Financial services are traditionally a data-heavy industry full of transactional data, user data and textual data.
While these data sets are very rich in information and could provide valuable insights, 80% of the information is unstructured and analyzing them manually with human efforts would be impossible.
That’s why natural language processing (NLP) comes with so many advantages for the financial sector. In this post, we'll cover some popular use cases including real world examples.
What is natural language processing?
Natural language processing as a subfield at the intersection of linguistics, computer science, and artificial intelligence. It is a set of technologies used to enable computers to understand and process texts written in natural language. This includes company emails, CRM entries, market reports and analysts’ recommendations, and much more.
Before the machine learning era, the vast majority of computer systems were based on hand-written sets of rules, making these systems troublesome to deliver and of limited scalability. Machines also lacked the understanding of the processed text, further limiting the use cases of the early NLP systems.
Using NLP, the words are transformed into multidimensional vectors. They show the relations between words by mapping their connections based on thousands of examples in the texts the models were trained on. The understanding of the meaning of a particular word, the relations between words in the sentence and the neighboring sentences came as a revolution in the way computers process texts.
How NLP supports financial services
Delivering the ability to scrape through text and interpret words in their context, NLP finds multiple use cases in the financial sector.
Trends spotting and predicting stock behaviors
The automated monitoring of social media can come as a great way to spot the investment opportunity ahead of the market.
“Many of you remember the “GameStop” craziness, which made data scientists scrape data from Reddit trying to get ahead of the curve and find the next meme stock,” comments Alan Feder Principal Data Scientist from Invesco and one of the speakers during the recent Data Science Salon conference. “You can look at tweets to predict which cryptocurrency will become popular soon,“ he adds.
Information retrieval to gain company insights
A financial analyst’s bread and butter are to read thousands of pages of financial documents.
“A common problem identified by S&P Global Market intelligence is that our clients need to review large bodies of unstructured data to find a domain-specific signal,” says Moody Hadi, Group Manager - New Product Development & Financial Engineering at S&P Global Market Intelligence. “You need to read through thousands of pages to find what the company disclosed about themselves. The process is manual, highly inefficient, and dependent on human expertise” he adds.
The NLP-based tool can scrape texts and find relevant information efficiently and swiftly, significantly boosting the efficiency of the analyst.
Summarization of company news
Another good example of augmenting the analyst’s work is automated text summarization tools. They can scan through hundreds of pages and deliver summaries containing only the relevant information.
“There are a myriad of ways this can be useful. Everyone has too much to read and too little time to do it and summarization can be used to decide whether they need to get into the whole thing. Long sections of management discussions and analysis can be shrunk into more manageable lengths. Or if there is a long news article about the company you consider investing in, summarizing it can get it just to the point.” comments Alan Feder. “You can even summarize the long emails from your boss!”
Watch Alan's full presentation about NLP in Finance - Beyond the Search for Alpha in the video below and learn more about methodologies within NLP and how they are being used in finance to help professionals do their job more efficiently and effectively:
Sentiment analysis in documents and news
With the understanding of the relations between the words used in the sentence, the machine can dig through text to understand the sentiment towards a subject. Thus, the analyst can get a short summarization based not only on facts delivered but also the emotions of the authors of the documents and news reports.
According to research published in Nature, the sentiment in the media toward one company can affect a whole network of companies in the same industry. Also, there is a statistically significant association between strong media sentiment and an abnormal market return. Companies such as S&P Global use NLP to analyze earning calls and media reports related to company announcements, which help them predict the healthiness of a certain company.
Thus, having swift access to the sentiment expressed in the media coverages and the documents published by companies can deliver a significant competitive advantage over other investors.
Document classification
The NLP-based system can also deliver real-time analysis of the data flowing through the institution. For example, it can scan documents and, depending on the content, send it to the proper department or specialist, saving them time and effort.
Scanning contracts and documents to support compliance
The automated systems can scan through the documents in search of the patterns and boost compliance. According to the data shared by IBM and Chartis Research, it is possible to automate up to 90% of the process of transforming the legal and regulatory documents of hundreds of pages in length into rules to be applied to systems and company policies.
The NLP-based solutions can also scan the documents in search of patterns associated with fraudulent activities or money laundering.
Chatbots boosting customer experience
The NLP-based solutions are seen not only in the back office but are also used in contacts with customers. These assistants are available 24/7, always patient and well-informed.
According to Drift, up to 80% of companies are using a chatbot or a similar conversational marketing solution. Moreso, among this 20% not having one, 74% think about implementing one soon.
Financial services are not an exception in the trend of delivering great customer experiences and the need to be more fitted to the audience’s needs.
Summary
The NLP-powered solutions can be useful in the back office, customer relations, and compliance among others. But in fact, the core of natural language processing is the support it delivers for human specialists.
“Even in other industries, you can use NLP products and skills to amplify the performance of your specialists. Even if they already are doing a great job” - comments Alan Feder. “Just try a bunch of them and see what works best!” he concludes.
Read more about how NLP is applied for industry specific tasks in this post about The Role of Natural Language Processing in Healthcare.
