Speech recognition is almost as natural as breathing for us, but for a computer, it has taken more than half a century to solve this ‘problem’. Previously, the fundamental drawbacks to Speech Recognition were its poor accuracy, sensitivity to noise, over dependence on training to a particular voice and similar problems meant it worked in principle, but not in practice. This has been hugely improved now, it is often reaching the high nineties in percentage terms for several reasons: the general increase in the availability of affordable computing power, the advent of the cloud and the vast numbers of people now using it. Last year, IBM announced a major milestone in conversational speech recognition by building a system that achieved a 6.9 percent word error rate. Since then, it has continued to push the boundaries of speech recognition. Today, it reached a new industry record of 5.5 percent where these are measured on a very difficult speech recognition task: recorded conversations between humans discussing day-to-day topics like “buying a car.” This recorded corpus, known as the “SWITCHBOARD” corpus and has been used for over two decades to benchmark speech recognition systems.

The 21st century has seen many improvements in this field. In the 2000s DARPA sponsored two speech recognition programs: Effective Affordable Reusable Speech-to-Text (EARS) in 2002 and Global Autonomous Language Exploitation (GALE). The National Security Agency has made use of a type of speech recognition for keyword spotting since 2006. This technology allows analysts to search through large volumes of recorded conversations and isolate mentions of keywords. Google‘s first effort at Speech Recognition came in 2007 when its first product “GOOG-411” a telephone based directory service was released. Now, Google voice search is supported in over 30 languages and particularly in 2015, Google’s speech recognition reportedly experienced a dramatic performance jump of 49% through new techniques involving deep learning.

These advancements in Speech Recognition Technology had diversified its application. It has been implemented by many Healthcare and Military organizations:

Health care

Medical documentation

In the health care sector, Speech Recognition is implemented in front-end or back-end of the medical documentation process. In Front-end speech recognition, the provider dictates into a speech-recognition engine, the recognized words are displayed as they are spoken, and the dictator is responsible for editing and signing off on the document. Whereas, in Back-end or deferred speech recognition the provider dictates into a digital dictation system, the voice is recognized and a draft document is made out of it which is routed along with the original voice file to the editor, where the draft is edited and finalized. Deferred speech recognition is widely used in the industry currently.

Therapeutic use

Particularly in short-term-memory re-strengthening of brain AVM patients, the use of speech recognition software in conjunction with word processors has shown significant benefits. Further research needs to be conducted to determine cognitive benefits for individuals whose AVMs have been treated using radiologic techniques.

Military

High-performance fighter aircraft

Significant progress in the test and evaluation of Speech Recognition in fighter aircraft has taken place in the last decade. Of particular note have been the US program in speech recognition for the Advanced Fighter Technology Integration (AFTI)/F-16 aircraft (F-16 VISTA). In this program, speech recognizers have been operated successfully in fighter aircraft, with applications including setting radio frequencies, commanding an autopilot system, setting steer-point coordinates and weapons release parameters, and controlling flight display.

Also, speaker-independent systems are being developed and are under test for the F35 Lightning II (JSF). This system has produced word accuracy scores in excess of 98%.

Training air traffic controllers

Training for air traffic controllers (ATC) represents an excellent application for speech recognition systems. In the current scenario, many ATC training systems need a person to act as a “pseudo-pilot”, engaging in a voice dialog with the trainee controller, which simulates the dialog that the controller would have to conduct with pilots in a real ATC situation. Speech recognition techniques can eliminate the need for a person to act as pseudo-pilot, thus reducing training and support personnel. The USAF, USMC, US Army, US Navy, and FAA as well as a number of international ATC training organizations are currently using ATC simulators with speech recognition from different vendors.

 

Editor’s note: Original Sources

---

 

• https://en.wikipedia.org/wiki/Speech_recognition
• Schutte, John (15 October 2007). “Researchers fine-tune F-35 pilot-aircraft speech system”. United States Air Force. Archived from the original on 20 October 2007
• https://www.ibm.com/blogs/research/2017/03/speech-recognition/

Biometrics has a wide array of potential usages, which is why different agencies and governments are developing the software in order to provide more efficient and secure identification processes.

US Military

The United States military is in the beginning phases of utilizing behavioral biometric software in order for soldiers and other military personnel to access technology. What used to be the territory of a common access key card or passcode is now being modified to be individualized, greatly reducing the risk of threat. This behavioral technology goes further than a thumbprint or an iris scan. It analyzes the way a soldier walks, or the way a technician types on a keyboard and uses a mouse. The same systems somebody is using is constantly collecting data on them. This continuous proof of presence will deliver a cyber security solution that promptly recognizes breaches, can assist in forensic investigations, and guarantees regulatory compliance.

Airport Security

Biometrics have started to become the long-awaited answer for numerous security dilemmas in the United States. U.S. Customs and Border Protection has begun implementing biometric software at airports nationwide. Following a long overdue mandate (15 years), new biometrics will solve the difficulty of scanning the identity of thousands of departing foreign visitors every day. The combination of cloud computing and the latest facial recognition technology will turn a process from about two minutes into only a couple of seconds, countering the rise in foreign travel into the US.

Border Security

The State Department is also developing biometric technology to enhance coordination on border security and migration activities on the US-Mexico border. According to a January report out of the Congressional Research Service, The U.S. and Mexican governments in 2015 actually approved a $75 million Mérida program to help Mexico “develop an automated biometrics system to help agencies collect and share information on criminals and migrants.” This is perhaps an alternative to the estimated $12 billion dollars it would take to build a border wall, according to Senate leader Mitch McConnell.

Editor’s note: Original Source ‘Fedscoop’, ‘Defenseone’

---

Patrick Tucker. “The Future of Military IT: Gait Biometrics, Software Nets, and Photon Communicators”

Defenseone. N.p.15 June. 2017, Web. 18 July. 2017

FedScoop. “Military testing behavioral ID technology that would replace CAC card”

FedScoop. N.p. 6 July. 2017, Web. 18 July. 2017.

Last summer, Congress passed and President Obama signed into law the FOIA Improvement Act of 2016 (Public Law No. 114-185), which adds to and amends the Freedom of Information Act.

The amendments create a “presumption of openness” limiting the federal government’s discretionary power to withhold requested information only when disclosure would result in “foreseeable harm.”

For those that transact business with or even simply communicate with the government (referred to as “submitters” in FOIA parlance), the FOIA changes mean that submitters such as government contractors and grant recipients must proactively respond when a FOIA request potentially targets confidential and/or proprietary data that has been shared with the government.

Importantly, the 2016 FOIA improvement Act did not change FOIA Exemption 4, which protects from disclosure “trade secrets and commercial or financial information obtained from a person [that is] privileged or confidential.” Under Exemption 4, the government is prohibited from disclosing trade secrets or other proprietary/confidential information that any submitter has shared with the government.

Unlike with some of the other FOIA exemptions, in their interpretation of Exemption 4, courts have determined that the government lacks any discretion to disclose trade secret or commercial confidential/proprietary information in response to a FOIA request.

The 2016 FOIA Improvement Act was passed to accelerate the FOIA process and to compel government FOIA officials to provide as much information as soon as possible in response to a FOIA request. The act now imposes a penalty (i.e., the waiver of the statutory FOIA fees) on the agency for failing to provide a timely FOIA response. The act also requires that the FOIA response segregate exempt information from releasable information in the same document, as an agency can no longer simply refuse to produce any document containing exempt information.

In addition, the Act requires the agency to produce electronic copies of documents/data, which can be instantly disseminated by the requesting party, rather than paper documents, in response to a FOIA request.

Furthermore, the act requires the creation of a federal government FOIA portal that allows the same FOIA request to be simultaneously submitted to multiple agencies. As a result, submitters must be poised to respond immediately as soon as the government provides notice that a FOIA request seeks disclosure of the submitter’s data and/or documents.

As an initial step, whenever any person or entity first shares information/data with the government that it does not want disclosed to any third party, the title page and each subsequent page of the confidential document or data should be plainly marked as containing “confidential and proprietary information which is exempt from disclosure under FOIA.”

Next, when the agency contacts the submitter (as FOIA requires) to tell them that a request seeks the disclosure of their information, the submitter should promptly respond by identifying:

1. The specific information within each responsive document that is exempt from disclosure.
2. The particular FOIA exemption (there are nine) that prohibits disclosure (as stated above, Exemption 4 protects trade secrets and confidential/proprietary data)
3. Why that exemption applies to each identified section of data/information that the submitter seeks to protect.

Also, the submitter (or submitter’s counsel) should attempt to maintain an open dialogue with the assigned agency FOIA official throughout the FOIA process to promptly address and resolve any disagreements about what should and should not be disclosed before the agency takes a final disclosure position, which is often difficult to unwind.

Finally, the submitter must be ready to assert a “reverse FOIA” action to prevent the disclosure of trade secrets or other confidential/proprietary information in the event that the agency disregards the submitter’s exemption recommendations before the agency releases the submitter’s trade secrets and confidential information in response to a FOIA request.

Editor’s note: Original Source ‘Washington Technology’

---

Doug Proxmire. “New FOIA rules open contractors to more risks of disclosure”

Washington Technology. N.p., Web. 17 February. 2017.

Spookily powerful artificial intelligence (AI) systems may work so well because their structure exploits the fundamental laws of the universe, new research suggests.

The new findings may help answer a longstanding mystery about a class of artificial intelligence that employ a strategy called deep learning. These deep learning or deep neural network programs, as they’re called, are algorithms that have many layers in which lower-level calculations feed into higher ones. Deep neural networks often perform astonishingly well at solving problems as complex as beating the world’s best player of the strategy board game Go or classifying cat photos, yet know one fully understood why.

It turns out, one reason may be that they are tapping into the very special properties of the physical world, said Max Tegmark, a physicist at the Massachusetts Institute of Technology (MIT) and a co-author of the new research.

The laws of physics only present this “very special class of problems” — the problems that AI shines at solving. “This tiny fraction of the problems that physics makes us care about and the tiny fraction of problems that neural networks can solve are more or less the same.

Deep learning

Last year, AI accomplished a task many people thought impossible: DeepMind, Google’s deep learning AI system, defeated the world’s best Go player after trouncing the European Go champion. The feat stunned the world because the number of potential Go moves exceeds the number of atoms in the universe, and past Go-playing robots performed only as well as a mediocre human player.

But even more astonishing than DeepMind’s utter rout of its opponents was how it accomplished the task.

“The big mystery behind neural networks is why they work so well,” said study co-author Henry Lin, a physicist at Harvard University. “Almost every problem we throw at them, they crack.”

For instance, DeepMind was not explicitly taught Go strategy and was not trained to recognize classic sequences of moves. Instead, it simply “watched” millions of games, and then played many, many more against itself and other players.

Like newborn babies, these deep-learning algorithms start out “clueless,” yet typically outperform other AI algorithms that are given some of the rules of the game in advance.

Another long-held mystery is why these deep networks are so much better than so-called shallow ones, which contain as little as one layer. Deep networks have a hierarchy and look a bit like connections between neurons in the brain, with lower-level data from many neurons feeding into another “higher” group of neurons, repeated over many layers. In a similar way, deep layers of these neural networks make some calculations, and then feed those results to a higher layer of the program, and so on, he said.

Magical keys or magical locks?

To understand why this process works, Tegmark and Lin decided to flip the question on its head.

“Suppose somebody gave you a key. Every lock you try, it seems to open. One might assume that the key has some magic properties. But another possibility is that all the locks are magical. In the case of neural nets, I suspect it’s a bit of both,” Lin said.

One possibility could be that the “real world” problems have special properties because the real world is very special, Tegmark said.

Take one of the biggest neural-network mysteries: These networks often take what seem to be computationally hairy problems, like the Go game, and somehow find solutions using far fewer calculations than expected.

It turns out that the math employed by neural networks is simplified thanks to a few special properties of the universe. The first is that the equations that govern many laws of physics, from quantum mechanics to gravity to special relativity, are essentially simple math problems, Tegmark said. The equations involve variables raised to a low power (for instance, 4 or less).

What’s more, objects in the universe are governed by locality, meaning they are limited by the speed of light. Practically speaking, that means neighboring objects in the universe are more likely to influence each other than things that are far from each other, Tegmark said.

Many things in the universe also obey what’s called a normal or Gaussian distribution. This is the classic “bell curve” that governs everything from traits such as human height to the speed of gas molecules zooming around in the atmosphere.

Finally, symmetry is woven into the fabric of physics. Think of the veiny pattern on a leaf, or the two arms, eyes and ears of the average human. At the galactic scale, if one travels a light-year to the left or right, or waits a year, the laws of physics are the same, Tegmark said.

Tougher problems to crack

All of these special traits of the universe mean that the problems facing neural networks are actually special math problems that can be radically simplified.

“If you look at the class of data sets that we actually come across in nature, they’re way simpler than the sort of worst-case scenario you might imagine,” Tegmark said.

There are also problems that would be much tougher for neural networks to crack, including encryption schemes that secure information on the web; such schemes just look like random noise.

“If you feed that into a neural network, it’s going to fail just as badly as I am; it’s not going to find any patterns,” Tegmark said.

While the subatomic laws of nature are simple, the equations describing a bumblebee flight are incredibly complicated, while those governing gas molecules remain simple, Lin added. It’s not yet clear whether deep learning will perform just as well describing those complicated bumblebee flights as it will describing gas molecules, he said.

“The point is that some ’emergent’ laws of physics, like those governing an ideal gas, remain quite simple, whereas some become quite complicated. So there is a lot of additional work that needs to be done if one is going to answer in detail why deep learning works so well.” Lin said. “I think the paper raises a lot more questions than it answers!”

 

Editor’s note: Original Source ‘LiveScience’

---

Tia Ghose. “The Spooky Secret Behind Artificial Intelligence’s Incredible Power”

LiveScience. N.p., Web. 7 October. 2016.