Links table
Abstract and 1 introduction
2. The background
2.1 Effective teaching practice
2.2 Reactions for teacher training
2.3 Signs sequence to generate comments
2.4 Great language models in education
3. The method
3.1 Data set and 3.2 sequence modes
3.3 GPT made signs of sequences
3.4 Measurements
4. Results
4.1 Results on RQ1
4.2 Results on RQ2
5. Discussion
6. Restrictions and future works
7. Conclusion
8. Thanks and appreciation
9. References
Excessive
Lesson principles
B. Inputs to set GPT-3.5
Jim is scattered from the relationship on the results based on the results
D. Detailed Results for the GPT-3.5 performance that has been seized
3.1 Data set
Our study has received IRB moral approval with the protocol number: Study2018 0000287 from Carnegie Mellon University. The study used a collection of data containing responses from 65 volunteer teachers who participated in the active praise lesson. The demographic collapse of these teachers was as follows: 52 % were eggs, 18 % of Asians, and 52 % of males, with more than half 50 years or larger. The goal of giving effective praise is to provide teachers with skills to enhance students’ motives by providing effective praise. We have collected 129 respondents from the teachers who completed the lesson, and these responses are sorted according to the type of praise (i.e. praise based on the effort, the praise -based praise, and the praise -based praise). It is worth noting that the data set contains only one of the person’s praise -based praise (“You are very smart”), which leads to excluding it from the analysis. Consequently, our study focused mainly on the analysis of praise -based praise and results.
3.2 Signs sequence
We aim to provide illustrative reactions that can highlight the components of the voltage -based and on -based praise of the teacher. Thus, we decided to use the signs sequence method. By doing this, we created clarification instructions in addition to specific examples of an effort and outcome Based on studies [55, 8, 7]. To carry out the explanatory comment, we have appointed two expert teachers who have completed a lesson for the first time Give effective praise From our platform and then they started commenting on the signs of praise that represent the features associated with an effort and outcomeFor 129 teacher responses.
In the pursuit of our understanding of effective praise in teaching dialogues, our study benefits INSIDE-S.UTSIDE (IOSigns mode system [31] In our study. the IO The scheme can capture the information necessary for our analysis, allowing us to maintain the focus on the basic aspects of praise in the teacher’s responses without the need to distinguish between the beginning or end of the entities, which correspond to our needs. the IO Scheme, characterized by its simplicity and efficiency, icon stickers as an inner mark (I) And an external sign (S.). The i mark is intended for praise components, while the sign is for the non -managed words. For example, when commenting on the components of praise for the teacher’s praise, “You are doing a great effort,” the phrase “great” and “effort” are determined as part of the voltage (i.e., an effortThe remaining text is determined in the response that it is the outside (i.e., “S.‘) One of the components of praise. By commenting on praise ingredients for each teacher’s response, we can get a list of symbols as shown in Figure 2.
When a reliability evaluation between the range of our study, we note that although Cohen Kapa is considered the record scale of the Inter-Anotatar Agreement for most of the tasks of the explanatory comments [44]The suitability of the tasks of applying serial marks – identification of the entity or similar tasks where the stickers are assigned to specific words or symbols in a sequence – limited [4, 16]. Specifically, a sequence mode may lead to partial agreements between broadcasters (for example, consensus on the type of symbolic stickers but not on the exact limits), which may not be taken effectively by Cuba Cohen because it does not represent a partial agreement [4]. In addition, in serial signs mode, a large percentage of symbols is usually classified as “S.(The distribution of symbolic posters appears in our study in Table 2), which leads to the distribution of an unbalanced poster. Given that Kapa Cohen undertakes an equal possibility in choosing each category, it may not provide a meaningful measure of agreement in situations in which the vast majority of stickers belong to one category, which makes the scale less beneficial or even misleading [4]. Looking at these restrictions, the F1 degree is often preferred to reliably evaluate the range in the tasks of setting serial signs as proposed in previous studies [4, 11]. Since Kappa from a distinctive symbol level can also provide some insight, we offer both Kappa and F1 to provide a comprehensive vision of the comments agreement in our study. Our results – 0.49 for Cohen’s Kappa and 0.79 F1 – are considered acceptable for our mission purposes as suggested [4, 15]. To address the contradictions between two conditions, a third expert was invited to resolve the contradictions. Distributing praise in our data set is as follows: 59 responses with voltage -based praise only, 22 with results based only, 31 contains both types, and 17 lacks signs either, which shows the various nature of praise within responses.
