為了您的方便，有利於英語語言版本的這個博客成幾種語言的翻譯提供了一個翻譯小部件. 如果您選擇使用這種自動的翻譯設施, 請理解可能有偏差自動的翻譯和英文原版之間. IBM 並不負責任何自動化的翻譯偏差和提供翻譯的版本 "正如" 沒有任何形式的保證.
BondEvalue was the 2017 Watson Build Asia Pacific Geography Champion. The company earned the accolade for an innovative mobile app that uses IBM Watson to identify relevant information for bond traders.
Imagine you’re a bond trader. To make the right decisions about buying or selling bonds, you need fast access to relevant information. But there is simply too much information to sort through. Each day, there might be hundreds or thousands of news items published on a single corporation. Only a small fraction of those news items are important for your work.
Large bond trading firms have deep understanding about what constitutes relevant information and also extensive resources to search through the flood of news for what’s important. But individual traders and smaller firms are at a distinct disadvantage.
At BondEvalue, our vision is to level the playing field. We want to give individual bond traders and small firms simple and immediate access to relevant information. Our plan was to create a mobile app that would capitalize on the power of artificial intelligence (AI) to place vital information right in the palms of users. Through the Watson Build Challenge, we were able to move forward with that plan, advancing from concept to prototype and ultimately to launch.
Making the most of AI
AI has amazing potential for helping to find relevant information in huge volumes of data. But there can be substantial training required.
It’s a little like raising a baby. Let’s say there was a way to measure the IQ of newborn babies, and you discovered that yours was a genius. You would still need to provide education and training so your child could reach his or her full potential.
Our team at BondEvalue has a wealth of bond trading experience—more than a century of experience in financial markets among all our team members. As we embarked on this path of incorporating AI capabilities into our solution, the question was: How do we transfer our experience and knowledge to Watson? When Watson encounters news stories or new information, we want it to behave like an experienced bond trader.
We discovered that part of the process of transferring knowledge includes redefining terms that have industry-specific meanings. 舉個例子, in the bond market, an “issue” is when a bond is made available by a corporation or government organization. But in standard natural language programming, an issue is a problem.
Another example: In the bond market, when a “yield” goes up, it’s bad news for the bond holder because the price goes down. But in natural language, an increasing yield is generally a good thing. We had to make sure that Watson understands these and other terms in the way a bond trader understands them.
Overcoming challenges through Watson Build
Through the Watson Build Challenge, we were able to navigate through these and other challenges with incorporating AI capabilities into our solution. The IBM team provided a broad array of resources that helped us shape our business proposal and then develop our prototype.
尤其是, we greatly benefitted from working with our Watson Build mentor, Sarma Gadepalli. In our weekly meetings, we presented our latest questions and problems, and he either provided answers on the spot or found the right person at IBM who could help. It was very useful to have a single, consistent point of contact throughout the program. And in fact, Sarma was such a huge help to us that we’ve continued to work with him after the Watson Build Challenge ended.
We were also fortunate to have access to a variety of other people from IBM, right up to senior management in the APAC region. We had productive discussions with IBM team members during every phase of the Watson Build Challenge, through the final presentations at the end of the program.
Moving from prototype to mobile app launch
Hearing that we were the Watson Build Asia Pacific Geography Champion was very exciting. We had spent many months—and long nights—finding a solution that worked, and it felt great to receive that validation from IBM.
But every one of the 400-plus teams that completed a prototype for the Watson Build Challenge deserves to be considered a winner. Building a prototype is a huge undertaking.
當然, there’s even more work to transform a prototype into a product ready for launch. Prototypes are designed to work in specific, well-defined situations. We designed ours to find relevant information for maybe 100 bond-issuing organizations. But for the launch, our app had to work for all bonds, at all times. We generated about 10 times the training data after submitting our prototype and before launching the app.
幸運的是, all of the collaboration we did with IBM helped provide a strong foundation for moving ahead toward the launch. And so far our mobile app has been a great success.
Looking ahead, we’re now working with IBM to identify banks and other financial services businesses that might be interested in offering our solution. Because our app is built on Watson, businesses have the confidence that it works, and they know it is powerful.
Offering a little advice to future Watson Build participants
I highly encourage other teams to participate in the next Watson Build Challenge. This is your moonshot. Once you’ve identified a really tough problem that’s worth solving, IBM can help you as you build an innovative solution to solve it.
To learn more about the BondEvalue app or sign up for a free trial, 訪問: bondevalue.com.