Accuracy of IBM Watson speech recognition is low - machine-learning

I develop an application that uses speech-to-text to transcribe audio to text. The accuracy is low. Some sentences have no meaning. Is there a way to improve the accuracy of speech-to-text?
Here's an example:
http://book.vidalab.co/books/alice-in-wonderland
Alice in Wonderland, in section 2:
"over at home to go white pawn this way you see ads"
should be "over at home to go white pawn this way you see Alice"
"rat in white"
should be "red and white"
"and the white army tries to win and the red on the Trice twin"
should be "and the white army tries to win and the red army tries to win"

You can try different services, for example Speechmatics, it's not very good at getting speakers but words are much more accurate than from Watson, the result is like this:
Credits of Alice in Wonderland by Alice girs Timberg this is a box recording all of her vocal recordings are in the public domain for more information or volunteer. Please visit libber Vox dot org.
I just listed stage directions read by McKayla Curtis Lewis Carroll.
Read by Shannon Brown Alice read by Amanda Friday the Red Queen read by Shauna canat White Queen read by Elizabeth Klatt White Rabbit read by Todd Humpty Dumpty read by Jeff Machado written read by Brett Hirsch.
The Mock Turtle read by Ted the alarm Mad Hatter read by Elliot gage the March Hare by Charlotte Duckett's dormouse read by Kimberly Krauss frog read by Larry Wilson Duchess read by L.A. Cheshire Cat read by Sarah Herschell Tweedle-Dee read.
By Charlotte Brown.
Do you do do I read by the sea a solo the King of Hearts read by Ted alarm the Queen of Hearts read by eating Ray Headrick knave by glorious Joe Carter pillar back at 2 loss to spot read by Dave Harris.
Five Spot read by Dave Harith. Seven of spades read by Dave Hereth end of credits.
Surnames recognition is very complex task, not many companies are doing it properly.

There are two major parts in any STT system: acoustic model and language model. The first one is about audio and speaker and handles things like: noise, pronunciations, accents and so on. Language Model is about structure of a given language and the words used in the speech.
If you would like to test an STT, use the recordings which are as close as possible to your target speech. A system which performs very well for general speech, or, for example, medical transcription, may be not a good in handle speech about archeology or poetry.e

Related

How would someone create a machine learning algorithm that extracts the speaker from a book/novel?

Basically organizes the content based on the speaker?
Excerpt From: Robert Louis Stevenson. “The Strange Case of Dr. Jekyll and Mr. Hyde.”
Example Input:
But Lanyon's face changed, and he held up a trembling hand. "I wish to see or hear no more of Dr. Jekyll," he said in a loud, unsteady voice. "I am quite done with that person; and I beg that you will spare me any allusion to one whom I regard as dead.
Example Output:
[
“Narrator”: “But Lanyon's face changed, and he held up a trembling hand.”,
“Lanyon”: “I wish to see or hear no more of Dr. Jekyll”,
“Narrator”: “he said in a loud, unsteady voice.”,
“Lanyon”: “I am quite done with that person; and I beg that you will spare me any allusion to one whom I regard as dead.”
]
I have not heard of the algorithm that does exactly this. But there are two well known problem that could be useful: named entity recognition (to find all potential speakers) and anaphora resolution (to decide who "he" or "she" is in each case).
You would also need to train a classifier that for each quoted chunk of text to decide whether it is a direct speech. And you would probably need another classifier to decide for each identified piece of speech and for each identified speaker in the context, how likely is that this speech actually belongs to this speaker.

Have figure appear close to the reference when ithe mages are included at the end of the document in latex

Hi I have a long document such as this wikipedia entry on Alice in wonderland below. I reference the figure on the rabbit in the first line but I import the figure at the very end of the document. I now want the figure to appear close to the reference in the pdf. Does any one have any suggestions? [h] does not work...
The White Rabbit (Figure \ref{fig:rabbit})
Chapter One – Down the Rabbit Hole: Alice is feeling bored and drowsy while
sitting on the riverbank with her older sister, who is reading a book with
no pictures or conversations. She then notices a White Rabbit wearing a
waistcoat and pocket watch, talking to itself as it runs past. She follows
it down a rabbit hole, but suddenly falls a long way to a curious hall with
many locked doors of all sizes. She finds a small key to a door too small
for her to fit through, but through it she sees an attractive garden. She
then discovers a bottle on a table labelled DRINK ME, the contents of which
cause her to shrink too small to reach the key, which she has left on the
table. She eats a cake with EAT ME written on it in currants as the chapter
closes.
Chapter Two – The Pool of Tears: Chapter Two opens with Alice growing to
such a tremendous size that her head hits the ceiling. Alice is unhappy and,
as she cries, her tears flood the hallway. After shrinking down again due to
a fan she had picked up, Alice swims through her own tears and meets a
Mouse, who is swimming as well. She tries to make small talk with him in
elementary French (thinking he may be a French mouse) but her opening gambit
Où est ma chatte? (Where is my cat?) offends the mouse and he tries to
escape her.
Chapter Three – The Caucus Race and a Long Tale: The sea of tears becomes
crowded with other animals and birds that have been swept away by the rising
waters. Alice and the other animals convene on the bank and the question
among them is how to get dry again. The Mouse gives them a very dry lecture
on William the Conqueror. A Dodo decides that the best thing to dry them off
would be a Caucus-Race, which consists of everyone running in a circle with
no clear winner. Alice eventually frightens all the animals away,
unwittingly, by talking about her (moderately ferocious) cat.
Chapter Four – The Rabbit Sends a Little Bill: The White Rabbit appears
again in search of the Duchess's gloves and fan. Mistaking her for his
maidservant, Mary Ann, he orders Alice to go into the house and retrieve
them, but once she gets inside she starts growing. The horrified Rabbit
orders his gardener, Bill the Lizard, to climb on the roof and go down the
chimney. Outside, Alice hears the voices of animals that have gathered to
gawk at her giant arm. The crowd hurls pebbles at her, which turn into
little cakes. Alice eats them, and they make her smaller again. Upon leaving
the house, she comes upon a puppy, with whom she plays for a while before
finding a mushroom with a hookah-smoking blue Caterpillar.
Chapter Five – Advice from a Caterpillar: The Caterpillar questions
Alice and she admits to her current identity crisis, compounded by her
inability to remember a poem. Before crawling away, the caterpillar tells
Alice that one side of the mushroom will make her taller and the other side
will make her shorter. She breaks off two pieces from the mushroom. One side
makes her shrink smaller than ever, while another causes her neck to grow
high into the trees, where a pigeon mistakes her for a serpent. With some
effort, Alice brings herself back to her normal height. She stumbles upon a
small estate and uses the mushroom to reach a more appropriate height.
The Cheshire Cat
Chapter Six – Pig and Pepper: A Fish-Footman has an invitation for the
Duchess of the house, which he delivers to a Frog-Footman. Alice observes
this transaction and, after a perplexing conversation with the frog, lets
herself into the house. The Duchess's Cook is throwing dishes and making a
soup that has too much pepper, which causes Alice, the Duchess, and her baby
(but not the cook or grinning Cheshire Cat) to sneeze violently. Alice is
given the baby by the Duchess and to her surprise, the baby turns into a
pig. The Cheshire Cat appears in a tree, directing her to the March Hare's
house. He disappears, but his grin remains behind to float on its own in the
air, prompting Alice to remark that she has often seen a cat without a grin
but never a grin without a cat.
Chapter Seven – A Mad Tea-Party: Alice becomes a guest at a mad tea party
along with the March Hare, the Hatter, and a very tired Dormouse who falls
asleep frequently, only to be violently woken up moments later by the March
Hare and the Hatter. The characters give Alice many riddles and stories,
including the famous Why is a raven like a writing desk?. The Hatter reveals
that they have tea all day because Time has punished him by eternally
standing
still at 6 pm (tea time). Alice becomes insulted and tired of being
bombarded with riddles and she leaves, claiming that it was the stupidest
tea party that she had ever been to. She makes her way back to the hallway
and, using the pieces of
mushroom, manages to get the key, unlock the door and enter the garden.
\begin{figure}
\centering
\includegraphics[width=0.4\textwidth]{./Figures/rabit.png}
\caption[Rabbit] {Picture of rabbit.}
\label{fig:rabbit}
\end{figure}
I'm afraid this is impossible to do in pure LaTeX. LaTeX's figure placements are restricted to 'here', 'top', 'bottom' and 'page'. None of these are sufficient for your case. 'h' from 'here' means close to where the figure is defined, not referenced.
It is however possible to have your .tex file processed by some other tool to move the figure definition closer to the reference and then have the 'here' placement applied, or to not style the image as a float at all.
Tools that can do this include awk and perl.

How to get a single video from YouTube using its id?

I am trying to use the YouTube Data API v3 to get a single video from its YouTube id. I am using the Google API python client. When I try to execute the following code:
search_response = youtube.search().list(
id="QkhBcLk_8f0",
part="snippet",
).execute()
I always get this error:
Traceback (most recent call last):
File "search.py", line 55, in <module>
youtube_search(args)
File "search.py", line 24, in youtube_search
part="snippet",
File "/usr/local/lib/python2.7/dist-packages/googleapiclient/discovery.py",
line 669, in method
raise TypeError('Got an unexpected keyword argument "%s"' % name)
TypeError: Got an unexpected keyword argument "id"
But I know there is a id parameter in the API, as it is listed in the reference.
Anybody know what am I doing wrong?
You should use youtube.videos() method.
youtube.videos().list(id="QkhBcLk_8f0",part="snippet").execute()
{u'etag': u'"YxyobdYztCvdjXOUqpUttvF39GM/nr0_6QcUDyGR0D_Gxz762lsqqfU"',
u'items': [{u'etag': u'"YxyobdYztCvdjXOUqpUttvF39GM/n7wasYBTHMM3SB4Jtsxu6JeDFPA"',
u'id': u'QkhBcLk_8f0',
u'kind': u'youtube#video',
u'snippet': {u'categoryId': u'27',
u'channelId': u'UCzu2OUGZlNtXPom3KNPGFzg',
u'channelTitle': u'FFreeThinker',
u'description': u'http://facebook.com/ScienceReason ... Great Minds, Great Words: Richard Feynman - The Uncertainty of Knowledge ... The Nature and Purpose of the Universe.\n\nPlaylist "Great Minds, Great Words":\n\u2022 http://www.youtube.com/user/FFreeThinker#grid/user/CC4F721030F8D4D1\n\n---\nPlease SUBSCRIBE to Science & Reason:\n\u2022 http://www.youtube.com/FFreeThinker\n\u2022 http://www.youtube.com/ScienceTV\n\u2022 http://www.youtube.com/Best0fScience\n\u2022 http://www.youtube.com/RationalHumanism\n---\n\nRichard Feynman (1918-1988) was an American physicist known for his work in the path integral formulation of quantum mechanics, the theory of quantum electrodynamics and the physics of the superfluidity of supercooled liquid helium, as well as in particle physics (he proposed the parton model).\n\nFor his contributions to the development of quantum electrodynamics, Feynman, jointly with Julian Schwinger and Sin-Itiro Tomonaga, received the Nobel Prize in Physics in 1965. He developed a widely used pictorial representation scheme for the mathematical expressions governing the behavior of subatomic particles, which later became known as Feynman diagrams. During his lifetime, Feynman became one of the best-known scientists in the world.\n\nHe assisted in the development of the atomic bomb and was a member of the panel that investigated the Space Shuttle Challenger disaster. In addition to his work in theoretical physics, Feynman has been credited with pioneering the field of quantum computing, and introducing the concept of nanotechnology (creation of devices at the molecular scale). He held the Richard Chace Tolman professorship in theoretical physics at the California Institute of Technology.\n\nFeynman was a keen popularizer of physics through both books and lectures, notably a 1959 talk on top-down nanotechnology called "There\'s Plenty of Room at the Bottom" and "The Feynman Lectures on Physics". Feynman also became known through his semi-autobiographical books ("Surely You\'re Joking, Mr. Feynman!" and "What Do You Care What Other People Think?") and books written about him, such as "Tuva or Bust!"\n\nHe was regarded as an eccentric and free spirit. He was a prankster, juggler, safecracker, proud amateur painter, and bongo player. He liked to pursue a variety of seemingly unrelated interests, such as art, percussion, Maya hieroglyphs, and lock picking.\n\nFeynman also had a deep interest in biology, and was a friend of the geneticist and microbiologist Esther Lederberg, who developed replica plating and discovered bacteriophage lambda. They had several mutual physicist friends who, after beginning their careers in nuclear research, moved for moral reasons into genetics, among them Le\xf3 Szil\xe1rd, Guido Pontecorvo, and Aaron Novick.\n\n\u2022 http://en.wikipedia.org/wiki/Richard_Feynman\n.',
u'liveBroadcastContent': u'none',
u'localized': {u'description': u'http://facebook.com/ScienceReason ... Great Minds, Great Words: Richard Feynman - The Uncertainty of Knowledge ... The Nature and Purpose of the Universe.\n\nPlaylist "Great Minds, Great Words":\n\u2022 http://www.youtube.com/user/FFreeThinker#grid/user/CC4F721030F8D4D1\n\n---\nPlease SUBSCRIBE to Science & Reason:\n\u2022 http://www.youtube.com/FFreeThinker\n\u2022 http://www.youtube.com/ScienceTV\n\u2022 http://www.youtube.com/Best0fScience\n\u2022 http://www.youtube.com/RationalHumanism\n---\n\nRichard Feynman (1918-1988) was an American physicist known for his work in the path integral formulation of quantum mechanics, the theory of quantum electrodynamics and the physics of the superfluidity of supercooled liquid helium, as well as in particle physics (he proposed the parton model).\n\nFor his contributions to the development of quantum electrodynamics, Feynman, jointly with Julian Schwinger and Sin-Itiro Tomonaga, received the Nobel Prize in Physics in 1965. He developed a widely used pictorial representation scheme for the mathematical expressions governing the behavior of subatomic particles, which later became known as Feynman diagrams. During his lifetime, Feynman became one of the best-known scientists in the world.\n\nHe assisted in the development of the atomic bomb and was a member of the panel that investigated the Space Shuttle Challenger disaster. In addition to his work in theoretical physics, Feynman has been credited with pioneering the field of quantum computing, and introducing the concept of nanotechnology (creation of devices at the molecular scale). He held the Richard Chace Tolman professorship in theoretical physics at the California Institute of Technology.\n\nFeynman was a keen popularizer of physics through both books and lectures, notably a 1959 talk on top-down nanotechnology called "There\'s Plenty of Room at the Bottom" and "The Feynman Lectures on Physics". Feynman also became known through his semi-autobiographical books ("Surely You\'re Joking, Mr. Feynman!" and "What Do You Care What Other People Think?") and books written about him, such as "Tuva or Bust!"\n\nHe was regarded as an eccentric and free spirit. He was a prankster, juggler, safecracker, proud amateur painter, and bongo player. He liked to pursue a variety of seemingly unrelated interests, such as art, percussion, Maya hieroglyphs, and lock picking.\n\nFeynman also had a deep interest in biology, and was a friend of the geneticist and microbiologist Esther Lederberg, who developed replica plating and discovered bacteriophage lambda. They had several mutual physicist friends who, after beginning their careers in nuclear research, moved for moral reasons into genetics, among them Le\xf3 Szil\xe1rd, Guido Pontecorvo, and Aaron Novick.\n\n\u2022 http://en.wikipedia.org/wiki/Richard_Feynman\n.',
u'title': u'Great Minds: Richard Feynman - The Uncertainty Of Knowledge'},
u'publishedAt': u'2010-03-04T15:12:56.000Z',
u'tags': [u'great',
u'minds',
u'words',
u'richard',
u'feynman',
u'uncertainty',
u'knowledge',
u'nature',
u'purpose',
u'universe',
u'god',
u'religion',
u'atheists',
u'atheism',
u'science',
u'physicists',
u'quantum',
u'mechanics',
u'electrodynamics',
u'superfluidity',
u'nobel',
u'prize',
u'theoretical',
u'physics',
u'atomic',
u'bomb',
u'space',
u'nano',
u'technology'],
u'thumbnails': {u'default': {u'height': 90,
u'url': u'https://i.ytimg.com/vi/QkhBcLk_8f0/default.jpg',
u'width': 120},
u'high': {u'height': 360,
u'url': u'https://i.ytimg.com/vi/QkhBcLk_8f0/hqdefault.jpg',
u'width': 480},
u'maxres': {u'height': 720,
u'url': u'https://i.ytimg.com/vi/QkhBcLk_8f0/maxresdefault.jpg',
u'width': 1280},
u'medium': {u'height': 180,
u'url': u'https://i.ytimg.com/vi/QkhBcLk_8f0/mqdefault.jpg',
u'width': 320},
u'standard': {u'height': 480,
u'url': u'https://i.ytimg.com/vi/QkhBcLk_8f0/sddefault.jpg',
u'width': 640}},
u'title': u'Great Minds: Richard Feynman - The Uncertainty Of Knowledge'}}],
u'kind': u'youtube#videoListResponse',
u'pageInfo': {u'resultsPerPage': 1, u'totalResults': 1}}

Getting/Indexing what Named-Entities have said/quoted

I am trying to semi-automatically identify what people/advisers/economists or the like say on the news:
1
But Barclays chief economist Kieran Davies said companies were simply opting against bank funding.
“I don’t think there’s actually a ­constraint on corporate credit from the banks. I think it’s more the case that ­corporates in recent years have either had sufficient cash or access to offshore funding to enable them to do their investment,” Mr Davies said.
2
The the amount of commercial lending for every dollar of residential property lending has plunged from $3.84 to $1.62 over the past 25 years, says ­analysis from Industry Super Australia.
“Consistent analysis demonstrates that we have a systematic issue ­transitioning national savings to real productive capital, such as nation-building infrastructure,” chief ­executive David Whiteley said.
3
Following the settlement, Macquarie’s then chief executive in Asia, Alex Harvey, said he was “quite ­encouraged about SATC, about the risk management framework we’ve now put in place over the last few years and the sort of opportunities that are in front of the trust.”
so with example #1, i would like to get what Mr Davies was saying,
and with example #2, what chief executive David Whiteley was saying,
and with example #3, what Alex Harvey was saying.
what is the best way of getting / indexing what those 3 people said in other 10000 or more articles ?
it feels like a mixed of Name-Entity-Tagging and Relationship-Extraction. Is there a more specific name to what I want to achieve ?

RedPitaya case or housing options?

The RedPitaya is a great looking instrument, but I'm afraid that I'll kill my new (expensive) device by stray voltage or ESD off my bench, within a few days.
Is it planned to make an optional "professional" case or similar to protect it?
Has anyone already created a 3D model so a printable case or housing could be made?
As a quick fix, I would:
use a piece of plain printer paper on the bench, underneath the Red Pitaya (it's typically more conductive than a typical plastic coating on the bench, but still not so cunductive as to short anything on the board bottom), and
more importantly, each time when approaching the bench, first touch the outside of one of the golden SMA jacks.
Probably any quick google search would answer the question but for the sake of completeness, I will answer this quesiton with what I found in my quick search.
Purchase Options
Now a days, there are several cases available for the pitaya:
Available on RS-components, Reichelt, amoung others:
RS Code: RS819-4077
Manufacturer: Red Pitaya
Manufacturer Ref: 1600 0715 001
Approximate cost: 20 Euros + Shipping + taxes
What appears to be 3d printed providers:
Nylon Plastic closed top case : Approximate 30 Euros
Nylon Plastic open top case
Printing Options
If you happen to have your own 3D printer then you can print one of many available designs.
Closed Top case
Open Top case
A Shielded Case Github Prject, on Thingiverse, on Youmagine
Others can be found on http://www.yeggi.com/ , http://grabcad.com/ ,...

Resources