- Education And Career
- Companies & Markets
- Gadgets & Technology
- After Hours
- Banking & Finance
- Energy & Infra
- Case Study
- Web Exclusive
- Property Review
- Digital India
- Work Life Balance
- Test category by sumit
Natural Language Processing Changing The Phase Of Retail In 2019
To being a cheap alternative to text recognition, and for delivering information in a cost-effective manner, NLP in the near future would be dynamic customer service centers to reckon with
Photo Credit : Shutterstock
Increasing adoption and awareness has been causing conversations on how Natural Language Processing (NLP) can equally benefit retailers and customers in the shopping process. Omnichannel retail gets a new depth with voice-integration, and opens up new opportunities when combined with VR and IR-led shopping experiences in brick-and-mortar stores, and through better contextual product search for the customer in the online space.
NLP is the next frontier of e-commerce, as increased adoption worldwide has precipitated into greater machine learning capabilities, driving further adoption, triggering a virtuous cycle. In any stream of business, a new technological breakthrough takes its own time to cross over from being a hype to being adopted by the masses. In retail, which is essentially seen as a humane activity involving multiple touchpoints across online and offline channels, language input has opened up fresh avenues of growth, and has also reduced several complications for the customer.
In any business process, screens that involve complex data inputs are a surefire formula to drive away customers. In such instances, voice cuts through the clutter and instantly enables customer delight. Of all human sensory inputs, voice is considered the most effortless and stress-free from the customers’ point-of-view. Artificial intelligence in voice processing has made it easy to sift through large repositories of data, to execute business actions.
Thanks to years of feeding the machine, the machine has begun to speak more like humans than we imagined. The latest in speech recognition has crossed 90 percent of that of human, and with its sheer consistency and scalability, AI holds an edge over real executives who handle data. With hardware improvements such as multi-directional voice capture, far-field recognition and speaker separation from sources, voice processing has reached new quality levels.
In the Indian context, barring language localization, which is still in the works, English with an Indian accent has attained equal capabilities as that of the US. When regional language processing would cross the chasm, the retail sector in India can unlock the huge potential of multi-lingual live translation, analysis and interpretations to aid the store managers. Further, emerging technologies would even have effortless ‘code-switching’, which switches between native language and English, as is used in India in a mixed fashion during speech. Significant research is now underway in terms of vernacularization of natural language processing.
Context over content: What can NLP do?
So far, the only grouse about inadequate natural language processing has been the lack of context awareness. Particularly, the retail sector faces even bigger challenges than organized, clear-cut, predictable, robust, process-driven businesses such as banking and finance, where every step can be translated into algorithms. In retail, there are no such defined protocols, which is making it a tougher challenge to enable complete replacement of store-managers. When AI cannot replace, it can definitely augment, and provide excellent context-relevant information in a live, interactive manner to serve the customer in the best way possible.
When it comes to application of NLP, there are two approaches, namely the speech-based token extraction, and the phoneme-based knowledge-agnostic approach. Speech-based state-of-the-art neural networks such as Google KWS have gotten better after being trained with keywords with a large number of utterances. While adding a new keyword in this model requires re-training the whole network, the takeaway is that the system has achieved an accuracy in the order of 94 percent.
On the other hand, knowledge-agnostic approach combines the power of neural networks and classical machine learning. It uses distance-based metrics to compare the similarities between keywords and spoken sentences to spot a keyword. The networks that are trained on large speech data to extract acoustic information, can be language independent. Further, adding a new keyword doesn’t require retraining the whole model. In this method, an accuracy of around 70 percent has been achieved with some basic modeling effort with a lot of scope for improvement with data and augmentation.
So, how does the program know a speaker’s identity? In this context, there are further two systems of NLP for user-level identification: Text dependent, and text-independent. While the former that identifies a user from transcripts owing to its text-intensive nature, the latter recognizes the user for every word he/she speaks, returning a better accuracy than the former and relies on a lot of pattern recognition of speech signals.
NLP is becoming mainstream, faster than you think
Together with augmented reality / virtual reality, the retail customer can enjoy a multi-modal shopping experience, with tailor-made services based on previous recorded interactions, and history of purchases. Domino’s partnership with Amazon to use voice to place and track pizza delivery orders, hailing a cab, or programs that take quick stock decisions by analyzing the sentiment of an annual report, are many examples of real-time NLP implementation already in place.
In addition to being a cheap alternative to text recognition, and for delivering information in a cost-effective manner, NLP in the near future would be dynamic customer service centers to reckon with. In the text-heavy occupations such as the legal sector, NLP reduced a great amount of burden by helping executives process PDFs, speeches and reports, freeing up their time to do the actual work! Personalized bots that make shopping fun, faster, and better, by employing semantics and sentiment analysis would be the order of tomorrow in retail.
Overall, NLP in retail is still a work-in-progress, with surprises waiting to happen at every turn as we march ahead in this digital age. The efficacy and relevance of voice-data painstakingly collected over decades would bear bigger fruits in the near future. Also, open source research and tools are further driving the adoption of voice in retail by truly democratizing the field.
Disclaimer: The views expressed in the article above are those of the authors' and do not necessarily represent or reflect the views of this publishing house. Unless otherwise noted, the author is writing in his/her personal capacity. They are not intended and should not be thought to represent official ideas, attitudes, or policies of any agency or institution.