nlp-lab/Project/naive_approach/README.md
2018-07-27 13:15:30 +00:00

29 lines
1.5 KiB
Markdown

# naive_approach
This directory contains the functions necessary to run the Naive Approach.
Prerequisites:
* the file [emoji_descriptions_preprocessed.csv](../Tools/emoji_descriptions_preprocessed.csv) has to be located in the specified folder [../Tools](../Tools)
* pandas has to be installed
For testing, import [naive_approach.py](naive_approach.py) and execute the following commands:
1. `prepareData(stem, lower)`
* preprocesses the emoji descriptions and returns a dictionary with the indexed emojis
* parameters:
* `stem`: Apply stemming (default=`True`)
* `lower`: Apply lowercasing (default=`True`)
2. `predict(sentence, lookup, emojis_to_consider, criteria, lang, embeddings, n=10, t=0.9)`
* evaluates an input sentence and returns a list of predicted emojis
* parameters:
* `sentence`: Input sentence (required parameter)
* `lookup`: dictionary with emoji data (return value of prepareData, required parameter)
* `emojis_to_consider`: set of emojis to include in prediction, or `"all"` (default=`"all"`)
* `criteria`: criteria to evaluate the values of the description - message matching.
* options: `"sum"`, `"mean"`, `"max_val"`, `"threshold"` (default: `"threshold"`)
* `lang`: language to use (default: "eng")
* `embeddings`: word embeddings
* options: `"wordnet"`, `"word2Vec"`, `"fastText"`, default: `"wordnet"`
* `n`: number of top ranked emojis to return (default=`10`)
* `t`: threshold for the `"threshold"` criteria (default=`0.9`)