The research field of Natural Language Processing is developing at a rapid pace. But what applications are there for companies?
Author: Mark Bosshard
Natural Language Processing, or NLP for short, is an interdisciplinary field between computer science, linguistics and artificial intelligence (AI) and, more recently, deep learning. Since the 1950s, the domain has been concerned with the digital processing of natural language, both in spoken and textual form.
Certainly there are companies in which natural language processing can generate more value than in others. For all companies with many digital processes, the potential value of NLP technology should be evaluated. Because communication is omnipresent today and happens at a rapid pace increasingly via screens, microphones and loudspeakers. In the following we show seven modern application areas in which NLP replaces processes or makes them more efficient. Cloud providers such as AWS, Azure or Google Cloud often help with out-of-the-box offers to get started quickly and gain initial experience with little effort.
Google Translate or DeepL , the younger star among translation machines, are well-known products for translating text into other languages. Meanwhile, clouds such as Azure, AWS or the Google Cloud ML-as-a-Service also offer enterprise services. These allow documents to be completely translated in one call with a single line of code. This can be used for everything from translating a foreign-language customer letter for processing by support staff to a 'pre-translation' for professional translators and greatly accelerates the translation process.
Requesting support from a company via Facebook Messenger or Whatsapp, just like you would contact a colleague, is becoming an everyday occurrence. Yesterday, during office hours, we were happy to listen to music on hold in a hotline queue and then be able to choose from a variety of inappropriate options by pressing a button. Chat bots not only make any company look modern, but can also, for example, combined with live chat, mean great added value for customers and a reduction in the workload of the support team. Voice assistants are also already used regularly on smartphones by one in five Swiss people. A new platform in which your services can be used in an innovative way to reach the smart speaker in the living room. In-house training bots are increasingly used, which provide employees with a "learning nugget" as a short conversation every day. Gartner expects that by 2021, one in four employees in the digital job market will be interacting with a virtual assistant every day.
How do you process hundreds of customer feedbacks that come in every day? Or can incoming e-mails from critical customers be prioritized to avoid jumping to a competitor at all costs? The sentiment analysis makes it possible to recognize whether a customer is satisfied or dissatisfied with his or her writing, even with a large amount of incoming text. This method, which is particularly valuable for sales and marketing, can be used either as an in-house development or via ready-to-use ML-as-a-Service services from cloud providers.
Do all your employees write external documents correctly in a foreign language? Or could the spelling or language style be optimized? Even the latter has been made possible by modern Spell-Checking algorithms. Usable for both internal and external documents, Spell-Checkers make embarrassing typos a thing of the past. By the way: Grammarly is a browser extension that can be used by anyone to review your text and even suggest stylistic improvements, no matter what website you are typing on.
Who hasn't used a search field in a tool that "doesn't work" - or just produces hits only with 1:1 matches. The optimization of document structures with tags and labels, but also a better search algorithm can help here. With Azure Cognitive Search, Google Cloud Search or Amazon Comprehend, all three ipt cloud partners offer exciting and mature solution modules. We would be happy to advise you individually on how you can upgrade your search to the latest state of the art.
How does an inquiry get to the right employee as quickly as possible? Document classification can help to forward incoming e-mails to the right employee without the need for manual triage. This saves work and creates speed for external inquiries to your company. In a Machine Learning (ML) project, keyword lists, classical ML algorithms but also deep learning approaches can be used to create the optimal triage algorithm for your inbox.
What is talked about in an email? Can software generate meaning from individual elements of the text? Filtering out a product or event addressed in the text and identifying it as an entity in the text is part of the research topic Natural Language Understanding (NLU). Many researchers consider the actual processing of text to be largely solved. NLU is the next big challenge. However, the current possibilities of NLU are still limited. The goal is to use information - like a described event in a customer feedback - aggregated to arrive at a basis for decision-making.
With digitization, much of what used to be handwritten on analog documents has become more accessible. But something has remained: People prefer to reproduce thoughts in a prose text rather than fill out their concerns in a prescribed form. AI-based NLP allows unstructured information to make sense quickly and to bring hidden potential to light, be it to extract hidden information from a large amount of text, to answer a question with appropriate answers, to translate text or to group it.
We would be happy to advise you on how you can digitize and simplify voice-supported processes through in-house developments or through the numerous ready-to-use cloud services available.