Data-driven analysis of play using SpeechBlocks

Mina Soltengheis and Ivan Sysoev

The rich, fine-grained data gathered from SpeechBlocks opens up unique opportunities to go beyond test results and look closely, progressively and longitudinally at the child’s learning process itself and use the lens of data to study what playful learning looks like.

To reveal the stories hidden in the data, we processed the log files into the diagrams called PlayTrees. PlayTrees capture essential events that happened during a session of child’s play, such as merges and splits of words, tap and drag actions and utterances pronounced by the speech synthesizer. They allow a person (a researcher or a literacy expert) to grasp the essence of a play session in one glance.

The following video shows a sample PlayTree that tells a small story from the data as the child plays with SpeechBlocks.

We use the raw data and Play Trees to understand both what children are building and how they’re building it. We first want to understand whether the words created are sensible by looking at the directionality of the text, misspellings, the context in which they were created, the pronounceability, and instances of word or letter repetition, etc.