EMG Sage User Manual

1 Introduction

The application provides some services to facilitate English self-learning. An arbitrary text to be studied can be registered via either direct typing, capturing with a smartphone camera or loading from a PDF file. The application can then automatically generate several language exercises based on that input text:

1.    Vocabulary exercises: explaining words occurring in the text. Users can listen to the word-pronunciation as well as practice the pronunciation. The application can return immediately the evaluation on the pronunciation precision.

2.    Grammar exercises: resuming some main grammar key-points found in the text for users.

3.    Reading comprehension: generating some questions and correct answers for users to check their understanding of the textual content.

4.    Conversation exercises: performing the vocal dialogue between application and users to verify their understanding as well as their pronunciation while “talking” to the application to answer the vocal questions. The application can provide immediate feedback on the correctness of the auditive responses.

Besides text analysis, the application provides a basic chat service for users. Users can ask the application for any information. This information and its correctness are determined by the Open AI 3.5 engine. Two ways of chat can be realized: via either text as convention or voice with a microphone.

2 How to register a text for analysis

An arbitrary text such as a journal article, a paragraph of a book can be registered for generating associated exercises via three ways:

1.    Typing directly (copy / paste between several applications is also possible) using the virtual keyboard of a smartphone,

2.    Taking a photo of the text via the smartphone’s camera,

3.    Loading the content directly from its stored format PDF.

To do so, the Text analysis button must be clicked (Figure 1). A new window with upper and lower text areas are shown then as in Figure 2. The upper area is served for storing the textual content to be analyzed. There are several ways to insert a textual content into this area:

1.    Text can be typed directly into this window by clicking the Edit text button (Figure 2),

2.    Text can be pasted (it is previously copied in other application) to this window via clicking the Paste text button,

3.    Text can be captured via camera by clicking the Camera input button. The taken image is then converted into text thanks to the so-called OCR (Optical Character Recognition) technique. In Figure 3, several sources of the image are supported in the current version of this Application: a picture can be taken directly from the smartphone’s camera; a picture from the album on the smartphone or a picture previously analyzed by the Application. A content in the format PDF can be also loaded.

3 How to retrieve the auto-generated exercises

After registering a text to the Application as described in the previous section, the Application offers several auto-generated English exercises corresponding to the input textual content.

3.1 Vocabulary exercises

On clicking the Vocabulary button (Figure 4), the Application will generate a list of words occurring in the content together with their definition. On the right-hand side of the word, the Pronunciation button is found. While being clicked, the button will activate the window as in Figure 5. In this Graphic User Interface (GUI), the Vietnamese meaning of the word can be seen. Its reference pronunciation can be replayed while clicking the symbol Speaker. On clicking and keeping pressed the symbol Microphone, user can record his pronunciation of the associated word. The app can then evaluate the correctness of this evaluation thanks to an AI-based auto-verification. The result is then visualized as in the Figure 6.

3.2 Grammar exercises

On clicking the Grammar button (Figure 7), the Application will automatically resume some grammar key-points happening in the text.

3.3 Reading comprehension exercises

The application provides several ways to check the reading comprehension of the user as the following:

·         The application automatically generates a series of questions related to the content of the text. Correct answers are also provided for users’ self-evaluation. This mode is activated whenever the button Question&Answer is clicked (Figure 7).

·         The application automatically generates a series of questions of type “Fill in the blank” related to the content of the text. Correct answers are also provided for users’ self-evaluation. This mode is activated whenever the button Fill in the blank is clicked (Figure 8).

 

3.4 Conversation exercises

The current version of the Application can offer a dialogue feature between users and the Application, simulating as if the user talked to a (virtual) assistant. This mode is activated when the button Question is clicked (Figure 9).

According the content of the registered text, the application can speak up vocally a question. The user can then answer that question also in a vocal manner. The application will evaluate the response. The result will be pronounced to the user. The process will be repeated (ten times) for the next questions.

To relisten to a question, click on the associated speaker symbol (Figure 9)

To say loudly the answer, click and keep pressing the associated microphone symbol.

3.5 User-defined exercises

Besides the pre-created exercises as described in the previous sections, the current release of the Application can also provide a personalized version of exercises. On clicking the symbol ...  (Figure 9), a window as in Figure 10 will show up. The prompt syntax of each exercise can be modified in this GUI. The prompt here implies the pre-programmed questions requested to the AI engine to create the information for users.

A new, user-defined exercise can also be defined. Clicking the Add new button on the Figure 10, the new window as in Figure 11 is shown. A title of the new exercise and the syntax of the prompt can be registered. Figure 11 illustrates a customized prompt user want to request: for the registered text, search for 5 word of type nouns. Note that the term <paragraph> in Figure 11 is a placeholder for the input text so that the Application can understand the content of the registered text as a context for the prompt.

 

Another example of defining a customized analysis cn be the following: to request a summary of an arbitrary text, a following prompt can be assigned to a new button: Acting as an English teacher, create a summary of the following text <paragraph>.

More complicated and more comprehensive requests can also be formulated like that to take the advantage of the natural language processing of the AI engine to generate more interesting information in accordance with the registered text.

4 How to retrieve information via chat

Users can ask the application for any kind of information. The questions can be raised to the application via textual or vocal manner. Textual chat is activated whenever the button Textual chat is clicked (Figure 1). A new window will be then shown to facilitate the text typing and to show the textual answer of the Application (Figure 12).

In the vocal mode, user should keep pressed the button Vocal chat (Figure 1) while speaking out clearly a question. A similar window as in the textual mode will appear to show the textual as well as to play the voice as the answer to the raised question.