Natural language processing (NLP) is a technology in the artificial intelligence (AI) sector that uses machine learning (ML) and similar advancements to analyze speech and language.
Companies can use NLP to rapidly analyze huge amounts of data and extract insights that were previously too time consuming and labor-intensive to feasibly get and apply.
See the five case studies below to learn how various companies used NLP and what kinds of results they achieved:
1. Hearst Newspapers
Hearst Newspapers is one of the world’s largest mass media publishers. The company oversees an average of 3,000 new articles per day, and team members faced increasing challenges when sorting, labeling, and categorizing all the new incoming content. The employees had such heavy workloads that they often had to prioritize certain articles while leaving others uncategorized to avoid falling behind.
The company pursued improvements by using the Google Cloud Natural Language API to build machine learning (ML) models that understand an article’s topic regardless of the content’s structure. Although Hearst Newspapers previously used an older system to try and automate content classification, decision-makers found this newer solution was superior.
“Google Cloud Natural Language API is unmatched in its accuracy for content classification,” says Naveed Ahmad, senior director of data, Hearst Newspapers.
“At Hearst Newspapers, we publish several thousand articles a day across more than 30 properties. With [NLP], we can quickly gain insight into what content is being published and how it resonates with our audiences.”
Hearst Newspapers can now sort newspaper articles into more than 700 predefined categories. It also identifies specifics, such as the locations and people mentioned in articles.
The granular data provided allows the Hearst Newspapers team to better target narrower customer groups when serving them advertising. This process enhancement means the company now has a much broader content inventory when matching different media types with certain ads.
Previously, Hearst Newspapers reserved only a small portion of its content for enhanced ad targeting. Now, all of its media gets automatically classified and can be used to get better targeting results.
NLP product: Google Cloud Natural Language API
Outcomes: Company representatives experienced better speed and accuracy during content creation, leading to decreased manual labor and valuable takeaways about how people consume media.
2. Lingmo International
Lingmo International is a language-translation company specializing in offering products ranging from wearable devices to smartphone apps. Danny May, the company’s CEO, founded the company after having his passport stolen in China and trying to use a translation app to communicate with the authorities. The obstacles he encountered showed him there had to be a better way.
The company May built provides context-based language translations for people who need to communicate quickly and successfully while having limited knowledge of a new language. Lingmo International must regularly retrain the models associated with its products to ensure continued accuracy.
However, that endeavor took up to 10 weeks, and decision-makers wondered if NLP might accelerate the task. So they chose several IBM Watson products to help with various stages of the training.
“We looked at technology from a number of vendors who were able to achieve impressive results in lab conditions,” May says.
“But in real-world tests, once background noise and other difficulties were introduced, they proved to be substantially less accurate than IBM Watson. Watson enables us to train our models on text instead of requiring multiple voice samples of every word or phrase. That dramatically reduces the amount of data we need to feed in and accelerates our training cycles by up to 50%.”
The NLP technologies chosen by Lingmo International are a key component of an earpiece the company offers. It can translate between nine different languages and give wearers results that are 85% accurate within an average of five seconds. The company now finishes its training cycles in half the time, and company leaders have seen an accuracy boost.
“With Watson, we’re often seeing 90 or 95% accuracy in practice, and we’re picking up dialect words that only locals know,” May says. “The feedback we’re getting from users is fantastic, because they can see that it’s a solution that doesn’t just work under lab conditions — it’s practical and useful in real-world situations.
“With Watson, we’re also able to take advantage of new data sources to make our translation services even more capable. For example, we worked with IBM to teach our models how to watch movies, which gives us a whole new level of insight into cultural word usage and trends.”
Industry: Language translation
NLP products: Watson Speech to Text, Watson Language Translator, Watson Text to Speech, and IBM Watson Natural Language Classifier
Outcomes: The company accelerated its AI model training, improved accuracy, and enabled the translation of languages in near real-time, despite nuances caused by dialects and regional slang.
Experts assert that AI can help companies disrupt the marketplace and use innovation to become more competitive. Most decision-makers know outstanding customer experiences are necessary for marketplace resilience and growth. That was the case for leaders at schuh, a Scottish shoe retailer well-known in the U.K. and Ireland that now operates around the world.
Executives hoped natural language processing could improve the speed and quality of customer support. Blair Milligan, head of systems development at the company, says it’s not just about how many customers a brand has, but how those people feel during their respective engagements.
“It’s not about the numbers — it’s about finding an emotional solution, Milligan says. “It’s about the brand. It’s about the quality of your customer interactions, and that’s really hard to get right.”
The company chose several solutions from AWS, with Amazon Comprehend and Amazon Transcribe being the two based in NLP. The schuh customer support team uses Amazon Comprehend to analyze customer emails and the associated sentiments. It can automatically assess the emails in the support team’s queue and tell which ones contain positive or negative emotions, before customer support specialists start their shifts.
The support tickets get sorted and color-coded by issue, then sent to the representative best equipped to deal with the customer’s concern.
“Using Comprehend to put a customer problem in front of the right person really gives us the best chance of retaining that customer going forward,” Milligan says.
Comprehend also detects sentiment in outgoing messages sent by representatives. If it identifies negativity, managers get alerts to check for quality assurance.
Regarding Amazon Transcribe, schuh managers want to do real-time analysis of in-progress calls, allowing them to intervene when necessary. Representatives reportedly feel more confident when resolving customer needs since the company started using NLP technology.
NLP products: Amazon Comprehend and Amazon Transcribe
Outcomes: The company achieved more efficient customer service and higher satisfaction rates as well as better staff productivity.
4. Vodafone South Africa
Leaders at Vodafone South Africa wanted to take the company through a digital transformation that would respond to changing customer needs and reflect the brand’s expanding assortment of product offerings.
They used NLP products from Microsoft to improve the performance of a chatbot named TOBi.
Vodafone customers ask questions ranging from contract updates to billing questions. The NLP technology can examine customer queries and extract the most relevant information while understanding a conversation’s overall meaning.
“It was clear that Microsoft was a great fit for what we were looking for,” says Kevin Knowles, leader of Vodafone’s Care Transformation Program.
“We needed a flexible, scalable natural language understanding solution that could match our growth ambitions for TOBi. It was essential that tools would be easy to use for the conversational designers in our business, built upon a modular technology to support our ambition to make TOBi the biggest chatbot in the world.”
Executives wanted the chatbot to respond to queries across multiple channels, including the telephone. So they used speech-to-text and text-to-speech capabilities within Azure Cognitive Services to make the bot sound realistic when responding to customers’ voice-based queries. It was also possible to make adjustments that reflected different emotions.
Paul Jacobs, the group head of operations transformation at Vodafone, says NLP allowed the company to move away from dry chatbot responses and make the conversations interesting.
“We used the AI and natural language processing capabilities in Azure to give TOBi a clear personality that could make conversations natural and fun, which drives better engagement,” Knowles says. TOBi now handles 60% of our customer interactions.”
Vodafone South Africa’s future NLP aspirations include making the technology multilingual and training it to differentiate between similar requests. Leaders believe such progress will further strengthen the company’s market position and make it better equipped to serve a diverse customer base.
NLP products: Azure Cognitive Services and Azure Language Understanding API
Outcomes: Customers received 24/7 access to assistance through an intelligent chatbot that offered the same sales upgrades as human representatives.
A major part of successful drug development involves understanding all the current literature about any relevant topics.
That’s why the team at pharmaceutical company Sanofi used an NLP platform to find useful insights within a tremendous body of existing research while learning more about biomarkers for multiple sclerosis (MS).
“The goal of the project was to identify new biomarkers for MS by exploring the association of HLA alleles and haplotypes with diseases and drug sensitivity,” says Dongyu Liu, associate director of translational sciences at Sanofi.
“We used NLP technology to extract information from unstructured text from millions of pieces of scientific literature.”
This approach uncovered 33 new links between diseases and drug sensitivities not yet mentioned in academic papers, in addition to the 22 already described in such work. That larger body of information could improve scientific outcomes as researchers look for better ways to manage and perhaps even cure MS.
Liu says that Sanofi’s use of NLP goes far beyond the MS biomarker project.
“We have been using the NLP AI (specifically the Linguamatics I2E platform) for disease mapping, mutation analysis, targeted identification and prioritization, and biomarker discovery,” Liu says. “My colleagues also applied NLP to the later part of drug development for such things as pharmacovigilance, opportunity scouting, and competitive intelligence.
Liu says that such applications of NLP are all about enhancing results.
“We want to understand how drugs work outside the clinical trial environment and demonstrate value so we can improve outcomes,” Liu says. “By applying NLP AI to all these sources, which are mostly stored as unstructured data, we can extract information and transform the data from an unstructured to structured format.
“We can then apply other machine learning or AI operations to get even more value from the information.”
NLP product: Linguamatics I2E
Outcomes: Advancing drug discovery and research via a multiple sclerosis biomarker project that created a searchable database accessible to research teams.
How Will You Use NLP?
These case studies show what’s possible when corporate leaders become committed to seeing what NLP could do for them. Use these examples to get inspiration about how natural language processing could improve how you do business and assist customers.