It might at first sight sound like a rather rarefied field of data science, but textual analytics (TA) is fast becoming a must-have tool for CIOs whose job it is to select technologies with strategic thrust.
Its value lies in the opportunity to automate an information-gathering process that was previously nearly impossible because it was time-consuming and piecemeal. TA in fact invents a new business process: it is like having a tireless analyst plough through the massive amounts of language content generated around standard business operations to harness marketing and other insights.
In contrast to many ways of generating business intelligence, textual analytics focuses on “unstructured” data. This quite literally means the life and times of text – or more generally language content.
Although human language is a slippery medium, companies supplying TA are increasingly finding that content that might at first sight appear to be “unstructured” (i.e. when compared to numerical data) can be highly structured when viewed from the emerging expertise of natural language processing.
Simple ‘positive’ and ‘negative’ assessments may be useful tags for certain types of textual information on a social media, but textual analytics is making rapid progress to more valuable insights into the meanings of human communications.
LT-Innovate sees three key touch points that will matter for suppliers of future text analytics solutions:
- Drill-down semantics: Language technology is rapidly extending its powers of analysis: while the first generation of sentiment analysis tools tended to simplify opinions into binary distinctions, new semantic tools are radically expanding the power of TA solutions to deliver finer-tuned results of what is communicated. One such field is that of “emotion recognition.” This means that TA can help identify more subtle expressions of approval, interest, concern or rejection than has been possible so far, offering a finer-grained understanding of what customers, partners and the market are saying.
- Multilingual processing. Multiple languages pose a key challenge for any TA application. Some TA products can run text analyses in up to 20 or more languages today, a remarkable feat given the subtleties of human expressive resources. But as companies start to target some of the long tail of markets and communities, the ability to decode thoughts, feelings and opinions expressed on social media in many dozens more languages may well tempt more and more businesses. Rich linguistic resources will be needed for this, but at the same time, the results will need to be translated into a single central repository for analysis. This requires highly tailored translation technology at the right price.
- Specialisation. TA is increasingly targeting a broader palette of corporate language content. In some case this will include the content of contact centre conversations, the content of internal conference calls, or the input to online surveys. Which means that TA will need to integrate seamlessly with specialised technologies such as speech recognition and audio recording to gain purchase over the text implicit in new types of media? This will almost certainly lead to specialisations of TA in terms of business tasks, type of content, or industrial and commercial sector.
This is just the beginning. In the longer run, it is very likely that the kind of linguistic analysis currently offered by the TA community will be embedded in a broader “cognitive computing” environment. For this scenario, learning systems will be programmed to range over corporate data to find patterns that suggest directions to be taken or decision to be made.
At the same time, CIOs will need to think about expanding the range of TA to other media than text – for example, video as image and as conversational content (and hence text) will also enter the content mix to be leveraged by analytics. Such developments will put huge pressures on the trust we hope to
have in the value of our analyses as they extend over more data types.
Yet at the same time, these integrated, easy-to-use solutions will need to appeal to CIOs and others who need to rapidly engage with the voice of their customers, the voice of their staff members, the voice of their service suppliers, and more generally the competitive voice of their market segments.
LT-Innovate believes that TA opens up a promising market in which highly-specialised language technology can provide effective responses to a business need that simply cannot be met by traditional solutions. The vital step is to recognise that one size will not fit all. Solution providers therefore need to know exactly what their potential customers need. A compelling case for textual analytics indeed!
This is why LT-Innovate -- the Forum for Europe's language technology industry -- and US consultancy Alta Plana Corporation, headed by industry analyst Seth Grimes, are holding a brand new event entitled LT-Accelerate. This will take place in Brussels (Belgium) on 4-5 December and is devoted to the people, players and end users of Textual Analytics.
