PresentationPDF Available

Recognizing blends: First experiments with PYBOSSA



In this presentation, we show how the PYBOSSA crowdsourcing platform can be used for collection of linguistic knowledge about the Dutch language. As a pilot study, we ask speakers of Dutch to recognize and analyze blends, a certain type of compound word.
Recognizing blends:
First experiments with
Peter Dekker
Tanneke Schoonheim
Instituut voor de Nederlandse Taal
enetCollect WG3&5 meeting Leiden, 24-25 October 2018
Introduction of INT
Instituut voor de Nederlandse Taal (Dutch Language Institute)
Scholarly institute in the field of the Dutch language
Central position in the Dutch-speaking world
Developer, keeper and distributor of corpora, lexica, dictionaries
and grammars
Provider of necessary building blocks of the study of Dutch
Introduction of INT
Current staff of INT:
17 (computational and corpus) linguists,
lexicographers, terminologists,
4 linguistic assistants and trainees
5 software engineers, 1 system administrator
5 administration and communication
Introduction of INT
Current projects at INT:
Contemporary and historical dictionaries and dictionary portals
Contemporary and historical corpora and lexica
Grammar portal, spelling database, terminology lists
Infrastructure, tools and data for linguistic research (CLARIN)
Introduction of INT
Relatively new projects at INT:
Development and hosting of products for educational purposes, such as
Bilingual dictionaries (New Greek, Portuguese, Estonian)
Dutch Word Combinations
Corpus Eenvoudig Nederlands (Corpus of Elementary Dutch)
Can crowdsourcing help us developing these and other educational products?
Research objective
Task solved by public: answer unknown
User details not important
Traditional (socio)linguistic research/survey:
Answer of task known beforehand, in many cases
User details important
Our research objective: combination
PYBOSSA can be suitable, more solutions exist
PYBOSSA installation
Hosted version ( vs hosting on own server
Own server: existing infrastructure at INT, more flexibility
Clear installation guide on PYBOSSA website
Complexity: PYBOSSA consists of multiple software packages
Ansible script: recipe for reproducible installation
Experiments with blends: Data
Compound of two words, where parts of the words are lost
Signifies a new meaning, related to the words it consists of
glamping (glamour + camping)
mup (millenial + yup)
Blends in English: Gries, S. T. (2004). Shouldn't it be breakfunch? A quantitative analysis of blend structure in English.
Linguistics, 639-668.
Experiments with blends: Data
Blends are part of two related projects at INT:
Algemeen Nederlands Woordenboek (ANW; Dictionary of Contemporary Dutch)
Neologism portal (upcoming)
Neologism Workflow at INT:
1. Data from newspapers and websites
2. Processed automatically, new words put aside
3. Lexicographer selects neologisms, creates entries in
a. neologism portal (all neologisms)
b. ANW dictionary (rooted neologisms only)
Experiments with blends: Crowd
Where did we find the crowd?
Newsletter Instituut voor de Nederlandse Taal
Congress Internationale Vereniging voor Neerlandistiek
Experiments with blends: Jobs
Can the crowd help in recognizing and analyzing blends?
Two jobs created in PYBOSSA:
Blend recognition
Recognize blends in a text
Blend analysis
Analyze the words a blend consists of
10 tasks per job
User interface design
Freedom in UI design: design using HTML and Javascript
PYBOSSA only loads and saves tasks from database
User details
Blends analysis
Blends recognition
Results: Age
Results: Gender
Results: Location
Results: Blends analysis
n = 326
Results: Blends analysis for preferendum
Analysis Frequency
referendum, pre 16
[Don’t know] 11
Multiple word forms (noun, verb) for prefer
Multiple interpretations (at least when word is
presented without context):
referendum + to prefer
referendum + pre
Results: Blends recognition
n = 223
Results: Blends recognition for twittie
twittie 122
twittie, fittie 56
fittie 16
twittie, tweet, fittie 5
[Do not know] 4
twittie: twitter + fittie ‘fight’ (slang)
fittie itself also occurred in text:
misinterpreted as blend
More input fields (3) than real blends per task (1):
stimulates giving more blends
User feedback
English language of PYBOSSA, while tasks are
about Dutch
Too many buttons
Task not always clear
Make welcome page attractive
Experiences with PYBOSSA
Freedom when developing tasks
Share tasks with other researchers
Everything else (account system and loading/saving tasks) handled by PYBOSSA
Quick answers from developers via bug tracker
No ready-made translation for all languages
PYBOSSA not designed for asking user details
When uploading large number of tasks, there is no clear end of job (you have to code that yourself)
User cannot easily go back to a previous task
User identification by IP address does not always work
Possible alternative for some purposes: Google Forms
Future experiments
Neologisms and dialects
User detail prediction as reward
Interesting issue:
Is PyBossa better suited for these tasks than for instance Google forms?
Is crowdsourcing useful for the analysis of blends?
Yes, because it gives an insight in how blends are interpreted by non-linguists.
Is PYBOSSA useful for this kind of crowdsourcing?
Yes, powerful platform, with its own strengths and drawbacks.
The paper addresses an investigation of the area of collaborative practices and crowdsourcing applications, in particular those that contribute to vast linguistic data/repositories collection, i.e., data on sociolinguistics, lexicography, as well as terminology and translation. The concept of crowdsourcing is defined in terms of collaboration practices such as various types of innovation and industry, and primarily of linguistic applications. A lexicological analysis of language forms, in their dialectal and social variations and the tools relevant for the crowdsourcing activity (TAALRADAR) developed at the Dutch Language Institute in Leiden are demonstrated and discussed. Linguistic research, language and translator training are identified as the areas benefitting from the collaborative and crowdsource tasks and collected data. Their consequences aiming to contribute to current methodological vistas in linguistics and language/terminology acquisition will be presented by showing evidence of how creation of a common collaborative space, typically in terms of cloud technology at present, for linguists, language learners, translator trainees and other specialists, can work and contribute to the development of language proficiency, enrichment of language use techniques and terminology acquisition, as well as the quality of language product. The competences focusing on these techniques, particularly those related to how crowdsourcing applications lead towards a more general phenomenon which involves collaborative knowledge acquisition and Internet-based task sharing, are presented and their position in language and knowledge acquisition is discussed.
ResearchGate has not been able to resolve any references for this publication.