m-aliabbas/idrak_ai_task

This repo contains the task assign by idark ai.

idrak_ai_task

This repo contains the task assign by idark ai. You can access the notebook from colab https://colab.research.google.com/drive/1r49W2dayUfltp7zQx1_4qY602IyLNn4F?usp=sharing or you can run locally idrak_ai_tasks (1).ipynb . The notebooks or python files contains all the details/documents and helps in comments and docs strings. Altough here we are sharing a brief details.

1. The question/answers text file will be process first. Empty and duplicate rows will be removed. Questions will be cleaned from unwanted symbols. ProcessQuestionDB class will be used for this purpose.

data_processor=ProcessQuestionsDB(input_df_path='S08_question_answer_pairs.txt',\
                output_df_path='idrak_ai_qa_pairs.csv',seprator='\t',question_col='Question',\
               answer_col='Answer')

2. Features will be extracted using FeaturesDBGenerator class

features_db_generator=FeaturesDBGenerator(dataset_path='idrak_ai_qa_pairs.csv',extractor_model_weight_path='FE.pt',output_db_path='qa_db.json')

3. Query Object will be initialize that contain user query. And Answer to this object will assign by QuestionReterival system.

query=Query('When did Political career of lincoln start?') #defining a query object

4. QuestionReterival class is responisble for reading features database,extract features from query question. And then match query features with database using cosine similarity.

qr=QuestionReterival(features_db_path='qa_db.json')

5. Searching the query answer.

qr.search_question(query)

6. Displaying the answer

query.get_answer()
query.get_similar_question()

or

print(query)

You can use gui_aap.py

Ussage:

Idraak_Tutorial.mp4 this video contains all the process

(Step 1) activate virtual envoirnment

(Step 2) Installing require modules

pip install -r requirements.txt

(Step 3) To extract features

python extract_feature.py

Expected output

--- Extracting Features ---
--- Features Extracted ---
--- Features are Stored to qa_db.json ---
--- Model is saved to FE.pt ---

(Step 4) Inference

For inference we made a GUI based application. This application is developed in Flat. Flat is a python based module , which create flutter like applications.

python gui_app.py

It will run a flutter based application in which you will pass the question and it give us answers.

Thanks