Conversational AI on platforms: opportunities and challenges

On this platform, Carlos Muñoz-Romero and Nieves Ábalos Serrano, co-founders and CEO and CPO of Monoceros Labs, respectively, shed light on the arrival of conversational AI to video-on-demand platforms, opening doors to content discovery and improving user experience.

The conversational artificial intelligence (AI) is transforming the way platforms and companies interact with their users. Open doors to offer new experiences, more personalized and intuitive, and that help to get the most out of the services or content that companies and platforms, such as those of video on demand (VOD), they offer.

This area of Conversational AI is a umbrella which encompasses different technologies based on natural language processing algorithms (NLP), machine learning (ML) y deep learning (DL). They also include the large language models (LLM), well known for ChatGPT from OpenAI, Gemini from Google and open-source Llama of Goal.

The objective of this area is to maintain a conversation with user, performing the following steps in each turn: recognize voice of the user and/or understand what they say in natural language, decide what action carried out based on the information available to satisfy the user's request, and generate a response in text and voice through a digital or synthetic voice.

In this way, these technologies allow platforms to interact with their users and converse in natural language through voice or text, to offer personalized recommendations, offer customer service, and provide information about content, in a more natural, intuitive and attractive way. If we focus on the personalized content recommendationHowever, although the opportunities are enormous, there are also significant challenges that must be addressed to maximize their potential. Next, we will explore some of the opportunities and challenges that presents conversational AI in content recommendation.

Opportunities

1. Advanced customization

One of the main advantages of conversational AI is its ability to offer highly personalized recommendations. By analyzing the interactions between the user and the platform, and the user's preferences, recommender systems can adjust their real time suggestions. For example, a chatbot can understand a user's preferences based on their responses and past behaviors, offering content recommendations that align with their specific interests based on context, which includes the time of day or the type of content consumed in the last few days. This ability to dynamic adaptation significantly improves user satisfaction, improving the user experience and therefore recurrence on the platform.

2. Natural and efficient interaction

Conversational interfaces allow more natural interaction, whether writing or speaking, and in many situations and contexts, more efficient with users. Instead of navigating through menus and lists, users can simply express your needs or interests in natural language. This conversational experience, if multimodal, that is, it is accompanied by displaying the content that is being talked about on the screen, further improves the experience that the user has on the content platform. This not only simplifies the content search and discovery process, but can also reduce friction and improve accessibility, especially for those less familiar with technology.

3. Improved user retention

Personalization and natural interaction contribute to a more positive user experience, which can result in higher user retention. Users are more likely to stay on a platform that understands your preferences and provides relevant recommendations efficiently. On platforms with a lot of content, the time it takes for the user to choose it increases. Spend half an hour choosing what to see if you only have one hour does not compensate, and it is one of the problems of platforms of this type. This is particularly crucial in a competitive environment where user retention is critical to the long-term success of any content platform.

Straight

1. Accuracy and relevance of recommendations

One of the most significant challenges is to ensure that the recommendations are accurate and relevant. Conversational AI systems must be able to correctly interpret user queries and provide suggestions that truly meet their expectations. This requires, in traditional conversational systems, ML and NLP algorithms adapted to the user's language and way of speaking, and large amounts of training data, which can be complex and expensive to develop and maintain.

Also, it is key to understand the content to make more precise recommendations. In traditional systems, it is necessary that the content of the platform be well managed and labeled manually or automatically, or that the best type of recommendation system be used according to the available data and the user experience that you want to offer. In more advanced systems, the capabilities of recent multimodal language models (LMM) In order to better understand and describe the audio, video and music content without the need to label it, but it is still early to determine the technical and economic feasibility of the application of these systems on a large scale.

Furthermore, if we are talking about conversational systems that use language models or LLMs, it will be necessary to work on ensuring that the responses are truthful adding other techniques that allow us to give reliability to the response based on the content of the platform, such as with the augmented generation technique by information retrieval (RAG).

2. User expectations management

Another challenge is managing user expectations. While conversational interfaces can improve the user experience, they can also generate frustration if they do not meet expectations. The available technologies have limitations in noise situations, languages not understood, accents, y efficiency in understanding the intention and information provided by the user during the conversation. Many of these limitations are resolved with a good conversational design, and with maintenance after launch, but it is still crucial that platforms set realistic expectations and ensure that their conversational AI systems are capable of handling a wide variety of queries and situations.

3. Bias in data and fairness in recommendations

AI algorithms can perpetuate Existing biases in training data, which can lead to answers and recommendations unfair or discriminatory. It is essential that AI developers implement mechanisms to detect and mitigate these biases, ensuring that responses and recommendations are fair and equitable for all users. For example, the algorithms used to recognize user requests may have been trained with little diversity of speech, accents and dialects, and may not understand many user groups well.

Conversation will be the key to success

The Conversational AI offers new user experiences in an environment in which the content is key. Through conversation, we can improve the user experience of content recommendation on platforms, through personalization and interaction in natural language. A better consumer experience of content on the platform will lead to improved user retention.

However, it also presents challenges, some more complex than others, associated with technological limitations and user expectations that must be taken into account to guarantee its success. Challenges that, although complex to face, will become increasingly common as technology advances. turning into opportunities.

And in the coming years, thanks to the rapid evolution of these conversational AI systems, the way that users will have to interact with content will be more natural and personalized, and this will be key to the success of content platforms.