We stand with Ukraine

Text mining applications for any sector

Expert.ai Team - 24 May 2016

Not only are organizations and individuals drowning in information, they also  cannot keep up its exponential growth. Text mining applications can be used to automate some of the most information-intensive business processes. In this post, we will explore some of the most effective text mining applications and how they can be used to automate processes in different sectors:

Manufacturing: Warranties

Analyzing repairs and maintenance data in the durable goods sector is a difficult, widespread problem that is also quite expensive. Analysis is made to not only identify individual components that may fail in normal use, but also to rapidly identify the causes of system-level failures due to unexpected interactions between system components (rather than just component-level issues). A text mining application can address analysis problems by accessing repair datasets generated by technicians who perform repairs and obtaining data from different sources such as the text narratives on repair history stored in different repositories.

A text mining application can also help control warrranty costs through analysis of unstructured information, which is often avoided due to the inherent complexity of the tasks required and the “messy” text data.

Business Value: Text mining applications mine data and unstructured information to extract the knowledge that organizations need to identify critical business issues. Integrating this information with the traditional data available for warranty costs analysis can help organizations more effectively address core business issues. For example, data extracted from warranty reports by a text mining application can help identify root causes of failures or problems by identifying the link described in the warranty report text.

Text mining applicatins in all sectors: Measuring the impact of marketing campaigns

With the growing influence of social media, it is now possible to monitor the market’s reaction to marketing campaigns before it impacts revenue. Using text mining applications, data can extracted from social media content and organizations can select and cluster the comments made on a campaign or advertisement, identifying variables such as “buzz” (how much people talk about the campaign), “sentiment” (how people speak about the campaign), “attributes” (what concepts are associated with the campaign and how they differ from the target message), etc.

Values attributed to these variables, made possible with text mining, can help organizations anticipate the impact of the campaign, suggest corrections or set new targets for results in real time.

Business Value: text mining application can be used to measure real-time social media reactions to a campaign can help an organization modify their message or if necessary, take correct actions before it’s too late. The resulting real-time intelligence supports better decision making for your marketing budget investment.

Media: Increased monetization

The growing crisis of traditional media suggests that publishers, if they want to survive, will have to find new methods and new business models to increase their online business revenue. Text mining applications offer several opportunities for maximize revenue from a publisher’s information assets.

Content optimization: Publishers are looking at text mining applications as a great opportunity for optimizing the revenue streams from a given set of content. For example, data collected about users’ clicking habits based on the position of an article on the home page provide important input to the publisher’s news room to decide whether a specific article is performing as expected (in terms of clicks), given its position on the page. In optimization terms, it is a typical problem of maximizing the revenue (directly linked to the clicks) given a relatively fixed amount of assets (the articles published on a given day).

Text mining applications that employsemantic understanding provide important additional data by analyzing the topics and feelings expressed in an article, which offers an additional dimension to the analysis. For example, the click through rate of a publication’s top left article that average 35% could dramatically increase or decrease  based on whether the article is about “good news” or if the news is related to local events.

By extracting data about the content, the semantic technology at the core of these text mining applications can further improve the publisher’s ability to optimize revenue.

Content Recommendation Engine: Another way publishers can maximize their revenue is to ensure that readers spend more time on their website. This can be accomplished through Content Recommendation Engines, which, thanks to text mining applications, create a list of targeted suggested articles to keep readers on the site.

Building and adjusting the output of your text mining application to ensure that it provides highly relevant suggestions, also for specialized domains, is a challenge for even the most advanced applications.

To achieve this, text mining applications require flexible capabilities to identify, for example, the most relevant concepts in an article, other relevant entities that can create dynamic links to other articles, as well as a convenient workspace that enables users to perform all the tasks required and adjust the recommendation strategy, all from the same interface.

Finance & Insurance: Fraud Detection

Text mining applications, integrated with optimization algorithms, can help insurance companies identify suspicious claims, including who is making the claim and when and how the claim is made, with the objective of identifying similarities with other seemingly unrelated cases.

Often, only a few degrees of separation exist between the participants, and text mining applications, by providing social network monitoring, can be helpful in connecting the dots. For example, take the following information—individuals with similar names who frequent certain car shops, combined with personal information such as phone numbers, email addresses and their role in the claim—and extend this to all claims filed.

Social network analysis made through advanced text mining applications helps uncover these previously unseen links; the integration of this data into additional resources—fraud scoring engines that already include traditional elements such as database searches, predictive modeling, network link analysis—can be used to determine the likelihood that a claim is fraudulent and help the organization prioritize their efforts accordingly.

Business Value: By using both structured and accurate unstructured data, plus effective optimization algorithms, companies can determine the likelihood that a claim is fraudulent, prioritize their efforts and significantly reduce their claims expenses.