Language Processing Pipelines Part 3
Formal Metadata
Title 
Language Processing Pipelines Part 3

Alternative Title 
FreeLing 4.1  An OpenSource Suite of Language Analyzers

Title of Series  
Part Number 
3

Number of Parts 
3

Author 

License 
No Open Access License:
German copyright law applies. This film may be used for your own use but it may not be distributed via the internet or passed on to external parties. 
Identifiers 

Publisher 

Release Date 
2019

Language 
English

Production Year 
2019

Production Place 
Dubrovnik, Croatia

Content Metadata
Subject Area  
Keywords  natural language processing language processing pipeline 
00:00
Pattern recognition
Group action
Line (geometry)
File format
Mathematical analysis
Function (mathematics)
Infinity
Ordinary differential equation
Plot (narrative)
Neuroinformatik
Formal language
Measurement
Writing
Finite element method
Particle system
Forschungszentrum Rossendorf
Representation (politics)
Metropolitan area network
Pressure
Data type
Rule of inference
Parsing
Execution unit
Graph (mathematics)
Theory of relativity
Information
Suite (music)
Open source
Mathematical analysis
Range (statistics)
Formal language
Repeating decimal
CAN bus
Type theory
Number
Uniform resource locator
Word
Frequency
Software
Logic
Predicate (grammar)
Function (mathematics)
Right angle
Form (programming)
Protein folding
03:07
Point (geometry)
Pattern recognition
Group action
Parsing
Graph (mathematics)
PRINCE2
File format
Mathematical analysis
Writing
Summierbarkeit
Data type
Rule of inference
Graph (mathematics)
Suite (music)
Tap (transformer)
Forcing (mathematics)
Open source
Bit
Formal language
Number
Arithmetic mean
Word
Process (computing)
Computer configuration
Function (mathematics)
04:45
Pattern recognition
Parsing
Graph (mathematics)
Codierung <Programmierung>
Mathematical singularity
Mathematical analysis
Function (mathematics)
Formal language
Number
Writing
Normal (geometry)
Form (programming)
Data type
Selforganization
Rule of inference
Parsing
Graph (mathematics)
Suite (music)
Tap (transformer)
Open source
ACID
Formal language
Degree (graph theory)
Plane (geometry)
Planar graph
Number
Process (computing)
Computer configuration
Function (mathematics)
Atomic number
Vacuum
06:28
Execution unit
Empennage
Slide rule
View (database)
Building
Demo (music)
Lemma (mathematics)
Token ring
Plastikkarte
Simulated annealing
Inclusion map
Normed vector space
Universal product code
Finitary relation
Process (computing)
output
Task (computing)
Data type
Selforganization
06:45
Point (geometry)
Meta element
Building
Service (economics)
State of matter
Multiplication sign
Source code
Set (mathematics)
Mereology
Formal language
Number
Product (business)
Mathematics
Different (Kate Ryan album)
Software
Repository (publishing)
Software framework
Maize
Information
Logic gate
Address space
Task (computing)
Module (mathematics)
Area
Service (economics)
Execution unit
Electric generator
Artificial neural network
Electronic mailing list
Independence (probability theory)
Line (geometry)
Formal language
Componentbased software engineering
Word
Process (computing)
Software
Repository (publishing)
Universe (mathematics)
Chain
Quicksort
Logic gate
Reading (process)
10:50
Mathematics
Hooking
Mathematics
Process (computing)
Natural number
Natural language
Process (computing)
Formal language
Fingerprint
00:00
of course you can process much love their logic chunk of tanks but for the season.
00:05
the. and i say i want the planned to parsing and may you have language out to detect a team not secure so want. you get analysis so you have a sentence here with the limousine blow and the participation them is the taking and you have a par three here is the dependency parsley but the sparse three candide and then the. listed in a call for but which is common in computation was taken and the people and the he also see that the. name entities are recognised paris as a location unesco as location so no person unesco's person so that's a problem and unesco is not really being recognized as it should be or that you can get also very detailed x.m.l. output as this one looks like that this his own. anyone have some type of selected output so if i go to semantic graph and i will receive and much more. details are so you see the dependency same tactic relations are up there and the cement think rafa is down there and if you look at the call for much then you have additional information so these are like asking tactic roles and. these are. word senses annotated following the knowledge available so these are worth senses here and there on the right hand side you have semantic role and also recognised of to matter. and that's a man to grab and gives you something like that so you have like four main predicates in this the recognized and you see the relations in cement a graph looks like that so you have hold meets that's derives from the from the now meeting. you but you have the underlying semantic action to meet you have prepare. and you have undertake which is next to undertake next and so once you see the a m one actually is and so no this is ten temperature elation and you have a zero and his relation to its so it's a meat. and undertaken so on so and i mean this is a graphical representation can be two weeks i mean this at leicester city of this network can be tweaked but that's just a question how to show what has been no to magically analyzed out of this. these sentence and then you can have the same sentence.
03:10
you slovenian. so it's translated and. the end of their i'm afraid we will we don't have some of the and more jews available so multiverse detector detection force living in doesn't wear and then we'll see what else doesn't work and will switch it off. on to his actions not available. so these are the the the more jobs that are missing or. the main city classification is not available something else is not available in check. the word sense does in ghana we got something a bit. yes but we didn't get the cement a graph it's very. for many very modest one a soul to organize means meeting unesco just that and but as you see dependency parsing is working and you have some some. words and sanitation this this one and the few you click there. then you open the right sense in the princeton with a. unknown but it's the wrong word sense of their this is three point one and this was analysed by two point zero so we tend to be translated in congress from from all diversion to the new or so it opened the wrong longer litter.
04:49
and then if we are.
04:53
and another. sentencing then the portuguese. it. degrees deserve their lives or two of his sentence. today language is a semantic graph is not available for portuguese a k. and it. not independent departments of their own gain. for the parsing years. i think the whole parsons on. apart from that. ok so as you see this just goes to curiously parsing so this is not do the dependence apart so it's a different form is a. and you get in calling for much which is a useful for further processing of course or x.m.l. output as well which is in more detail. so that's what the friehling actually took some of the background the process is from acts like and developed an added that to a number of languages so it's not the are very exhaustive number of languages but still there is some of these are not. all of them are covered with almost use that's that's something what we would like to build up for sure.
06:33
when i was so i've shown that and early move quickly and my top with the following things.
06:40
and. ok so that's the x.m.l. were well what you might find a very useful in your future where it can and then experiments is to try to find the existing solutions for some of the task that you would like to use or.
07:02
to do so i'll give you a list of different language repositories where you can find a lot of things i mean there will be communities quite already quite mature you can find tools and services for almost every task you need for almost that doesn't mean that. that's really the be already done for language that you want to cover but then you can find is for another languages for english at least and see how this can be ported maybe maybe that solution is language independent but no one has ever tried it on a different language is so one of the things would be european research infrastructure. they're letting attic particularly you can look at the and linguistic processing the framework for building all own line building graphically building linguistic processing china change there has been developed which is called of a british team has been been developed do have my number of a german universities together. but the them entrance point is and just wanted to be at the address of this the european research infrastructure is clarine dot edu and another repository which is actually a federation of european repositories for four different a researcher mehmood for different. in which technologies of resources and tools is called meant i share. and them and the in the next one is european languages sources agency the telegraph and week shake truly them. these two are mostly for free in l.a. you lose have to pay for some of resources and then in l.a. you in predominantly funny source is not so many tools but linguistic data consortium use an american repository at the university offender states at the time. and then you really there you can find also some resources that that they come for free and some are some around a places but i think these and these are four main repository as for language technologies in general and then there are some other frameworks that allow you to be loose.
09:21
processing chains one of the most popular in known as gates from the university of sheffield and the friehling a's show you just a feeling it it's been run by. polytechnic university in the catalonian from barcelona and you have of course and will be stanford set of tools and and they and not always so easily the you can not always build chains out of that so easily as in rebel east or where you have already complete change like these friehling. org eight but still they have a valuable in pieces of software that could be and very often language independent so did a day sort of the favor of the language independent approach and and of course today if you look at the commercial and and and. it takes products that packages like i.b.m. what's on only one might fix they will include already language technology modules in their analytical packages they offered that as it is a part of a commercial product and the fate tell you. that's the end of a p will look to make a difference particularly if we include a neural networks in training new generation of the modules and n.o.p. tools and well don't take my word for that read this guy and so you have.
10:55
the data for in bright future for and will be why natural language processing will change everything so it's been written three years ago and still is valid after that. i would surely seen a thank you for attention.
11:11
so poor. i.