Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ Instructions for Experiment „Natural Language Processing“
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ2 The Experiment Procedure Topic: “Natural Language Processing (NLP)” Objective of the Experiment Gain insight about extracting and evaluating concepts and generating questions Informed consent Please fill in the informed consent
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ3 The Experiment Procedure Step 1: Learning the text Step 2: Questionnaire 1 Step 3: Learning Activity I (Questionnaire 2) Step 4: Test (Questionnaire 3) Step 5: Learning Activity II (Questionnaire 4) Step 6: Questionnaire 5a Please allow yourself breaks between tasks!
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ4 Please learn the text (max. 30 minutes) See (Subject: „NLP – text“) Step 1
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ5 Please fill in Questionnaire 1 (max. 10 min) See (Subject „phd Questionnaire 1“) Please give all answers in English! Please do not use the text while answering the questions! Click on the „Submit“-Button once you have finished the task! Step 2
Experiment „Natural Language Processing“ Step 3a: Extracting concepts Please extract the main concepts of the NLP-text, using Questionnaire 2 (max. 10 min) (Example: Speed of light; theory of relativity, Einstein). Please separate your concepts with semicolons (;) or commas (,) See (Subject „ phd Questionnaire 2“) After you have finished this step, please continue working on Questionnaire 2 (Step 3b) as described on the next slide Christian Gütl, IICM, TU-GRAZ6
Experiment „Natural Language Processing“ Step 3b: Generating questions Please generate eight questions from the NLP-text, using Questionnaire 2. Construct two Open Ended questions using the text (Example: What do you know about theory of light?) Construct two Fill In The Blank questions using the text (Example: The founder of this university was ________.) Construct two Single Choice questions using the text. Their answer should be “no”! (Example: Albert Einstein was born in Norway) Construct two Multiple Choice questions with three distractors for each chapter. Only one of the distractors should be correct! (Example: The sun is _____. blue, green, yellow, red) Note that you have to generate one question for each category first and then another one afterwards Christian Gütl, IICM, TU-GRAZ7
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ8 Step 4 Please fill in Questionnaire 3 (max 15 Minutes) and allow yourself a break afterwards. See (Subject „ phd Questionnaire 3“). Please do not use the text while answering the questions!
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ9 Step 5a: Evaluation of concepts Evaluation Task I (max. 10 min.) Please evaluate the concepts you will find in Questionnaire 4. Do you think the concept is relevant for the text? 1 = not relevant; 5 = very relevant See (Subject „ phd Questionnaire 4“). After you have finished this task, please continue working on Questionnaire 4 (Step 5b) as described on the next slide
Experiment „Natural Language Processing“ Christian Gütl, IICM, TU-GRAZ10 Step 5b: Evaluation of questions Evaluation Task II (max. 35 min) Please evaluate the questions provided in Questionnaire 4, using the criteria as described below. For each criterion described below, please assign a score between 1 ( = very bad) and 5 (very good) to each question. Pertinence: Relevancy of the formulated questions. Has the question a precise and logical connection to the topic? Terminology: Appropriateness of the words chosen to formulate the question. Level: Extent in approaching the topic. Is the question trivial or expresses it a significant meaning? (1 = trivial, easy; 5 = difficult, significant meaning) Answer (only if provided!): Is the provided answer relevant? Distractors (only for multiple choice!): Quality of the computed distractors of the multiple choice questions
Experiment „Natural Language Processing“ Step 5c General Survey Please fill in the last section of Questionnaire 4 (max. 15 min) Christian Gütl, IICM, TU-GRAZ11
Experiment „Natural Language Processing“ Step 6 Evaluating questions Christian Gütl, IICM, TU-GRAZ12 Evaluation Task III Please evaluate the questions provided in Questionnaire 5a (see ), using again the criteria as described below. For each criterion, please assign a score between 1 ( = very bad) and 5 (very good) to each question. Pertinence: Relevancy of the formulated questions. Has the question a precise and logical connection to the topic? Terminology: Appropriateness of the words chosen to formulate the question. Level: Extent in approaching the topic. Is the question trivial or expresses it a significant meaning? (1 = trivial, easy; 5 = difficult, significant meaning) Answer (only if provided!): Is the provided answer relevant? Distractors (only for multiple choice!): Quality of the computed distractors of the multiple choice questions
Experiment „Natural Language Processing“ Thank you for participating in this Experiment! Christian Gütl, IICM, TU-GRAZ13