Academia.eduAcademia.edu

Figure 3 – uploaded by santhana megala

See full PDFdownloadDownload figure
The values a, b and c were the standard values of Low, Medium and High respectively and the values 1, m and n were the calculated values of Low, Medium and High respectively. Defining IF-Then rules is important in the Inference Engine. Sample IF — Then rules for the Inference Engine based on the Feature Extraction measures are mentioned below.

Figure 3 The values a, b and c were the standard values of Low, Medium and High respectively and the values 1, m and n were the calculated values of Low, Medium and High respectively. Defining IF-Then rules is important in the Inference Engine. Sample IF — Then rules for the Inference Engine based on the Feature Extraction measures are mentioned below.

Related Figures (16)

In previous researches, Probabilistic and the rule-based techniques were used for the text summarization, but Legal judgment summarization is a tedious process, and it is not easy to find out the important sentences like any other document. A single word that occurs only one time in the judgment may belong to an important one. Hence to obtain a good Judgment summarization, enhanced methods were needed. Figure 1 — Overall System Architecture of a Legal Judgement Summarization System
In previous researches, Probabilistic and the rule-based techniques were used for the text summarization, but Legal judgment summarization is a tedious process, and it is not easy to find out the important sentences like any other document. A single word that occurs only one time in the judgment may belong to an important one. Hence to obtain a good Judgment summarization, enhanced methods were needed. Figure 1 — Overall System Architecture of a Legal Judgement Summarization System
Figure: 2 — Architecture of the Text Summarization System based on Fuzzy Logic In the Defuzzification step, the output membership function step is divided into three membership functions, namely, "Unimportant", "Average" and "Important", which convert the result of the inference engine into a crisp output to obtain a final sentence score for each sentence.
Figure: 2 — Architecture of the Text Summarization System based on Fuzzy Logic In the Defuzzification step, the output membership function step is divided into three membership functions, namely, "Unimportant", "Average" and "Important", which convert the result of the inference engine into a crisp output to obtain a final sentence score for each sentence.
The most important benefits of CRFs over Hidden Markov Models i.e, HMMs is their conditional nature, ensuing in the relaxation of the independence assumptions required by HMMs in order to make a certain tractable inference. The architecture of the CRF approach, to generate a summary for the given legal judgment document was depicted in Figure 4.
The most important benefits of CRFs over Hidden Markov Models i.e, HMMs is their conditional nature, ensuing in the relaxation of the independence assumptions required by HMMs in order to make a certain tractable inference. The architecture of the CRF approach, to generate a summary for the given legal judgment document was depicted in Figure 4.
CRFs make a first-order Markov independence assumption with binary feature functions to link the output nodes of the graphical model in a linear chain by edges and thus can be understood as conditionally-trained finite state machines (FSMs) which are suitable for segmentation and sentence labeling [10].
CRFs make a first-order Markov independence assumption with binary feature functions to link the output nodes of the graphical model in a linear chain by edges and thus can be understood as conditionally-trained finite state machines (FSMs) which are suitable for segmentation and sentence labeling [10].
The Viterbi algorithm used to finding the shortest route through a graph is shown below:
The Viterbi algorithm used to finding the shortest route through a graph is shown below:
Let Ptr(k,t) be the function that returns the value of x used tc compute V,,, if t>1, or k if t=1. Then: Figure 5 - Keep tracking the likely sequence states using VA
Let Ptr(k,t) be the function that returns the value of x used tc compute V,,, if t>1, or k if t=1. Then: Figure 5 - Keep tracking the likely sequence states using VA
Algorithm : Sentence Score Generation The algorithm to find the sentence score for each sentence from the given judgment is shown below. Consider all the sentences S,,r€ {1,...,R} in the documents and all the Topics Tj,.j€ {1.,...,.K} and then by calculating the probability of the Sentence S, for the given the Topic T; i. P(S, |T;). Thus calculating the probability for the sentence S, belongs or represents the topic T; . Let the words of the sentence S, be {W),W5, ... Wg}. Algorithm to find sentence score for each sentence based on each topic for the entire corpus was given below:
Algorithm : Sentence Score Generation The algorithm to find the sentence score for each sentence from the given judgment is shown below. Consider all the sentences S,,r€ {1,...,R} in the documents and all the Topics Tj,.j€ {1.,...,.K} and then by calculating the probability of the Sentence S, for the given the Topic T; i. P(S, |T;). Thus calculating the probability for the sentence S, belongs or represents the topic T; . Let the words of the sentence S, be {W),W5, ... Wg}. Algorithm to find sentence score for each sentence based on each topic for the entire corpus was given below:
Figure 6 - Architecture to generate Legal Judgment summary using LDA After the pre-processing all the sentences present in the document were sent to LDA as bag of words and the outcome of the process is some different topics, based on the probabilistic model. Now a set of topics for the given corpus using LDA topic model is derived. Consider each judgment from the corpus and find the sentences present in the document using Sentence Boundary method.
Figure 6 - Architecture to generate Legal Judgment summary using LDA After the pre-processing all the sentences present in the document were sent to LDA as bag of words and the outcome of the process is some different topics, based on the probabilistic model. Now a set of topics for the given corpus using LDA topic model is derived. Consider each judgment from the corpus and find the sentences present in the document using Sentence Boundary method.
A rich set of features were included in this paper to identify the rhetorical roles present in the Legal judgment. International Journal of Computer Sciences and Engineering
A rich set of features were included in this paper to identify the rhetorical roles present in the Legal judgment. International Journal of Computer Sciences and Engineering
Table: 3- Precision, Recall & F- Measure Value for the Seven Segments using CRF& LDA
Table: 3- Precision, Recall & F- Measure Value for the Seven Segments using CRF& LDA
Table 4 — T- Test table for the Null Hypothesis HO based on the average F-Measures obtained for CRF & LDA. Based on the above results it clearly shows that the Null Hypothesis that stated was rejected because t value calculated is greater than the t critical value ie., 7.039 > 2.306. Hence the p-value obtained for the calculated t score is 0.00054, which is less than 0.01, ie., 0.00054<0.01. Therefore the result is significant at p<0.01. The hypothesis In this paper, the average F-Measures of the sample Legal Documents were taken as the performance measures for the statistical test. A Null Hypothesis HO was set by stating that there is no difference between the results generated by the Conditional Random Field & Latent Dirichlet allocation. On the other hand, an Alternative Hypothesis H1 indicating that there is a difference between the results generated by the Conditional Random Field & Latent Dirichlet allocation. Based on the Statistical Paired t-test results denoted in Table 4, it clearly states that CRF method provides results better than the LDA Method.
Table 4 — T- Test table for the Null Hypothesis HO based on the average F-Measures obtained for CRF & LDA. Based on the above results it clearly shows that the Null Hypothesis that stated was rejected because t value calculated is greater than the t critical value ie., 7.039 > 2.306. Hence the p-value obtained for the calculated t score is 0.00054, which is less than 0.01, ie., 0.00054<0.01. Therefore the result is significant at p<0.01. The hypothesis In this paper, the average F-Measures of the sample Legal Documents were taken as the performance measures for the statistical test. A Null Hypothesis HO was set by stating that there is no difference between the results generated by the Conditional Random Field & Latent Dirichlet allocation. On the other hand, an Alternative Hypothesis H1 indicating that there is a difference between the results generated by the Conditional Random Field & Latent Dirichlet allocation. Based on the Statistical Paired t-test results denoted in Table 4, it clearly states that CRF method provides results better than the LDA Method.
Figure: 7 - Graphical representations of F-Measure value for CRF & LDA
Figure: 7 - Graphical representations of F-Measure value for CRF & LDA
Figure: 8- ROC Curves for Fuzzy Logic & Online Summarizer Table: 5 - Comparison between online summarizer and fuzzy logic based on TP & FP Rate ROC analysis is related in a direct and natural way to cost/benefit analysis of diagnostic decision making. ROC Curve for our system grows on the left top border, which shows a good accuracy.
Figure: 8- ROC Curves for Fuzzy Logic & Online Summarizer Table: 5 - Comparison between online summarizer and fuzzy logic based on TP & FP Rate ROC analysis is related in a direct and natural way to cost/benefit analysis of diagnostic decision making. ROC Curve for our system grows on the left top border, which shows a good accuracy.
Figure: 10 — Sample Output for the Structured Summary using Classification Technique
Figure: 10 — Sample Output for the Structured Summary using Classification Technique
Algorithm based on the Rhetorical Roles present in the Legal Judgment document. Figure: 9 — Sample Output for Judgement Summarization using Fuzzy Logic
Algorithm based on the Rhetorical Roles present in the Legal Judgment document. Figure: 9 — Sample Output for Judgement Summarization using Fuzzy Logic
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved