Joint Learning of Words and Meaning Representations for Open-Text Semantic Parsing

摘要: Open-text semantic parsers are designed to interpret any statement in natural language by inferring a corresponding meaning representation (MR – formal of its sense). Unfortunately, large scale systems cannot be easily machine-learned due lack directly supervised data. We propose method that learns assign MRs wide range text (using dictionary more than 70,000 words mapped 40,000 entities) thanks training scheme combines learning from knowledge bases (e.g. WordNet) with raw text. The model jointly representations words, entities and via multi-task process operating on these diverse sources Hence, the system ends up providing methods for acquisition wordsense disambiguation within context parsing single elegant framework. Experiments various tasks indicate promise approach.