Aided Intelligence for Finance

Quantitative methods work fine and help the operators' decision-making process by suggesting a "normal" path of the prices but, at the end, what is important is the news and people's reaction to it. ayfie enables quantitative analysis enhanced by qualitative evidence. ayfie's toolset is able to summarize, reduce or categorize the input qualitative data and produce analysis directly related to the actions that will be taken by the operators as an enhancement of their decision-making process.

In the financial community, news, rumors and facts are among the most important factors that determine the operators behavior. Operators, in fact, are much more influenced by news than by analysts' forecasts or historical price analysis of shares. When a news such as "the inflation rate is expected to increase next month" arrives, the consequences are immediately visible and operators base their decisions on their personal experience and on other's people behavior, rather than on expensive and complex forecasts produced by complicated neural-networks financial forecasting systems. 

The financial operators and information providers understood the importance of qualitative data as the key-point in the trading decision-making process long time ago. Therefore, the emphasis has been on providing as much relevant qualitative information as possible. Financial operators receive, in real-time, news regarding information on things such as companies, macroeconomics or politics. They also have access to huge quantities of past information.

The language and information needs to be managed in an effective and efficient way, thus ayfie.

Sifting Through the Fluff

In their daily work, portfolio managers and traders need to stay up to date with all relevant information about events that might influence stock prices in their portfolio. They also have to keep an eye on other analysts’ views of stock performance and other metrics.

Although several services providing this kind of information are available, it is a significant, manual effort to sift through all the relevant sources and find all the important bits and pieces. It is virtually impossible for a single person to retrieve all news on all stocks contained in a larger portfolio. This leads to either aggressive filtering or lots of person-hours spent on getting the data - time that is not available for using the information to act on it.

This task is especially tedious. There is is a lot of overlap between the news outlets, which should be ignored. But, there are also dissenting opinions, which should be analyzed.

Two news items can be dealing with:

  • the same company / stock, but with different events concerning it.
  • the same event, but with different interpretations / perspectives.
  • the same event with the same perspective, potentially just listing facts without interpretations.

A human reader will mentally deduplicate news of the third type, aggregate and analyze news of the second type and categorize and filter news of the first type.

A combination of rule-based information extraction and statistical analytics can greatly reduce the effort involved in retrieving and digesting all this information by applying the pre-filtering steps automatically.

Deduplication of Events

Events that pertain to a specific company can take many different forms in news outlets. Some might just repeat the respective dpa or Reuters text, others will change the wording, add interpretations and additional facts. While the latter makes for a better and possibly more interesting read, it is not ideal for consuming a large amount of information in a short time frame.

Consider the following news quotes:

Adidas AG appointed Kasper Rorsted as chief executive officer to succeed Herbert Hainer, charging the head of Dial soap maker Henkel AG & Co. with restoring growth to the German sportswear company. Rorsted, 53, will join the Adidas board on Aug. 1 and take over as CEO two months later, the Herzogenaurach, Germany-based company said Monday.

Adidas said Kasper Rorsted, currently chief executive at Henkel AG, a German maker of cleaning products, adhesives and beauty-care items, would become a board member at Adidas as of August and CEO in October. Henkel had announced Mr. Rorsted’s resignation earlier in the day.

Adidas has named Henkel's Kasper Rorsted as its new chief executive, boosting its shares as investors expressed hopes that the Dane has the credentials to boost profits at the German sportswear firm. Rorsted, 53, will leave Henkel in April after eight years at the consumer goods firm and take charge of Adidas on Oct. 1, succeeding Herbert Hainer, 61, who has headed the firm for 15 years as the longest-serving boss of a leading German company.

All three articles deal with the same basic fact but use very different wordings to express it. If all three of those news outlets were part of a financial information service feed, an analyst would have to skim over three text snippets without gaining any additional insights.

Applying so-called information extraction and codified linguistic knowledge to this problem makes it possible to identify these texts as duplicates with a very high level of precision.

This is not an easy task since it requires knowing the predicate "employed", which has a person, a role, a company, a start and an end date as arguments. It also requires being able to deal with a wide variance of time expressions such as "Aug. 1 … two months later…", "in October", "on Oct. 1".

By applying pre-built, local grammars for person names, companies and time expressions, along with synonyms for how the predicate "employed" might be expressed, it is possible to derive the same formal representation from all three texts.

employed (Kasper Rorsted, CEO, Adidas AG, October 1st)

After performing this operation for all news texts in all news outlets, it is then an automatic task to identify the parts of the texts that are duplicates to be suppressed and to identify those which contain new information. Thus, only the most recent or most reliable instance of every fact will be presented, while the others only serve as supporting evidence.

Adidas AG appointed Kasper Rorsted as chief executive officer to succeed Herbert Hainer, charging the head of Dial soap maker Henkel AG & Co. with restoring growth to the German sportswear company. Rorsted, 53, will join the Adidas board on Aug. 1 and take over as CEO two months later, the Herzogenaurach, Germany-based company said Monday. (also reported by WSJ and Reuters)

Additional sources will then only be displayed if they provide additional information or a different perspective on what is already known.

Aggregation of Structured Information

While the problem of duplication is certainly an important one and has a high impact on the efficiency with which information can be consumed and structured, there are many more interesting applications of this technology.

For example, by extending the simple "employed"-predicate we saw in the earlier example, we can continuously build a table of the changing key employments in companies, even if we only have incomplete information.

Based on this information, it is now possible to make sense of several news articles at once. For instance, we may follow the career of an executive and correlate what his advent or departure might mean for the success of a company.

Person Role Company Start End
Kasper Rorsted CEO Adidas AG October 1st 2016 -
Kasper Rorsted - Henkel AG - April 30th 2016
Simon Perkins MD Ncell April 12th 2016 -
Nigel Morrison CEO Skycity March 1st 2008 April 29th 2016

By aggregating news on the company, you could also find out which company has the most personnel fluctuation in high ranking positions. This might be a good indicator of trouble ahead. 

Considering how much valuable information can be deduced from just properly analyzing one simple type of fact, it is obvious that many more applications become possible when we add the analysis of opinions.

Use Case: Stock Price Analysis

The graph has been created by automatically analyzing predictions on stock price developments for one company by several well-known stock analysts.

In the same way we used the "employed"-predicate to structure news about personnel changes, we have applied local grammars that turn analysts’ ratings into a structured table.

This information can be used in several ways to make predictions about a stock. One can, for example:

  • average analysts’ target estimations,
  • pay particular attention to predictions that lie outside one standard derivation from the average or
  • get alerted when the variance in the analysts’ prediction increases – hinting at an event that is interpreted very differently by different analysts.

The Tables Behind the Text

These few simple examples only hint at how much can be gained from transforming unstructured text information into structured data.

By finding the "tables behind the text", information extraction based on local grammars can make the life of people who need to consume large amounts of information in their daily work much easier and much more efficient.

The same technology has helped us improve scientific research, health services, efficacy when dealing with failures in industry components and many other subjects.

Best of Breed in Aided-intelligence for Finance

General support tool is to summarize, reduce or categorize the input qualitative data, rather than produce any analysis directly related to the actions that will be taken by the operators as a result of their decision-making process. Therefore, the emphasis is to provide the financial operator with information which is a summary of the most important data, rather than actually trying to suggest the action to be taken. The trend is captured and identified, but the interpretation of the final information and the decision is left to the financial operator. Similarly, ayfie can tell the operator that the main underlying event of a particular group of news is, for example, about a probable increase in the inflation rate, but the interpretation of the results is left to the operator.

In this use, ayfie is there to help financial operators to overcome the actual qualitative data overload, simplifying and reducing the amount of qualitative information that is needed to support their decision-making process.

ayfie's extraction system can infer knowledge from the source data (articles) which is not directly available within that data. So, for example, a text can contain information regarding a takeover, even if the word "takeover" or "acquisition" has never been cited. This analysis can be used by the financial operator for explanatory tasks such as analysis of price behavior related to qualitative information, which normally has to be performed by hand. 

ayfie identifies the number and relevance of each time a particular news item has been cited and reports this information to the user using a semantic comparison. Further insight into how the information has been presented is provided as context.

ayfie can potentially be used as prediction / forecasting tool, suggesting to the financial operator final decisions (buy/sell decisions). Operators often base their decisions on (real-time) news, rather than on quantitative predictions. In this view, the decision-making process of the financial operators can be described in the following steps:

  1. the operator reads the news;
  2. the operator gives his/her own interpretation of the new information;
  3. the operator compares and analyses the new information with the knowledge he/she already owns;
  4. a final decision (buy/sell) is taken.

The system would process the new data, identify the relevant information, process it according to specific domain-knowledge for that situation and present it to the operator with the summary / template of the original article and the suggested decision to be taken.

Leveraging Machine Learning, Linguistic Analysis and years of experience solving complicated problems.

Contact us