Answer Extraction by Recursive Parse Tree Descent

We develop a recursive neural network (RNN) to extract answers to arbitrary natural language questions from supporting sentences, by training on a crowdsourced data set (to be released upon presentation). The RNN defines feature representations at every node of the parse trees of questions and supporting sentences, when applied recursively, starting with token vectors from a neural probabilistic language model. In contrast to prior work, we fix neither the types of the questions nor the forms of the answers; the system classifies tokens to match a substring chosen by the question’s author. Our classifier decides to follow each parse tree node of a support sentence or not, by classifying its RNN embedding together with those of its siblings and the root node of the question, until reaching the tokens it selects as the answer. A novel co-training task for the RNN, on subtree recognition, boosts performance, along with a scheme to consistently handle words that are not well-represented in the language model. On our data set, we surpass an open source system epitomizing a classic “pattern bootstrapping” approach to question answering.