In many industries, we’re seeing the growing use of chatbots to handle customer inquiries and frequently asked questions. While this is convenient for customers, the bots often lack the human touch found in natural conversation.
While everyday conversation may seem simple, it actually poses considerable challenges, including knowledge of conversational topics and the recognition of individual speech patterns. The unstructured nature of conversation is particularly difficult for machine learning.
The challenge
Our client, a digital solutions provider that develops conversational artificial intelligence (AI) systems, was looking to develop a customized AI that recreates the personality, way of speaking and preferences of a character or human being through conversational chatbot applications. The AI system, known as Personality Reverse, needed to respond not just to questions and answers, but also engage users in everyday conversation.
In order to build an AI to communicate naturally with users, our client required a data collection partner capable of delivering a wide range of text data. This text data had to be reflective of natural conversation and needed to be labeled for emotional intent.
Our holistic and customized solution
TELUS Digital was selected to create and annotate the training data based on our data collection and annotation expertise. Our flexible approach and dedication to customized project guidelines were keys to being selected. Winning factors included:
- Price: We provided a clear and reasonable pricing plan based on the content and data amount necessary for the project.
- Expertise and experience: In order to best support client data requirements, we worked closely with the client through regular meetings and email correspondence to align on project goals, clarify challenges and remove blockers. Through this process, we created a set of guidelines for the delivery of training data.
- Quality assurance: To ensure an appropriate level of quality, we first provided a sample of training phrases. We then created a detailed set of annotator guidelines to align with the client’s specific needs, and worked closely with the project manager to ensure quality at delivery.
The TELUS Digital difference
From our global AI Community of 1,000,000+ data experts, we put together a qualified team of data collectors and annotators specifically for the project.
The team collected 5,000 sets of question and answer text samples, labeling each set for emotional intent.
With this data, the client developed a personal AI system capable of conversation with their users. The client is committed to improving and developing their AI projects, and plans for future AI companions to utilize voice and facial recognition technology to identify user emotions and respond with appropriate conversational topics.