Write tests to measure effectiveness of field detection logic

  • Status Encerrado
  • Budget $30 - $250 USD
  • Total Bids 2

Descrição do Projeto

I have a javascript function that given a PNG of a form will find fields that can be filled out by a user of the form.

I want to measure the false positive and false negative rate for this function.

To do this one can:

1. Collect a corpus of sample PDFs

2. Filter the corpus of sample PDFs to those that have field data defined in them

3. Extract field data from the remaining PDFs

4. Create a PNG of the pages in the remaining PDFs

5. Execute the field detection logic against the PNGs

6. Compare the embedded field data to the detected field data

7. Produce a report of the results

I would do this with two scripts, the first of which would collect of the corpus of PDFs to be used as test vectors.

This script would likely:

1. Search google for "filetype:pdf form" and download the corresponding files.

2. Extract field data. This can be done with [url removed, login to view]

3. Convert from a PDF to PNGs. This can be done with [url removed, login to view]

The second script would measure the effectiveness of the of our field detection using the results of the first.

This script would likely:

1. Enumerate through the results of the first script

2. Run the field detection logic against the PNGs from the first script

3. Look at the fields extracted in the first step comparing results from our field detection logic.

4. Produce a markdown table of the results.

The markdown table would look something like this:

| Filename | # Fields | # Fields Detected | False Positive | False Negative | Accuracy |


| [url removed, login to view] | 10 | 10 | 0 | 0 | 100% |

| [url removed, login to view] | 20 | 20 | 0 | 0 | 100% |


False Positive = The count of fields where we found a field that did not exist

False Negative = The count of fields where the form contained a field we did not find

Accuracy = The % of fields that were expected to be found that were actually found

Both scripts could use [url removed, login to view] for command line parsing.

I work only with individuals, no agencies please.

Obter orçamentos grátis para um projeto como este
Habilidades Requeridas

Pretende fazer algum dinheiro?

  • Definir seu orçamento e prazo
  • Esboce sua proposta
  • Seja pago por seu trabalho

Contratar freelancers que também ofereça neste projeto

    • Forbes
    • The New York Times
    • Time
    • Wall Street Journal
    • Times Online