Skip to content

README.md

Latest commit

 

History

History
75 lines (60 loc) · 2.98 KB

README.md

File metadata and controls

75 lines (60 loc) · 2.98 KB

Call Classification Project

This project focuses on classifying customer support calls into predefined categories using unstructured call transcripts.

The source data consists of raw, unstructured transcripts. Your primary objectives are to parse these transcripts into a usable representation and design a model (or pipeline) that can classify each call into the specified categories.

Problem Tasks

  1. Transcript Parsing
    The call data is unstructured; you must parse each transcript into a structured form suitable for modeling (e.g., tokenized text, features, or embeddings).

  2. Model / System Design
    Develop or design a model that can categorize calls into the given categories. Depending on your chosen approach, you may:

    • Implement a working classification model (possibly using pre-trained NLP models), or\
    • Provide a clear system/architecture diagram that shows how you would perform the classification end-to-end.

    Notes:

    • A full, from-scratch model training and data labeling effort is not required.\
    • You are encouraged to use pre-trained models, existing NLP libraries, or other high-level methods to demonstrate a viable solution.

Output / Deliverables

  1. One of the following:
    • Model:
      A running model that performs the call classification task (it may be pre-trained). Include enough instructions/code for someone to run it end-to-end on the provided data.\
    • Diagram:
      A diagram describing how the model/system can be designed from start to end. This should clearly show data flow, preprocessing steps, modeling components, and outputs.
      • You may use any diagramming format (e.g., PowerPoint, draw.io, Visio, Canva, etc.).\
      • Ensure the image is shared with your submission, either as its own file (e.g., included in a PPT) or embedded in the notebook.
  2. Presentation (PPT)
    • A slide deck of up to 5 pages summarizing:
      • Your overall approach.
      • Modeling or system design decisions.
      • Any experiments or illustrative examples.
      • Key results or expected outcomes.
  3. Notebook Submission
    • Submit the notebook call_classifier.ipynb containing:
      • Your exploratory work (if any).
      • Parsing / preprocessing code.
      • Model code or pseudo-code / illustrative examples.
    • Make sure all relevant cells have been executed and outputs are visible.

Notes on Exporting / Sharing

Ensure that all outputs are saved in the notebook before exporting. You can export your notebook as HTML via Jupyter's "File → Download as → HTML" or similar workflow.