A.I. Natural Language Processing

	Starts	Organizer
	Sat 01 Jun 2013 15:00	Everyone interested
	Ends	Event Owner
	Sat 01 Jun 2013 16:00	User:Skmp

Let the skynet begin

Ευθύμης Πετρόπουλος-Τράκας "Πρώτη συνάντηση, για την Επεξεργασία Φυσικής Γλώσσας. Σας περιμένω όλους εκεί, με μπόλικη διάθεση, υλικό προς μελέτη και φυσικά ιδέες!"

(coordination and note taking by skmp)

http://doodle.com/zrag5ak9atvzd8yd

definitions

• • nlp - natural language processing
  • nlg - natural language generation

• People
  • theodim
  • skmp
  • Nikos Katzouris
  • Orestis Roditis
  • Timos Petropoulos
    • Interests ~ Cryptography (as applicable to nlp ?), Apply newer algos to greek
  • Panagiotis Katsivelis
    • Interests ~ nlg
  • Vasilis Salapatas

referenced projects

• • newsum
  • Standfold parser
  • asoee
    • pos-tagger (nlp.cs.aueb.gr)
    • Name entity recogniser (NER)
  • Eclipse IDE
    • pydev extension
  • simple nlg
  • NLTK (Natural Language Toolkit) in Python
  • wordnet
  • ispell
  • alchemy (sentiment analysis)
  • Openbox
  • iperdiavgeia
  • alibaba
  • Wiktionary
  • europarl
  • Ethinko tipografio
  • podio
  • oasana
  • wordforms
  • Duolingo

General chat

• • Introduction of each person
  • Plan is for a Study group (once per one or two weeks)
  • We'll pick a subject and work on it
    • apokodikopiisi grammikis a ~
  • Statistical vs structured algorithms
  • language discussion (it always must happen :p)
    • Standfold parser & related libraries (java-based)
  • mailing list tag decided to be [nlp]
  • language chat again
    • perhaps we can decide on a common base ?
  • nlg
    • using a dictionary
    • There's nothing for greek right now
  • nltk like, but updated w/ support for greek
  • hosting probably on github
  • back to project talk
  • ispell / spell checking
  • Sentiment analysis
    • alchemy
  • nothing in Greece really
    • Not really open/open source
  • turney algorithm
    • tries to match betwen bad/good
    • Implement in Greek ?
  • Dimokritos, OpenGov (useful for obtaining greek corpora?)
  • iperdiavgeia -> search diavgeia api
  • provlepsi aftoktonion meso fb
  • Sentiment analysis
  • topic detection is (more general than sentiment analysis)
  • Aggregation of open knowledge
  • its hard to get data
  • Scale of required corpus
    • 2M "entries" is a good number
  • alibaba
  • ptixiaki tou panagioti
    • glosiko modelo, paragogi diafimiseon apo perigrafi, using nlg
    • 47k corpus is wayy too small
  • scrapping
    • using browser
  • dictionaries - parsing existing ones from the 'net
    • Existing ones kind of suck
  • Open data in greek we can use
  • europarl
    • hugely big corpus
    • can be used to train
  • Ethinko tipografio
    • Another open source of information
  • reverse dictionary
  • wordforms

project ideas

• • apokodikopiisi grammikis a ~
  • basic dictionary-based nlg for greek (panagiotis)
  • nltk like, but updated w/ support for greek (who ?)
  • Dictionary/language graph for greek ? (theodim)
  • spell checker improvement (theodim)
  • turney-based sentiment analysis for Greek
  • topic detection (theodim)
  • Corpus -- there's nothing in greek -- perhaps crawl for it ? (panagiotis)

decisions

• • doodle next meetup
  • investigate pontio
  • For next time, we'll have presentations of the tools everyone has used (more like a study group and not format presentations)
  • get registed on wiki !
  • get registed on hackerspace discuss!

Contact

Μέσω της λίστας γενική συζήτησης βάζοντας το label [NLP] στην αρχή του subject.