A natural language parser is a program that works out the grammatical structure of sentences, for instance, which groups of words go together (as "phrases") and which words are the subject or object of a verb. Probabilistic parsers use knowledge of language gained from hand-parsed sentences to try to produce the most likely analysis of new sentences. These statistical parsers still make some mistakes, but commonly work rather well. Their development was one of the biggest breakthroughs in natural language processing in the 1990s. You can try out our parser online.

Google has open sourced its language parsing model, SyntaxNet, calling the English version Parsey McParseface. The system understands human language with an incredible degree of accuracy, but attention has centred around its choice of name, which comes after people voted to name a science research ship Boaty McBoatface - it was in fact named after Sir David Attenborough. The open sourcing of Google's parsing model means that the broader community can employ the tool to up the game of artificial intelligence (AI). This means that machines could understand sentences from a standard database of English language journalism, the first step in their journey to take over the world. "At Google, we spend a lot of time thinking about how computer systems can read and understand human language in order to process it in intelligent ways," explained Google senior staff research scientist, Slav Petrov.

As much as we love to fawn over artificial intelligence (AI), it's still not great at recognizing and parsing natural language. That's why Google is open sourcing its new language parsing model for English, which it calls'Parsey McParseface.' Before you even ask, the name has no meaning. When Google was trying to figure out what to call its language parsing technology, someone suggested Parsey McParseface; it's a bit like Apple's Liam, which has no clever backstory either. The overall AI model model is called SyntaxNet (please make your SkyNet jokes now); 'ol Parsey is just for English. Some of the biggest names in tech are coming to TNW Conference in Amsterdam this May.

A TensorFlow implementation of the models described in Andor et al. (2016). At Google, we spend a lot of time thinking about how computer systems can read and understand human language in order to process it in intelligent ways. We are excited to share the fruits of our research with the broader community by releasing SyntaxNet, an open-source neural network framework for TensorFlow that provides a foundation for Natural Language Understanding (NLU) systems. Our release includes all the code needed to train new SyntaxNet models on your own data, as well as Parsey McParseface, an English parser that we have trained for you, and that you can use to analyze English text. So, how accurate is Parsey McParseface?

As much as we love to fawn over artificial intelligence (AI), it's still not great at recognizing and parsing natural language. That's why Google is open sourcing its new language parsing model for English, which it calls'Parsey McParseface.' Before you even ask, the name has no meaning. When Google was trying to figure out what to call its language parsing technology, someone suggested Parsey McParseface; it's a bit like Apple's Liam, which has no clever backstory either. The overall AI model model is called SyntaxNet (please make your SkyNet jokes now); 'ol Parsey is just for English. Our biggest ever edition of TNW Conference is fast approaching!

The "SP theory of intelligence", with its realisation in the "SP computer model", aims to simplify and integrate observations and concepts across AI-related fields, with information compression as a unifying theme. This paper describes how abstract structures and processes in the theory may be realised in terms of neurons, their interconnections, and the transmission of signals between neurons. This part of the SP theory -- "SP-neural" -- is a tentative and partial model for the representation and processing of knowledge in the brain. In the SP theory (apart from SP-neural), all kinds of knowledge are represented with "patterns", where a pattern is an array of atomic symbols in one or two dimensions. In SP-neural, the concept of a "pattern" is realised as an array of neurons called a "pattern assembly", similar to Hebb's concept of a "cell assembly" but with important differences. Central to the processing of information in the SP system is the powerful concept of "multiple alignment", borrowed and adapted from bioinformatics. Processes such as pattern recognition, reasoning and problem solving are achieved via the building of multiple alignments, while unsupervised learning -- significantly different from the "Hebbian" kinds of learning -- is achieved by creating patterns from sensory information and also by creating patterns from multiple alignments in which there is a partial match between one pattern and another. Short-lived neural structures equivalent to multiple alignments will be created via an inter-play of excitatory and inhibitory neural signals. The paper discusses several associated issues, with relevant empirical evidence.

Human communication takes many forms, including speech, text and instructional videos. It typically has an underlying structure, with a starting point, ending, and certain objective steps between them. In this paper, we consider instructional videos where there are tens of millions of them on the Internet. We propose a method for parsing a video into such semantic steps in an unsupervised way. Our method is capable of providing a semantic "storyline" of the video composed of its objective steps. We accomplish this using both visual and language cues in a joint generative model. Our method can also provide a textual description for each of the identified semantic steps and video segments. We evaluate our method on a large number of complex YouTube videos and show that our method discovers semantically correct instructions for a variety of tasks.

A lot of discussion around Matt Jockers' Syuzhet package (involving Annie Swafford, Ted Underwood, Andrew Piper, Scott Weingart and many others) has focused on issues of validity -- whether sentiment analysis is accurate enough for the task, whether the Fourier transform is an appropriate method for dimensionality reduction, whether the emotional trajectories themselves are valid measurements of anything at all (Scott has a good enumeration of the various issues here.) Andrew's discussion of the validity of inherently subjective measurements inspired me to solicit at least one data point from readers that we can use for one question under discussion with Syuzhet: what does a human judgment of the "emotional trajectory" of a work look like, and how often do readers agree with each other on this task? This method of soliciting human judgments for inherently subjective tasks is at the core of NLP and a lot of machine learning -- syntactic parsing, part of speech tagging, named entity recognition, topic classification, sentiment analysis, and lots of other tasks all rely on humans making judgments that are often surprisingly difficult in practice; learning algorithms in these cases are not so much learning any notion of "truth" but simply to reproduce the human judgments they're given. Agreement rates between humans is often seen as a proxy for the complexity of the task; if humans can't agree, it can be a sign that the task is ill-defined or underspecified. Word sense disambiguation is one good example of this, with low inter-annotator agreement rates [Snyder and Palmer 2004]; while sentiment analysis was originally designed with product/movie reviews in mind (does person X like product Y?) -- i.e., attitude with respect to a particular target -- I think the more general sentiment-as-tone problem (is this tweet happy or sad?) is much less well specified as a problem with an answer that can be judged by anyone but the original author. One aspect of those kind of annotations that I think is much less explored (which Piper points to and I think would be an extremely interesting area to work on) is the case where multiple judgments are simultaneously valid -- different interpretations of the same phenomenon, each backed by their own argument.

As you may know we recently launched a new service offering, our News API, and over the past week or so we've been using it to run some little experiments around analyzing news content. We wanted to use the News API to collect and analyze popular news headlines. We set out to find both similarities and differences in the way two journalists write headlines for their respective news articles and blog posts. Note: For a more technical, in-depth and interactive representation of this project, check out the Jupyter notebook we created. This includes sample code and more in depth descriptions of our approach.

This paper investigates the problem of cross-lingual transfer parsing, aiming at inducing dependency parsers for low-resource languages while using only training data from a resource-rich language (e.g., English). Existing model transfer approaches typically don't include lexical features, which are not transferable across languages. In this paper, we bridge the lexical feature gap by using distributed feature representations and their composition. We provide two algorithms for inducing cross-lingual distributed representations of words, which map vocabularies from two different languages into a common vector space. Consequently, both lexical features and non-lexical features can be used in our model for cross-lingual transfer. Furthermore, our framework is flexible enough to incorporate additional useful features such as cross-lingual word clusters. Our combined contributions achieve an average relative error reduction of 10.9% in labeled attachment score as compared with the delexicalized parser, trained on English universal treebank and transferred to three other languages. It also significantly outperforms state-of-the-art delexicalized models augmented with projected cluster features on identical data. Finally, we demonstrate that our models can be further boosted with minimal supervision (e.g., 100 annotated sentences) from target languages, which is of great significance for practical usage.