Monthly Archives: December 2016

Prototype display enables viewers

3-D movies immerse us in new worlds and allow us to see places and things in ways that we otherwise couldn’t. But behind every 3-D experience is something that is uniformly despised: those goofy glasses.

Fortunately, there may be hope. In a new paper, a team from MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) and Israel’s Weizmann Institute of Science have demonstrated a display that lets audiences watch 3-D films in a movie theater without extra eyewear.

Dubbed “Cinema 3D,” the prototype uses a special array of lenses and mirrors to enable viewers to watch a 3-D movie from any seat in a theater.

“Existing approaches to glasses-free 3-D require screens whose resolution requirements are so enormous that they are completely impractical,” says MIT professor Wojciech Matusik, one of the co-authors on a related paper whose first author is Weizmann PhD Netalee Efrat. “This is the first technical approach that allows for glasses-free 3-D on a large scale.”

While the researchers caution that the system isn’t currently market-ready, they are optimistic that future versions could push the technology to a place where theaters would be able to offer glasses-free alternatives for 3-D movies.

Among the paper’s co-authors are MIT research technician Mike Foshey; former CSAIL postdoc Piotr Didyk; and two Weizmann researchers that include Efrat and professor Anat Levin. Efrat will present the paper at this week’s SIGGRAPH computer-graphics conference in Anaheim, California.

Glasses-free 3-D already exists, but not in a way that scales to movie theaters. Traditional methods for TV sets use a series of slits in front of the screen (a “parallax barrier”) that allows each eye to see a different set of pixels, creating a simulated sense of depth.

But because parallax barriers have to be at a consistent distance from the viewer, this approach isn’t practical for larger spaces like theaters that have viewers at different angles and distances.

Other methods, including one from the MIT Media Lab, involve developing completely new physical projectors that cover the entire angular range of the audience. However, this often comes at a cost of lower image-resolution.

The key insight with Cinema 3D is that people in movie theaters move their heads only over a very small range of angles, limited by the width of their seat. Thus, it is enough to display images to a narrow range of angles and replicate that to all seats in the theater.

What Cinema 3D does, then, is encode multiple parallax barriers in one display, such that each viewer sees a parallax barrier tailored to their position. That range of views is then replicated across the theater by a series of mirrors and lenses within Cinema 3D’s special optics system.

“With a 3-D TV, you have to account for people moving around to watch from different angles, which means that you have to divide up a limited number of pixels to be projected so that the viewer sees the image from wherever they are,” says Gordon Wetzstein, an assistant professor of electrical engineering at Stanford University, who was not involved in the research. “The authors [of Cinema 3D] cleverly exploited the fact that theaters have a unique set-up in which every person sits in a more or less fixed position the whole time.”

Practical applications for non native English

After thousands of hours of work, MIT researchers have released the first major database of fully annotated English sentences written by non-native speakers.

The researchers who led the project had already shown that the grammatical quirks of non-native speakers writing in English could be a source of linguistic insight. But they hope that their dataset could also lead to applications that would improve computers’ handling of spoken or written language of non-native English speakers.

“English is the most used language on the Internet, with over 1 billion speakers,” says Yevgeni Berzak, a graduate student in electrical engineering and computer science, who led the new project. “Most of the people who speak English in the world or produce English text are non-native speakers. This characteristic is often overlooked when we study English scientifically or when we do natural-language processing for English.”

Most natural-language-processing systems, which enable smartphone and other computer applications to process requests phrased in ordinary language, are based on machine learning, in which computer systems look for patterns in huge sets of training data. “If you want to handle noncanonical learner language, in terms of the training material that’s available to you, you can only train on standard English,” Berzak explains.

Systems trained on nonstandard English, on the other hand, could be better able to handle the idiosyncrasies of non-native English speakers, such as tendencies to drop or add prepositions, to substitute particular tenses for others, or to misuse particular auxiliary verbs. Indeed, the researchers hope that their work could lead to grammar-correction software targeted to native speakers of other languages.

Diagramming sentences

The researchers’ dataset consists of 5,124 sentences culled from exam essays written by students of English as a second language (ESL). The sentences were drawn, in approximately equal distribution, from native speakers of 10 languages that are the primary tongues of roughly 40 percent of the world’s population.

Every sentence in the dataset includes at least one grammatical error. The original source of the sentences was a collection made public by Cambridge University, which included annotation of the errors, but no other grammatical or syntactic information.

To provide that additional information, Berzak recruited a group of MIT undergraduate and graduate students from the departments of Electrical Engineering and Computer Science (EECS), Linguistics, and Mechanical Engineering, led by Carolyn Spadine, a graduate student in linguistics.

After eight weeks of training in how to annotate both grammatically correct and error-ridden sentences, the students began working directly on the data. There are three levels of annotation. The first involves basic parts of speech — whether a word is a noun, a verb, a preposition, and so on. The next is a more detailed description of parts of speech — plural versus singular nouns, verb tenses, comparative and superlative adjectives, and the like.

Next, the annotators charted the syntactic relationships between the words of the sentences, using a relatively new annotation scheme called the Universal Dependency formalism. Syntactic relationships include things like which nouns are the objects of which verbs, which verbs are auxiliaries of other verbs, which adjectives modify which nouns, and so on.

The annotators created syntactic charts for both the corrected and uncorrected versions of each sentence. That required some prior conceptual work, since grammatical errors can make words’ syntactic roles difficult to interpret.

Berzak and Spadine wrote a 20-page guide to their annotation scheme, much of which dealt with the handling of error-ridden sentences. Consistency in the treatment of such sentences is essential to any envisioned application of the dataset: A machine-learning system can’t learn to recognize an error if the error is described differently in different training examples.

Repeatable results

The researchers’ methodology, however, provides good evidence that annotators can chart ungrammatical sentences consistently. For every sentence, one evaluator annotated it completely; another reviewed the annotations and flagged any areas of disagreement; and a third ruled on the disagreements.