The Brand-Tuned Language-to-User Data Classifier tool, on its own, is extremely valuable to any product marketing team, but Writesof uses it to analyze billions of words of content in a massive neural network database. By using this neural network, powered with machine learning, our NLG Personas are able to make decisions on what to write and who to write it to. In the same manner, human writers can use it to determine word choice, communication style, and grammar structures that are most likely to result in a positive, more engaging user experience. Your customer data serves as feedback for the tool to rank language by performance. do the same, making more inform editing decisions for personalized product communications.
We use a machine-learning-capable version of this tool internally to power our NLG systems, which make language choices by analyzing user engagement and conversion data. Instead of using one brand’s data set, we incorporate language from several brands, by product category. Language is parsed into overlapping chunks of text, each classified by its objective, subjective, and contextual attributes, in relation to user data. Objective linguistic attributes are standard qualitative and quantitative features of language, such as Part of Speech and Word Location. Subjective linguistic attributes are qualitative features, such as emotion and reading level, which relate to users in differing ways. Contextual attributes are linguistic attributes that connect words together, such as a message’s tone or communication objective.
These are just a few generalized examples of how Writesof defines and measures language. In fact, Writesof uses some of the most advanced Natural Language Processing (NLP) tools, many of which were internally developed specifically for product communications. NLP technologies, some of which have been refined over decades, are what enable our NLG Personas to probabilistically determine word choice, language structure, and communication style.