/Meaningly

NLP text searching tool that returns the most similar sentences

Primary LanguagePythonMIT LicenseMIT

Note: This project was built as a tool for a McGill University research study to provide better transcript searchability.

Meaningly

A NLP search tool for querying coded research study transcripts and finding similar sentences. API + GUI built on top of the Universal Sentence Encoder that lets the user enter sentences and find similar ones in a transcript. The user can add a custom threshold (between 0 and 1) which represents the percentage above which the text sentences must be correlated to the input.

GUI

Inputs

File

The path to the document with the transcript. Must be a word doc

List of phrases

List of sentences to search the text for. Include all punctuation and start each new sentence one a new line.

Threshold

When given a threshold, the program will only display sentences that are correlated to the input phrases above the given value. Default is 0. Must be a decimal value between 0 and 1.

Symbol

Not Implemented yet. Symbol after which the actual text/dialogue starts. Useful for parsing out names of speakers, etc. Currently the program begins reading the sentence after a ":" (colon).

Example:

GUI Filled

Results