Lyrics similarity data don't have computed similarity score and line number

[ccolonna]

Here the data file lyrics_similarity.json

The data file has at the moment null for this fields, a sample:

        {
            "lyrSimId": "lyr_sim_isophonics_135_isophonics_114_001_002",
            "compSimScore": null,
            "humanSimScore": null,
            "lineA": {
                "lineLabel": "Come on (Come on), Come on (Come on)",
                "recordingId": "isophonics_135",
                "lineNumber": "001",
                "recordingName": "Please Please Me",
                "artistName": "The Beatles"
            },
            "lineB": {
                "lineLabel": "So come, come on, come on",
                "recordingId": "isophonics_114",
                "lineNumber": "001",
                "recordingName": "Little Child",
                "artistName": "The Beatles"
            }
        },

See also this message by @andreamust explaining raw lyrics similarity data

I have just added the file containing the lyrics similarity to the json folder. The alignment of the lyrics has been a bit of a task, sorry for the delay. Some notes on the created file:

• it contains only the similarity between lyrics lines;
• the file is small because it only contains lyrics that have been annotated as "good" by human annotators. These annotations were made on a few samples so this is what we have;
• We have NO computed similarity or human annotated similarity at the moment (the json fields contain "null"); For the demo I assume that 1 or 2 particularly similar lines will be chosen by hand. The only change to be made will be on the interface side, i.e. removing the indication of the similarity value.

Does the demo interface accept lyrics annotation without the score ?

[ccolonna]

Also line numbers are fakes for this reason:

These ones are placeholders. The id of the similarity is composed by id_track_1 + id_track_2 + line_number_track_1 + line_number_track_2, so I had to invent some line number to make the id unique.

The interface should accept lyrics similarity annotation:

• without line numbers
• without computed similarity score