Chatbots and SR
Chatbots are programmed the way they only know what they are taught, so they cannot understand the context of humans and interact with them in natural conversational language. I don't think a chatbot can be an appropriate partner to practice a language, though in reading books for children it can be useful. When I was interacting with George I noticed that he replies to the last sentence only, giving inadequate and unsufficient answers, it avoids answering through asking questions. These are all disadvantages that the science tends to solve. A chatbot can be used in ESL classrooms for fun practice, but they are not useful for developing any language skill. Chatbots are programmed around pattern matching, they search for key words to give responses. Practicing chatbots with advanced learners will be fun, but practicing AI(artificial intelligence) with learners having lower proficiency may be helpful a little. A learner can interact with it at home and bring interactions to the class for error corrections, or when a teacher wants to teach WH-questions to the learners, asking those questions to a chatbot can be assigned as a homework. Learners can also speak into a microphone to have conversations in the target language with chatbots.
SR technology's goal is to interpret all words that are intelligibly uttered by anyone. This kind of technology is beneficial for language learning in the areas of pronunciation, speaking and reading. CAPT systems(pronunciation) translate and recognize spoken utterances and even evaluate the pronunciation quickly giving a score or a feedback, which is valuable in case the learner's language is limited or when the teacher feedback is not available. Learners improve their pronunciation by repeating the target word, slowing down, saying the words correctly. SR requires Ss to talk, to respond and to interact with the program using words, phrases, and conversations thus enhancing speaking skills. SR gives Ss a safe environment to relax and practice. There is a Joy School English app built for EFL kids, which encourages speaking in every activity. In the Joy School English activity called "Do It Dotty" Ss give commands to Dotty the ladybug to perform some actions. SR has also been applied to EFL reading to help Ss with reading difficulties. There is a software "Reading Companion" which helps elementary school children's learning. However the SR technology has limitations. In the SR process there can be many errors and contextual misinterpretations as the technology deals with human speech, variety of its accents. Two types of errors can occur in this technology-false-positive, when the system determines that the learner has pronounced a sound or a word correctly, but it was incorrect, and false-negative, when the word or a sound was pronounced incorrectly, but in real it was interpreted as correct.
SR technology's goal is to interpret all words that are intelligibly uttered by anyone. This kind of technology is beneficial for language learning in the areas of pronunciation, speaking and reading. CAPT systems(pronunciation) translate and recognize spoken utterances and even evaluate the pronunciation quickly giving a score or a feedback, which is valuable in case the learner's language is limited or when the teacher feedback is not available. Learners improve their pronunciation by repeating the target word, slowing down, saying the words correctly. SR requires Ss to talk, to respond and to interact with the program using words, phrases, and conversations thus enhancing speaking skills. SR gives Ss a safe environment to relax and practice. There is a Joy School English app built for EFL kids, which encourages speaking in every activity. In the Joy School English activity called "Do It Dotty" Ss give commands to Dotty the ladybug to perform some actions. SR has also been applied to EFL reading to help Ss with reading difficulties. There is a software "Reading Companion" which helps elementary school children's learning. However the SR technology has limitations. In the SR process there can be many errors and contextual misinterpretations as the technology deals with human speech, variety of its accents. Two types of errors can occur in this technology-false-positive, when the system determines that the learner has pronounced a sound or a word correctly, but it was incorrect, and false-negative, when the word or a sound was pronounced incorrectly, but in real it was interpreted as correct.
That's right. SR can make mistakes. But also depends on how well the system has been trained to understand non-native speaker speech and interpret it. In a way, it's very much like human intelligence based on experience.
ReplyDelete