Chatbots and Automated Speech Recognition (ASR) in EFL/ESL Classroom

The ability to understand human language;  to ask a question and to receive an answer, to generate realistic and logically consistent responses are the core of human communication. Moreover, with the rapid progression of technology, speech understanding technologies; chatbots and automated speech recognition (ASR) progressed as well, and now a range of speech understanding applications are available to language learners and teachers.  Furthermore, These applications enable EFL/ESL teachers and learners to address a variety of learning issues.  In other words, the successful implementations of these technologies give the learners the chance to improve their speaking skills. Namely, speech understanding technologies give the learners the opportunity to interact with a ‘speaker’ of English whenever they choose. For instance, automated speech recognition (ASR) can be implemented in the teaching and learning process (A2-B2, 12+, a variety of topics) as a useful pedagogical tool to boost meaning-focused speaking interaction. In other words, it gives the learners the opportunity to have meaningful face-to-face discussions with the virtual interlocutor.   An additional value of ASR is in their versatility to address a variety of learning issues and topics. Namely, ASR is a way for students to improve their pronunciation, gain awareness of their articulation and boost their speaking skills. 

However, ASR has also limitations, which should be taken into consideration.  The straightforward process of speech recognition by the machine is a complicated process by variability issues. One issue is connected with the peaks and troughs of the speech signal, that may not line up correctly when words are said at a slower or faster rate.  Moreover, the variability among speakers is another problem: voice quality, an accent that all affect speech characteristics. Furthermore, unlike printed text, spoken language does not have clear-cut boundaries between words. Therefore, sounds at the beginnings and ends of words could be produced slightly differently depending on their immediate context.

While there is still another way to provide language learners with conversation practice machine: 'chatbot'. The linguistic worth of  ‘chatbot’ software programs is to hold a conversation or interact, as a precursor to their potential as an ESL  learning resource.  Thus, Chatbots can be implemented in the teaching and learning process, as a suitable pairing of pedagogy and technology, if the following aspects are taken into consideration:

 Chatbots work well when the language input to them consists of one-clause sentences and when the topic is an everyday one.  It is noteworthy that chatbots which perform better converse in a single domain and have specialised vocabularies and discussion topics (A2-B2, 12+, vocabulary: single domain).

 Moreover,  Chatbots need correctly-spelt words in grammatically correct sentences. Some chatbots can deal with misspellings, but this is often unpredictable, as the ability of chatbots to maintain a cohesive exchange over multiple conversations is limited.  Thus, these limitations are non-trivial and cannot be overlooked. However, for practical purposes, the virtual dialogue can avoid the problems of speech synthesis and lack of conversational cohesion, and in this case, it remains a worthwhile option for CALL application.