Abstract: Information may be extracted from a text corpus. The text corpus may be parsed into a parse tree structure based on the parts of speech of the words of the text corpus. A path in the parse tree structure may be identified as linking an entity and a value, and the path may be applied to the same or other text corpuses to extract other instances of entity-value pairs. Extracted information, associated paths, or both may be validated in some instances.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for information retrieval. In one aspect, a method includes receiving a fact query; determining an expected type of answer to the fact query; identifying search results responsive to the fact query; identifying phrases from the resources corresponding to the search results that correspond to a form of the expected type; determining a score for each of the identified phrases; and presenting an answer to the fact query using the scores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for machine learning. In one aspect, a method includes receiving a collection of facts, each fact represented as an entity-attribute-value tuple; identifying expected values for one or more individual attributes, where the identifying expected values includes, for each particular attribute: identifying facts having the attribute, calculating a value score for facts of the collection of facts having the particular attribute for each particular value, calculating a global score for all facts of the collection having the attribute, and comparing the value score to the global score such that a value is identified as an expected value if the comparison satisfies a specified threshold.

Abstract: A system and method for predicting price fluctuations in financial markets. Our approach utilizes both market history and public news articles, published before the beginning of trading each day, to produce a set of recommended investment actions. We empirically show that these markets are surprisingly predictable, even by purely market-historical techniques. Furthermore, analyzing relevant news articles captures information features independent of the markets history, and combining the two methods significantly improves results. Capturing usable features from news articles requires some linguistic sophistication the standard naïve bag-f-words approach does not yield predictive features. Instead, we use part-of-speech tagging, dependency parsing and semantic role labeling to generate features that improve system accuracy.