Training Conversational AI to Provide Customized Movie Recommendations

SOLUTION

Innodata’s experienced team was able to support all the processes needed to build a conversational AI chatbot. First the team developed a taxonomy of movie topics, types, and moods. The taxonomy also included additional attributes customers search for, such as genre, actors, duration, and content rating, which would help filter the movie recommendations the chatbot would provide. In addition, the Innodata team utilized a combination of data aggregation technology to automatically scrape data. They then built a model that automatically extracted entities, which were used to add values, such as additional descriptors for a movie, to the search intent. These entities and intents help the chatbot provide accurate recommendations. The entities were ranked to identify intents and over 30 utterances were added across each intent, which allowed the chatbot to better understand customer requests and increase accuracy. With the data collected, the training groundwork in place, and a new taxonomy developed, the team was able to annotate the available data quickly and accurately. 

IMPACT

Innodata helped build the foundation for a conversational chatbot that provides movie recommendations. Once the AI solution provider deployed the new chatbot, it gained significant traction and brought new leads to the company. It allowed the client to offer a product that would give streaming service providers a new way to engage customers and improve the overall experience.