Menu
Published artificial intelligence blog on
Machine learning consists of processing and analyzing a huge volume of data in order to make predictions about behaviors, especially for use cases related to sales and web. A few weeks ago I became interested in a free online course offered by Amazon Web Services. I was curious to learn more about a concept that everyone is talking about lately, especially when talking about Artificial Intelligence and digital transformation: machine learning. From the introductory video of the course I was able to arrive at this definition of machine learning :
Machine learning consists of processing and analyzing a huge volume of data in order to make predictions about behaviors, especially for use cases related to sales and the net.
In fact, this definition fully matches an Amazon practice, this time in the role of a global internet product sales platform, that we have all experienced: below the item that we are about to buy, the section introduced by the phrase 'Customers who viewed this product also viewed', which provides other items of possible interest.
The introductory phrase of this section perfectly reflects the procedure carried out: Amazon collects millions of behaviors of its users regarding purchasing processes, it analyzes them in order to predict the future behaviors of its customers, and with high probability manages to increase sales on the internet. No cabe duda de que esta metodología resulta ser un arma extremadamente potente a la hora de hacer marketing dirigido, previsiones de demanda por supuesto convertir las visitas a la internet en ventas. ¿Pero qué pasa si se aplica la misma estrategia para dar soporte al cliente?
En el contexto de un canal de atención al cliente, como puede ser un chatbot, es importante pararnos a pensar en el concepto de predicciones, que resulta poco compatible con la exactitud la precisión que requiere el mundo del soporte. Si bien una predicción de venta recomendada no conlleva riesgos mayores – si se convierte la venta, Not bad, si no se convierte, también -, las respuestas que se dan a los usuarios por un canal de soporte oficial de la compañía pueden ser altamente sensibles: they need to be reliable, accurate and contextualized,both authorized and commercially. The chatbot is still just an additional channel to resolve questions about plane tickets, check contractual aspects of a bank account, download appliance manuals, check data of a lodge reservation, the balance of a phone line…
The knowledge about the company provided through the chatbot must not differ from the rest of the website nor from other customer service channels, whether these channels are automated or well managed by human agents. An inappropriate response, especially regarding a delicate inquiry, could have dire consequences. Let's imagine an insured service chatbot, on the website of an insurance company, que a una consulta sobre mama pecho deriva al usuario a respuestas sobre cirugía plástica, por ser más habituales según el machine studying, en vez de ofrecer una más respuestas sobre patologías, ginecología, and many others. La imagen de este aseguradora resultaría dañada debido a la poca fiabilidad del chatbot, más exactamente debido al uso erróneo de una metodología predictiva, como es el machine studying, con un propósito de atención al usuario.
In addition, para cumplir con este propósito, el requisito imprescindible del machine learning es disponer de un volumen – cuánto más amplio mejor – de datos históricos para procesar y analizar, a partir de los cuales construir modelos de comportamiento predictivos. Today, some macro-companies have these volumes of data, after spending years monitoring their support channels and processing the information manually to be able to rationalize and classify it.
However, the reality is that most companies do not have these traces of their customers' activity, therefore they cannot benefit from machine learning easily. the alternative left for them is to start creating data manually, as if they were traces collected from users, to 'train' an ad-hoc machine learning model. This training process is highly promoted by numerous chatbot providers, hasta tal punto que se ha creado la falsa creencia de que la única manera de implementar un chatbot es entrenando el sistema manualmente, durante meses, con cuantas más iteraciones mejor – siendo las iteraciones muchas maneras diferentes de expresar la misma intención.
Cultivan este mito aquellos que no cuentan con tecnología de Procesamiento del Lenguaje Pure para eludir la falta de datos previos. Also, la creación handbook de datos para encajar con modelos de machine learning implica un esfuerzo colosal para las compañías, que se traduce tanto en una dilatación del time-to-market del proyecto como en una desilusión para con la Inteligencia Artificial.
Por el contrario, A provider capable of using Pure Language allows solving the two problems posed by machine learning for a customer service chatbot: on the one hand, it does not require extensive volumes of data, on the other, it is based on linguistic engineering to provide semantically relevant responses, contextualized and highly reliable.
This does not mean that machine learning is an evil approach but rather that it is not always the most suitable approach depending on the use case. Although to boost web sales with recommendations it seems to be the most appropriate methodology – and with demonstrable success -, for something as smart as a customer service chatbot, it is likely to err in inaccuracy and lack of precision, disadvantages that end up harming the image of companies. Betting on Pure Language Processing, instead, allows guaranteeing the reliability of the chatbot, since of the company, as well as promoting an agile time-to-market and companies' trust in Artificial Intelligence.
In the face of this scenario, the ideal approach could consist of bringing together the best of both worlds: fully trust in Pure Language Processing as the main technology of the user support chatbot, and at the same time use some functionalities of machine learning to achieve certain purposes, such as analyzing customer interactions to suggest new content for the knowledge base, well, based on clickthrough statistics, adapt the interface to improve the user experience.
Related
Enter your comment here…
Enter your details by clicking on an icon to sign in:
email (necessary) (The address will not be made public)
Name (necessary)
Receive new entries by e-mail.
Sections
Vallebro Communication
Entrepreneur's Blog online and offline
%d bloggers like this:
Digital platform on the technological uniqueness of Roko's Basilisk, We promote the artificial intelligence of the future.