SenseNet is a system that:

• aligns senses from different dictionaries but with the same meaning
• performs sense-based reverse dictionary
• computes similarity based on the combination of deep learning and dictionaries

Senset is the atomic unit of the system, and represents a set of word senses which have the same meaning. It has the following attributes:

• senset_id (string): unique ID of a Senset
• word (string): the corresponding word of a Senset
• pos_norm (string): the normalized POS tag of a Senset
• senses (list of Sense): the senses whiche belong to a Senset

Sense represents a word sense defined by a dictionary and has the following attributes:

• sense_id (string): unique ID of a Sense
• word (string): the corresponding word of a Sense
• pos (string): the raw POS tag of a Sense
• pos_norm (string): the normalized POS tag of a Senset
• source (string): the name of the source dictionary of a Sense
• definition (string): the definition in English of a Sense

The system currently has the following characteristics:

• Every Senses of a Senset are from different dictionary.
• The source of a Sense is either wordnet or cambridge.

The python version used is 3.8

First, create a new virtual environment (recommended) and install the dependencies

pip install -r requirements.txt

Second, download the data for the system.

Once we have done, the structure of the project should look like

├── scripts
├── src
├── data 
│   └── v0.1.0
│       └── wn_bi-camb
│           ├── sense_file.jsonl
│           ├── senset_file.jsonl
│           ├── sense_embeddings.txt
│           ├── senset_embeddings.txt
│           └── sense_embedder/
├── requirements.txt
├── README.md 
├── demo_max-sensenet.ipynb
├── demo_mean-sensenet.ipynb
└── .gitignore

Please refer to the demo_max-sensenet.ipynb and demo_mean-sensenet.ipynb.