Poster presented at the 2nd meeting of the COST Action CA16105 - enetCollect : European Network for Combining Language Learning with Crowdsourcing Techniques, which took place at Alexandru Ioan Cuza University, in Iasi, Romania.
The poster shows how chatbots can play an important role in Language Learning applications.
1. Investigating the Role of Chatbots for Language Learning Applications
Paraphrases of Formal/Informal Language
Language learners need to learn and understand
formal and informal ways of communication to
fully grasp the knowledge of a language
Chatbots can be an additional feature to be
implemented into Language Learning applications
to (i) converse, (ii) entertain, and (iii) improve the
understanding and use of formal and informal
language
The eSPERTo project, whose goal is to develop a
paraphrasing system, has been showing how
paraphrases are crucial for Language Learning
Paraphrasing is a technique that consists of
rewriting a sentence or text using different words
One of eSPERTo’s goals is to extend and enrich the
knowledge base of a virtual agent (chatbot)
Chatbots with Paraphrastic Knowledge
Chatbots with paraphrastic knowledge can engage
into conversations, because they have language
understanding (≠ parrot mode)
Paraphrastic knowledge provides the capacity to
make semantic associations, correctly answer or
reply to the same question in different but
semantically-equivalent ways (i.e., using different
words and syntactic structures)
One of the studies performed with eSPERTo is to
provide an automatic solution for normalization of
informal syntactic “inadequacies” into formal
written standard structures, through the
application of transformational grammars
The study focused on paraphrasing and converting verbal
constructions typical of informal or spoken language into formal
language, or vice versa, in order to teach European and Brazilian
Portuguese language varieties. This will prepare learners for using
the language appropriately to the situation and context
Automation of Paraphrasing in Several Languages
and the Use of Crowdsourcing
Our proposal is to:
(i) Include an automation process in chatbots
(ii) explore the idea with broader linguistic
phenomena
(iii) expand the methodology to other languages
(iv) use crowdsourcing to enlarge the paraphrastic
knowledge database
Crowdsourcing can help develop a language
database of formal and informal phrases in specific
conversational scenarios in various languages. This
effort can lead to the creation and expansion of a
paraphrase corpus, which can be used in a broad
range of Natural Language Processing (NLP) tools
for a wide variety of applications.
§ Are supposed to engage in a conversation about any subject
§ Should handle a small talk and provide plausible responses
§ Enable texting/chatting in a quasi-conversational environment
Conversation
Learning
§ Enable learners to practice language in a spoken/written form
§ Conversations relate to specific/targeted domain topics
§ Learners often confront them with out-of-domain input, which
result in chatbots giving incorrect/inappropriate feedback
§ Chatbots give feedback to related tasks
Entertainment
§ Have simple faces/personalities/shapes
§ They are equipped with an additional variety of conversational
topics, applying the following strategies:
1) Admitting ignorance
2) Changing the conversation topic
3) Introducing small stories
§ Have gamification elements, such as rewarding learners for
achieving learning goals
1. A menina generosa queria-o surpreender todos os dias
2. A menina generosa queria surpreendê-lo todos os dias
"the generous girl wanted to surprise him everyday"
§ eSPERTo presents as a conversion option for the informal complex
verbal predicate with clitic in 1., its formal equivalent in 2.
§ The transformation requires the infinitive form of surpreender
‘surprise’ to change to surpreendê, which is used before the
enclitic dative pronouns o , a , os , as.
The ending changes from -er to –ê and the clitic receives an –l for phonetic reasons.
§ Style revision is an important feature of the eSPERTo project
§ eSPERTo has the capacity to identify and generate semantically
equivalent sentences and ways of expression
§ It provides suggestions that users can choose to revise their texts
§ It helps improve text quality by transforming sentences written in
an informal language into formal sentences, corresponding to the
standardized norm for the written language
Informal to formal speech (or vice versa)
queria-o surpreender queria surpreendê-lo
Video on eSPERTo with more suggestions for language learning:
www.youtube.com/watch?v=_0kxvX1m35Y
eSPERTo - Paraphrasing Informal into Formal Portuguese (Barreiro et al. 2018)
System Chatbots
FalaComigo – www.l2f.inesc-id.pt/w/FalaComigo_(Enhance_the_Cultural_Tourism_through_the_Interaction_with_Virtual_Characters)
Anabela Barreiro, INESC-ID
anabela.barreiro@inesc-id.pt
Branislav Bédi, University of Iceland
brb19@hi.is