Morphology beyond inflection. - PowerPoint Presentation

Slide1

Morphology beyond inflection.

Building a wordformation based lexicon for Latin

Eleonora Litta Modignani

PicozziMSCA FellowCIRCSE Research CentreUniversità Cattolica del Sacro Cuore, MilanoMarie Skłodowska-Curie grant agreement No 658332-WFL

Inflectional vs derivational

Morphology (for Latin)Word formation relations

not treated by computational lexical resources or morphological analysers.Halfway between morphology

and semanticsWFR not only build new words, but create new words with a shared semantic core, which can be useful for NLP tools.First tools for modern languages (Czech, Croatian, German).

The Word Formation Latin (WFL) project

Started by Passarotti and Mambrini in 2012Awarded Marie Curie

Fellowship Definitive derivational lexicon for Latin.

The lexical basis: LemLat

Morphological analyser for Latin.Data collected from

three dictionaries: Georges and Georges, Ausfuhrliches Lateinisch-Deutsches Handwôrterbuch (1913-1918)Glare, Oxford Latin Dictionary

(1982)Gradenwitz: Laterculi vocum latinarum (1904).40,014 lexical entries, 43,432 lemmas, and 26,205 lemmas from Forcellini’s Onomasticon (1940).Lexical basis for WFL: LES (LExical Segment) archive and List of Lemmas.

Word formation

Item-and-Arrangement model: word forms are 1) simple morphemes or

2) concatenation of morphemes absolving the following conditions: Baudoin’s assumption that both base and affixes are lexical elements (i.e. they are both morphemes), They are dualistic, having both form and meaning (Bloomfield’s “sign-base” morpheme theory) They both exist in a lexicon (Bloomfield’s “lexical morpheme” theory)

Formalising WFL

Word formation based lexicon built in three

steps:WFRs are detectedWFRs are applied to the lexical data3) Results

are manually checked and evaluated

Types of word f

ormationDerivation:

Affixal: Prefixal: duco => con-duco Suffixal: amo => am-a-bil-

isConversion : bonus (adj.) => bonum (noun)Compounding : magnus + facio = magnificus

Detection of word formation

rules (WFRs)

Semi-automatic finding of affixal rules (Passarotti & Mambrini 2012).List of possibile combination of PoS for conversion (e.g. V-to-V, V-to-N, V-to-A, etc.)

and compounding (A+V=N, A+V=A, etc.).WFRs formalised into a table according to category of change, type of word formation, input PoS and output PoS. 50 additional rules found so far while working through the data.

WFR list

MySql relational database

MySQL

queries

WFR

paired candidates: V-to-V prefix sub-

Finding prefixal

and suffixal candidates

Manual

checking

and disambiguatingfocaria - n ‘

kitchen maid’focarius - n ‘kitchen servant’focarius - adj. ‘belonging to fire’

Conversion A-to-N2

Conversion A-to-N1

Conversion N2-to-N1

Evaluation

Precision is higher for lower morphotactic mutations (e.g.prefixal

rules)Precision lower for obscure WFRs (e.g. compounding)Precision can also depend on workflow

Recall needs to be calculated when we have finding WFRs

Viewing the relations

Visualisation query system (by Chris Culy)Browsing

options:By WFRBy AffixBy PoSBy Lemma

Thank you!

WFL team

Eleonora LittaMarco PassarottiVisualisationsChris Culy (http://chrisculy.net/)DB Engineering

Paolo RuffoloWebsitehttp://progetti.unicatt.it/progetti-milan-wfl-homehttps://www.facebook.com/wordformationlatin/WFL has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 658332-WFL.