MIT researchers have developed a “semantic parser” that learns through observation to more closely mimic a child’s language-acquisition process, which could greatly extend computing’s capabilities.

Credit: MIT News

In computing, learning language is the task of syntactic and semantic parsers. These systems are trained on sentences annotated by humans that describe the structure and meaning behind words. Parsers are becoming increasingly important for web searches, natural-language database querying, and voice-recognition systems such as Alexa and Siri. Soon, they may also be used for home robotics.

But gathering the annotation data can be time-consuming and difficult for less common languages. Additionally, humans don’t always agree on the annotations, and the annotations themselves may not accurately reflect how people naturally speak.

In a paper being presented at this week’s Empirical Methods in Natural Language Processing conference, MIT researchers describe a parser that learns through observation to more closely mimic a child’s language-acquisition process, which could greatly extend the parser’s capabilities. To learn the structure of language, the parser observes captioned videos, with no other information, and associates the words with recorded objects and actions. Given a new sentence, the parser can then use what it’s learned about the structure of the language to accurately predict a sentence’s meaning, without the video.

This “weakly supervised” approach — meaning it requires limited training data — mimics how children can observe the world around them and learn language, without anyone providing direct context. The approach could expand the types of data and reduce the effort needed for training parsers, according to the researchers. A few directly annotated sentences, for instance, could be combined with many captioned videos, which are easier to come by, to improve performance.

In the future, the parser could be used to improve natural interaction between humans and personal robots. A robot equipped with the parser, for instance, could constantly observe its environment to reinforce its understanding of spoken commands, including when the spoken sentences aren’t fully grammatical or clear. “People talk to each other in partial sentences, run-on thoughts, and jumbled language. You want a robot in your home that will adapt to their particular way of speaking … and still figure out what they mean,” says co-author Andrei Barbu, a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Center for Brains, Minds, and Machines (CBMM) within MIT’s McGovern Institute.

Read more about it here (Massachusetts Institute of Technology. “Machines that learn language more like kids do: Computer model could improve human-machine interaction, provide insight into how children learn language.” ScienceDaily. ScienceDaily, 31 October 2018.)

Scientists report that they have developed a powerful new printer that could streamline the creation of self-assembling structures that can change shape after being exposed to heat and other stimuli. They say this unique technology could accelerate the use of 4-D printing in aerospace, medicine and other industries.

“We are on the cusp of creating a new generation of devices that could vastly expand the practical applications for 3-D and 4-D printing,” H. Jerry Qi, Ph.D., says. “Our prototype printer integrates many features that appear to simplify and expedite the processes used in traditional 3-D printing. As a result, we can use a variety of materials to create hard and soft components at the same time, incorporate conductive wiring directly into shape-changing structures, and ultimately set the stage for the development of a host of 4-D products that could reshape our world.”

Credit: H. Jerry Qi

4-D printing is an emerging technology that allows 3-D-printed components to change their shape over time after exposure to heat, light, humidity and other environmental triggers. However, 4-D printing remains challenging, in part because it often requires complex and time-consuming post-processing steps to mechanically program each component. In addition, many commercial printers can only print 4-D structures composed of a single material.

Last year, Qi and his colleagues at Georgia Institute of Technology, in collaboration with scientists at the Singapore University of Technology and Design, used a composite made from an acrylic and an epoxy along with a commercial printer and a heat source to create 4-D objects, such as a flower that can close its petals or a star that morphs into a dome. These objects transformed shape up to 90 percent faster than previously possible because the scientists incorporated the tedious mechanical programming steps directly into the 3-D printing process. Building on this work, the researchers sought to develop an all-in-one printer to address other 4-D printing challenges and move the technology closer to practical application.

The machine they ultimately devised combines four different printing techniques, including aerosol, inkjet, direct ink write and fused deposition modeling. It can handle a multitude of stiff and elastic materials including hydrogels, silver nanoparticle-based conductive inks, liquid crystal elastomers and shape memory polymers, or SMPs. SMPs, which are the most common substances used in 4-D printing, can be programmed to “remember” a shape and then transform into it when heated. With this new technology, the researchers can print higher-quality SMPs capable of making more intricate shape changes than in the past, opening the door for a multitude of functional 4-D applications and designs.

The researchers can also use the printer to project a range of white, gray or black shades of light to form and cure a component into a solid. This grayscale lighting triggers a crosslinking reaction that can alter the component’s behavior, depending on the grayscale of shade shined on it. So, for example, a brighter light shade creates a part that is harder, while a darker shade produces a softer part. As a result, these components can bend or stretch differently than other parts of the 4-D structure around them.

The printer can even create electrical wiring that can be printed directly onto an antenna, sensor or other electrical device. The process relies on a direct-ink-write method to produce a line of silver nanoparticle ink. A photonic cure unit dries and coalesces the nanoparticles to form conductive wire. Then, the printer’s ink-jet component creates the plastic coating that encases the wire.

Currently, Qi’s team is also working with Children’s Healthcare of Atlanta to determine whether this new technology could print prosthetic hands for children born with malformed arms.

Also, recently there has been advance in developing the world’s first-ever 4D printing for ceramics, which are mechanically robust and can have complex shapes. This could turn a new page in the structural application of ceramics.

Credit: City University of Hong Kong

Read more in these articles:

Fabrication technology in the fourth dimension (ETH Zurich. “Fabrication technology in the fourth dimension.” ScienceDaily. ScienceDaily, 8 May 2017.)

Original paper: Chen, T., Mueller, J. and Shea, K., 2017. Integrated design and simulation of tunable, multi-state structures fabricated monolithically with multi-material 3D printing. Scientific reports, 7(1), pp.1-8.

New 4-D printer could reshape the world we live in (American Chemical Society. “New 4-D printer could reshape the world we live in.” ScienceDaily. ScienceDaily, 21 March 2018.)

World’s first-ever 4D printing for ceramics (City University of Hong Kong. “World’s first-ever 4D printing for ceramics.” ScienceDaily. ScienceDaily, 18 August 2018.)

Original paper: Liu, G., Zhao, Y., Wu, G. and Lu, J., 2018. Origami and 4D printing of elastomer-derived ceramic structures. Science advances, 4(8), p.eaat0641.