In some blogs like Big Data, we have already explained for what purposes large amounts of data can be used. There are a lot of social networks and news portals today, and you have to screen them all at the same time in order not to miss anything. Especially now, when content marketing plays a major role more often, the early recognition of trend topics and developments hold great importance.
The different portals serve different purposes and are used accordingly: On Twitter there is a short-term, fast flow of information. On Facebook, however, the texts are longer, more detailed and can even "survive" a few days in the news feed. So how nice would it be to extract all the important trends on the Internet with the help of a tool to see them at a glance?
Follow trends without being an insider
The best solution is called SigniTrend. SigniTrend stands for "Significant Trend" and makes it possible for the first time to recognise trends and events in the shortest possible time in posts, news and comments and to present them in a bundled form at a glance (see Figure 1)
SigniTrend shows a kind of mind map of the most important keywords that circulate on the Internet. The larger these keywords are, the higher their significance, i.e. frequency measured by the general probability of occurrence per word. The most frequent keywords are shown for each news or trend. For the user only the interpretation remains, which makes it very easy to view the current trends.
What is the idea behind "SigniTrend"?
SigniTrend was developed in 2014 at the LMU Munich and further developed at the University of Heidelberg. It is part of the growing literature on the subject of event detection, which is becoming increasingly important with the advance of social media. SigniTrend uses an estimate of statistics for words and word pairs. Through clever interaction of special data structures, it is possible to track an open and previously unknown vocabulary and even all word pairs that occur together.
SigniTrend detects when a term occurs more frequently than statistically expected. Ubiquitous words and long-running topics are thus automatically ignored. Terms that correspond to new trends and events, on the other hand, appear without being specifically searched for. By visualizing very intuitively, by making each term appear larger the further its frequency exceeds the expectation for it, the interpretation is very easy. Connecting lines in different stroke widths indicate when words also occur significantly frequently as a pair.
In the following graphic the news in the period from 02.07 - 09.07.2018 are analysed for the search terms "cave", "Thailand", "Wild Boar" + "Tham Luan".
July 6th: Navy-SEAL diver dies in rescue attempt; Elon Musk sends engineers to Thailand
July 7th: Hundreds of rescue shafts are drilled; children send letters to their families; THW wants to send German pumps to the cave
July 8th: The first children are gradually being rescued; #ThaiCaveRescue; Comparisons in social media of drowning refugees in the Mediterranean
SigniTrend's possibilities and basic functions
Finally it is possible to recognize trends automatically within a short time. The countless hashtags that are often used in Twitter are understood more quickly with SigniTrend. The effort of searching through the individual hashtags and reading through tons of tweets with the respective hashtag is now a thing of the past. Of course, the hashtags of Facebook and Instagram are also analyzed. So a company or user does not have to be active on all social networks to follow all news and trends. Another special feature is that this tool can be integrated into a company's existing dashboard if desired. This provides an overview of the company's key figures and the most important trends relating to the specified brand.
Another feature that makes SigniTrend so practical is the specific search for certain words, brands or products. This means that not only general trends on the Internet are extracted, but also specific opinions and trends related to a keyword. The user doesn't have to search for it himself. After defining the search term, the output shows a "newsfeed" for the keyword.
How you can use SigniTrend for your company
SigniTrend is mainly used for trend recognition in relation to companies and their products and their brand. It shows what others think about your brand or product and creates the opportunity for better recommendations for action.
A similar approach was taken by Press Spokesman (Magazin für Kommunikation) . There it was analysed in which years people talked about which animal. With this knowledge, they gave better recommendations for action to the toy industry, for example. Another very important aspect is the recognition of possible shitstorms by following his own campaign or hashtag. Low-threshold trends that affect their product can be found much faster. A food manufacturer thus recognizes the latest "food trend" before the others and develops suitable products faster.
Since you don't have to search for a specific topic forever, you save a lot of time. And time is money, as we all know. In addition, companies save expensive online surveys, which usually do not provide reliable information anyway. The time-consuming preparation of such surveys is no longer necessary and all the needs of the customer group are covered. SigniTrend is very flexible. Depending on requirements, functions are added or modified to meet special needs.
Conclusion and summary
SigniTrend is the recognition of emerging topics (trends) in text streams such as news items, mails or tweets. It analyzes whether words occur significantly more frequently in a given period of time than would normally be expected. If this is the case, a trend develops. SigniTrend captures exactly these words and displays them graphically.
ADVANTAGES OF SIGNITREND AT A GLANCE:
- Processing of large amounts of data and reduction to the central content topics
- Discovery of "red threads" in content
- Recognizing developments
- Application for different domains (e.g. communication, news and social media)
- Enormous time saving
We would be happy to provide you with further useful information on how to learn more about your brand messages.
Schubert, Erich, Michael Weiler, and Hans-Peter Kriegel. SigniTrend: Scalable Detection of Emerging Topics in Textual Streams by Hashed Significance Thresholds. Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2014, doi:10.1145/2623330.2623740