teams can spend 12-16h checking a complex transaction or a company’s background. Operations want to reduce that time to just a couple of hours but without threatening the quality of the work. They can’t “afford” missing one hit, missing one person of interest, missing one threat…

When it comes to working with government agencies, extracting leads and intelligence from social media has become essential for several reasons:  

• Fake news management. This has been a real international buzzword! How to identify fake news quickly and contain them? Singapore is actually the first country to take a “legal” action against fake news.
• Threat and crime detection. Governments want to implement tools to help them crawl all the webs (clear, dark, and deep web) and get the most out of social media for predictive analytics applications. The Embers project, now called Presage, is a platform that has been able to predict major events in Venezuela and Brazil by analyzing blogs, social media posts, images….The initial goal of this research was to predict civil disorder to help police forces anticipate and prepare for such events. This research showed that they could, on average, predict an event, 7 days in advance with a result of 80 to 90 percent of accurate forecasts! Scary and exciting at the same time! Now, they’re working on the same model to predict disease outbreaks.
• Assessing public opinion. Governments’ interest in social media can also be more “commercial.” Just like a company would exploit it for brand management and satisfaction assessment, governments will use it to learn what people think about the president, the minister, a new policy, a new law, a public event.

The next step is to continue to leverage NLP for more automation and deeper analysis. For instance, customers ask for more granularity in emotion analysis in surveys, for their chatbot, in comments, in posts… There is a continuous need to go deeper and for more customization. 
Another request we hear more and more is document summary based on key elements and sentences. This leads us to machine generated text. We’re not in production yet for this feature, but getting close! Once this is mastered, a whole new world of automation will open its doors. I can imagine people not just wanting a summary about one document, but a summary of a set of documents.

Do you remember when I said earlier that companies try to predict the future and that imagination was the limit? The main question is: what is your ultimate goal? What do you need to do tomorrow that you cannot do today?

In a nutshell, what questions Corporates should ask themselves to start their text data analytics journey? 

Whether it’s a non profit organization, a bank, a startup, an agency, they should ask themselves the following questions:

Have I got all the information I need to make the right decisions?  
If the answer is no, then:

• What am I missing to be more competitive/more productive?
• What information am I missing that I wish I had?
• What type of content do I have available?
• How much do I really know about what’s in my text data?

Once you know what you have, think about how you use it:

• Am I using manual processes to perform critical business analytics?
• Do I increase risk by not using all of my data?
• Am I missing things I know I have? 

Finally, think about what you want:

• What do I want to know from my data?
• How do I want to visualize that information?
• Who should have access to which level of information?

This should help you drive your evaluation and transformation to get more from your text data. One thing to keep in mind though is that data is so vast and diverse! Exploiting it correctly takes many steps, and you should not be fooled! No provider can be an expert in everything: data cleaning, analysis of text, voice, images, videos, figures, data aggregation, visualization, etc.   
To survive, a company needs to evolve with its market and continuously integrate new technologies. It’s via the combination of human skills and technologies that a company will stay competitive and attractive for customers.

How has COVID-19 impacted your business? 

My company being global had actually made the “work from home” bet a couple of years ago, so for us the disruption to our organization and work habits was really little.

Our sales cycle is quite long so even though we expect a drop in our projected revenue, it won’t be catastrophic. Luckily, We don’t work in the retail or tourism industrywe don’t rely on the tourism or retail industries. Banks see an increase in the number of loan applications they have to process which means they need to ramp up their identity resolution resources and our national security government projects do not stop because of COVID-19.

Actually our R&D team has been working on several projects showing that NLP can be useful during this time: 

• social media analysis to discover outbreak hotspots
• model adaptation for medical text processing and healthcare chatbot
• fuzzy name matching for patient identity and claim verification

How dare you to be a woman in the international world of tech?

I don’t dare, I just am. I’ve never placed or behaved myself differently in meetings or social events because I’m a woman. I’ve always worked in environments where the majority of the stakeholders were men and never felt I was less considered or less heard.

Actually never, is probably not 100% true, I just don’t care enough to pay attention or give importance to unsupportive behaviors. When you have something relevant to say, people will listen, sometimes you just need to adapt the way you say it. I also have the feeling that there is real solidarity between women working in a masculine industry, or at least in the tech industry. We got to help each other out.
I’m very thankful for the women I’ve met in Singapore who have opened their network and shared their knowledge. That’s how we will all move forward.

How do you bring more diversity?

Well first, I have diverse experiences myself! I’m happy to speak English with my somewhat British accent while proudly wearing my Boston Red Sox cap and eating homemade quesadillas and frijoles (leftover from 6 months in Monterrey, Mexico). And on top of all that, I’ll be bragging how great french wines are!
More seriously, Singapore is a perfect example of diversity. You never know who the person you’re talking to is from and that’s the beauty of it. Team work here is by default international. Each person has his/her own culture, own reasoning, background, business habits, vision, own ambition.
I also volunteer for a local non-profit organization Hello Tomorrow Singapore, which is part of a global non-profit. Its raison d’etre is to bring together researchers, entrepreneurs, academics, and governments to raise awareness and for investment  in Deep Tech. This brings a very diverse and exciting ecosystem together.

The team I work with here is amazing:  knowledgeable and passionate about this initiative. The startups and researchers we meet are all working on breakthrough innovations in health, agriculture, well being, aeronautics. This group also facilitates   networking with players outside of Singapore such as Cambodia, Indonesia, Hong-Kong, Thailand. An interesting melting pot of thinkers and innovators. 

Diversity in culture, backgrounds, age, sex, experience is necessary to come up with breakthrough discoveries and reasoning. People with the same mindset, will get along for sure, but will get stuck quicker. Nearly 10 years ago, there was a study showing that “the existence of collective intelligence among groups of people who cooperate well, and the tendency to cooperate effectively is linked to the number of women in a group,” going further Professor Anita Woodley, who worked on the analysis of this research, added that “having group members with higher social sensitivity is better regardless of whether they are male or female.” Proof that diversity in cognitive skills and character impact an entire team for the better!

Jeanne, thank you ! 
Get in touch with us @ womenfrenchtech at gmail dot com
In collaboration with Amel Rigneau & Sophie Guo