Skip to content

Ensemble methods for improving extractive summarization of legal case judgements

Notifications You must be signed in to change notification settings

naturenurtureelite/Legal-AI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Legal-AI

All the datasets have been obtained from [https://github.com/Law-AI/summarization]

We have reused the IN-Ext and IN-Abs dataset from [https://github.com/Law-AI/summarization]

All the datasets are available in dataset folder. In addition to the original documents and the expert written gold standard summaries we have additionally provided the outputs of the base algorithms in the dataset folder


The are 15 summarization algorithms namely Lexrank,LSA, Luhn, Reduction, DSDR, LetSum, KMM, DSDR, DELSumm, Case Summarizer, SummaRunner/Attn_RNN, SummaRunner/CNN_RNN, SummaRunner/RNN_RNN, BERTSUM, Chinese Gist

The supervised summarization algorithms are:- SummaRunner/Attn_RNN, SummaRunner/CNN_RNN, SummaRunner/RNN_RNN, BERTSUM, Chinese Gist

The unsupervised domain independent algorithms are:- Lexrank,LSA, Luhn, Reduction, DSDR

The unsupervised domain specific algorithms are:- LetSum, KMM, DSDR, DELSumm, Case Summarizer

The unsupervised algorithms are:- Lexrank,LSA, Luhn, Reduction, DSDR, LetSum, KMM, DSDR, DELSumm, Case Summarizer

The supervised algorithms are:-SummaRunner/Attn_RNN, SummaRunner/CNN_RNN, SummaRunner/RNN_RNN, BERTSUM, Chinese Gist

The top 4 summarization algorithms are:- BERTSUM, SummaRunner/Attn_RNN, SummaRunner/RNN_RNN and DELSumm

The various ensembling techniques tried are- Simple Voting, Borda Count, Reciprocal Ranking, Vecsim-Greedy-Length, Vecsim-Greedy-Degree, Vecsim-Roundrobin-Length, Vecsim-Roundrobin-degree, Pagerank, Hits authority, HITS hub, Node2vec, Deepwalk.


All the codes are available in codes folder


The folder IN-Ext summaries contains the summaries generated by the various methods. There are multiple methods namely Simple voting, Borda Count, Reciprocal Ranking, Vecsim-Greedy-Length, Vecsim-Greedy-degree, Roundrobin-length, Roundrobin-degree, Pagerank, Node2vec, HITS-authority, HITS-hub. Every folder has 6 sub-folders having the summaries generated by 6 different variations namely All methods, Supervised, unsupervised, unsupervised domain independent, unsupervised domain specific, Top 4.


The folder IN-Abs summaries contains the summaries generated by the various methods. There are multiple methods namely Simple voting, Borda Count, Reciprocal Ranking, Vecsim-Greedy-Length, Vecsim-Greedy-degree, Roundrobin-length, Roundrobin-degree, Pagerank, Node2vec, HITS-authority, HITS-hub. Every folder has 8 sub-folders having the summaries generated by 8 different variations namely All methods, Supervised, unsupervised, unsupervised domain independent, unsupervised domain specific,Top 2,Top 3,Top 4.